diff --git a/go.mod b/go.mod index f302528c4..3c3572d36 100644 --- a/go.mod +++ b/go.mod @@ -12,7 +12,7 @@ require ( github.com/jmoiron/sqlx v1.3.5 github.com/kisielk/sqlstruct v0.0.0-20201105191214-5f3e10d3ab46 github.com/lib/pq v1.10.9 - github.com/mattn/go-sqlite3 v1.14.19 + github.com/mattn/go-sqlite3 v1.14.20 github.com/prometheus/client_golang v1.18.0 github.com/zmap/zcrypto v0.0.0-20230310154051-c8b263fd8300 github.com/zmap/zlint/v3 v3.5.0 diff --git a/go.sum b/go.sum index f861e0df0..38e87ba5b 100644 --- a/go.sum +++ b/go.sum @@ -236,8 +236,8 @@ github.com/mattn/go-isatty v0.0.7/go.mod h1:Iq45c/XA43vh69/j3iqttzPXn0bhXyGjM0Hd github.com/mattn/go-isatty v0.0.8/go.mod h1:Iq45c/XA43vh69/j3iqttzPXn0bhXyGjM0Hdxcsrc5s= github.com/mattn/go-isatty v0.0.9/go.mod h1:YNRxwqDuOph6SZLI9vUUz6OYw3QyUt7WiY2yME+cCiQ= github.com/mattn/go-sqlite3 v1.14.6/go.mod h1:NyWgC/yNuGj7Q9rpYnZvas74GogHl5/Z4A/KQRfk6bU= -github.com/mattn/go-sqlite3 v1.14.19 h1:fhGleo2h1p8tVChob4I9HpmVFIAkKGpiukdrgQbWfGI= -github.com/mattn/go-sqlite3 v1.14.19/go.mod h1:2eHXhiwb8IkHr+BDWZGa96P6+rkvnG63S2DGjv9HUNg= +github.com/mattn/go-sqlite3 v1.14.20 h1:BAZ50Ns0OFBNxdAqFhbZqdPcht1Xlb16pDCqkq1spr0= +github.com/mattn/go-sqlite3 v1.14.20/go.mod h1:Uh1q+B4BYcTPb+yiD3kU8Ct7aC0hY9fxUwlHK0RXw+Y= github.com/mattn/goveralls v0.0.2/go.mod h1:8d1ZMHsd7fW6IRPKQh46F2WRpyib5/X4FOpevwGNQEw= github.com/matttproud/golang_protobuf_extensions v1.0.1/go.mod h1:D8He9yQNgCq6Z5Ld7szi9bcBfOoFv/3dc6xSMkL2PC0= github.com/matttproud/golang_protobuf_extensions/v2 v2.0.0 h1:jWpvCLoY8Z/e3VKvlsiIGKtc+UG6U5vzxaoagmhXfyg= diff --git a/vendor/github.com/mattn/go-sqlite3/callback.go b/vendor/github.com/mattn/go-sqlite3/callback.go index d30569100..b794bcd83 100644 --- a/vendor/github.com/mattn/go-sqlite3/callback.go +++ b/vendor/github.com/mattn/go-sqlite3/callback.go @@ -100,13 +100,13 @@ func preUpdateHookTrampoline(handle unsafe.Pointer, dbHandle uintptr, op int, db // Use handles to avoid passing Go pointers to C. type handleVal struct { db *SQLiteConn - val interface{} + val any } var handleLock sync.Mutex var handleVals = make(map[unsafe.Pointer]handleVal) -func newHandle(db *SQLiteConn, v interface{}) unsafe.Pointer { +func newHandle(db *SQLiteConn, v any) unsafe.Pointer { handleLock.Lock() defer handleLock.Unlock() val := handleVal{db: db, val: v} @@ -124,7 +124,7 @@ func lookupHandleVal(handle unsafe.Pointer) handleVal { return handleVals[handle] } -func lookupHandle(handle unsafe.Pointer) interface{} { +func lookupHandle(handle unsafe.Pointer) any { return lookupHandleVal(handle).val } @@ -238,7 +238,7 @@ func callbackArg(typ reflect.Type) (callbackArgConverter, error) { switch typ.Kind() { case reflect.Interface: if typ.NumMethod() != 0 { - return nil, errors.New("the only supported interface type is interface{}") + return nil, errors.New("the only supported interface type is any") } return callbackArgGeneric, nil case reflect.Slice: @@ -360,11 +360,11 @@ func callbackRetGeneric(ctx *C.sqlite3_context, v reflect.Value) error { } cb, err := callbackRet(v.Elem().Type()) - if err != nil { - return err - } + if err != nil { + return err + } - return cb(ctx, v.Elem()) + return cb(ctx, v.Elem()) } func callbackRet(typ reflect.Type) (callbackRetConverter, error) { diff --git a/vendor/github.com/mattn/go-sqlite3/convert.go b/vendor/github.com/mattn/go-sqlite3/convert.go index 03850737f..f7a9dcd72 100644 --- a/vendor/github.com/mattn/go-sqlite3/convert.go +++ b/vendor/github.com/mattn/go-sqlite3/convert.go @@ -23,7 +23,7 @@ var errNilPtr = errors.New("destination pointer is nil") // embedded in descript // convertAssign copies to dest the value in src, converting it if possible. // An error is returned if the copy would result in loss of information. // dest should be a pointer type. -func convertAssign(dest, src interface{}) error { +func convertAssign(dest, src any) error { // Common cases, without reflect. switch s := src.(type) { case string: @@ -55,7 +55,7 @@ func convertAssign(dest, src interface{}) error { } *d = string(s) return nil - case *interface{}: + case *any: if d == nil { return errNilPtr } @@ -97,7 +97,7 @@ func convertAssign(dest, src interface{}) error { } case nil: switch d := dest.(type) { - case *interface{}: + case *any: if d == nil { return errNilPtr } @@ -149,7 +149,7 @@ func convertAssign(dest, src interface{}) error { *d = bv.(bool) } return err - case *interface{}: + case *any: *d = src return nil } @@ -256,7 +256,7 @@ func cloneBytes(b []byte) []byte { return c } -func asString(src interface{}) string { +func asString(src any) string { switch v := src.(type) { case string: return v diff --git a/vendor/github.com/mattn/go-sqlite3/doc.go b/vendor/github.com/mattn/go-sqlite3/doc.go index ac27633b3..a3bcebbcb 100644 --- a/vendor/github.com/mattn/go-sqlite3/doc.go +++ b/vendor/github.com/mattn/go-sqlite3/doc.go @@ -5,63 +5,63 @@ This works as a driver for database/sql. Installation - go get github.com/mattn/go-sqlite3 + go get github.com/mattn/go-sqlite3 -Supported Types +# Supported Types Currently, go-sqlite3 supports the following data types. - +------------------------------+ - |go | sqlite3 | - |----------|-------------------| - |nil | null | - |int | integer | - |int64 | integer | - |float64 | float | - |bool | integer | - |[]byte | blob | - |string | text | - |time.Time | timestamp/datetime| - +------------------------------+ - -SQLite3 Extension + +------------------------------+ + |go | sqlite3 | + |----------|-------------------| + |nil | null | + |int | integer | + |int64 | integer | + |float64 | float | + |bool | integer | + |[]byte | blob | + |string | text | + |time.Time | timestamp/datetime| + +------------------------------+ + +# SQLite3 Extension You can write your own extension module for sqlite3. For example, below is an extension for a Regexp matcher operation. - #include - #include - #include - #include - - SQLITE_EXTENSION_INIT1 - static void regexp_func(sqlite3_context *context, int argc, sqlite3_value **argv) { - if (argc >= 2) { - const char *target = (const char *)sqlite3_value_text(argv[1]); - const char *pattern = (const char *)sqlite3_value_text(argv[0]); - const char* errstr = NULL; - int erroff = 0; - int vec[500]; - int n, rc; - pcre* re = pcre_compile(pattern, 0, &errstr, &erroff, NULL); - rc = pcre_exec(re, NULL, target, strlen(target), 0, 0, vec, 500); - if (rc <= 0) { - sqlite3_result_error(context, errstr, 0); - return; - } - sqlite3_result_int(context, 1); - } - } - - #ifdef _WIN32 - __declspec(dllexport) - #endif - int sqlite3_extension_init(sqlite3 *db, char **errmsg, - const sqlite3_api_routines *api) { - SQLITE_EXTENSION_INIT2(api); - return sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, - (void*)db, regexp_func, NULL, NULL); - } + #include + #include + #include + #include + + SQLITE_EXTENSION_INIT1 + static void regexp_func(sqlite3_context *context, int argc, sqlite3_value **argv) { + if (argc >= 2) { + const char *target = (const char *)sqlite3_value_text(argv[1]); + const char *pattern = (const char *)sqlite3_value_text(argv[0]); + const char* errstr = NULL; + int erroff = 0; + int vec[500]; + int n, rc; + pcre* re = pcre_compile(pattern, 0, &errstr, &erroff, NULL); + rc = pcre_exec(re, NULL, target, strlen(target), 0, 0, vec, 500); + if (rc <= 0) { + sqlite3_result_error(context, errstr, 0); + return; + } + sqlite3_result_int(context, 1); + } + } + + #ifdef _WIN32 + __declspec(dllexport) + #endif + int sqlite3_extension_init(sqlite3 *db, char **errmsg, + const sqlite3_api_routines *api) { + SQLITE_EXTENSION_INIT2(api); + return sqlite3_create_function(db, "regexp", 2, SQLITE_UTF8, + (void*)db, regexp_func, NULL, NULL); + } It needs to be built as a so/dll shared library. And you need to register the extension module like below. @@ -77,7 +77,7 @@ Then, you can use this extension. rows, err := db.Query("select text from mytable where name regexp '^golang'") -Connection Hook +# Connection Hook You can hook and inject your code when the connection is established by setting ConnectHook to get the SQLiteConn. @@ -95,13 +95,13 @@ You can also use database/sql.Conn.Raw (Go >= 1.13): conn, err := db.Conn(context.Background()) // if err != nil { ... } defer conn.Close() - err = conn.Raw(func (driverConn interface{}) error { + err = conn.Raw(func (driverConn any) error { sqliteConn := driverConn.(*sqlite3.SQLiteConn) // ... use sqliteConn }) // if err != nil { ... } -Go SQlite3 Extensions +# Go SQlite3 Extensions If you want to register Go functions as SQLite extension functions you can make a custom driver by calling RegisterFunction from @@ -130,6 +130,5 @@ You can then use the custom driver by passing its name to sql.Open. } See the documentation of RegisterFunc for more details. - */ package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.c b/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.c index 454512ad2..7bf32a69d 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.c +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.c @@ -1,7 +1,7 @@ #ifndef USE_LIBSQLITE3 /****************************************************************************** ** This file is an amalgamation of many separate C source files from SQLite -** version 3.44.0. By combining all the individual C code files into this +** version 3.45.0. By combining all the individual C code files into this ** single large file, the entire code can be compiled as a single translation ** unit. This allows many compilers to do optimizations that would not be ** possible if the files were compiled separately. Performance improvements @@ -19,7 +19,7 @@ ** separate file. This file contains only code for the core SQLite library. ** ** The content in this amalgamation comes from Fossil check-in -** 17129ba1ff7f0daf37100ee82d507aef7827. +** 1066602b2b1976fe58b5150777cced894af1. */ #define SQLITE_CORE 1 #define SQLITE_AMALGAMATION 1 @@ -460,9 +460,9 @@ extern "C" { ** [sqlite3_libversion_number()], [sqlite3_sourceid()], ** [sqlite_version()] and [sqlite_source_id()]. */ -#define SQLITE_VERSION "3.44.0" -#define SQLITE_VERSION_NUMBER 3044000 -#define SQLITE_SOURCE_ID "2023-11-01 11:23:50 17129ba1ff7f0daf37100ee82d507aef7827cf38de1866e2633096ae6ad81301" +#define SQLITE_VERSION "3.45.0" +#define SQLITE_VERSION_NUMBER 3045000 +#define SQLITE_SOURCE_ID "2024-01-15 17:01:13 1066602b2b1976fe58b5150777cced894af17c803e068f5918390d6915b46e1d" /* ** CAPI3REF: Run-Time Library Version Numbers @@ -4268,15 +4268,17 @@ SQLITE_API void sqlite3_free_filename(sqlite3_filename); ** ** ** ^The sqlite3_errmsg() and sqlite3_errmsg16() return English-language -** text that describes the error, as either UTF-8 or UTF-16 respectively. +** text that describes the error, as either UTF-8 or UTF-16 respectively, +** or NULL if no error message is available. ** (See how SQLite handles [invalid UTF] for exceptions to this rule.) ** ^(Memory to hold the error message string is managed internally. ** The application does not need to worry about freeing the result. ** However, the error string might be overwritten or deallocated by ** subsequent calls to other SQLite interface functions.)^ ** -** ^The sqlite3_errstr() interface returns the English-language text -** that describes the [result code], as UTF-8. +** ^The sqlite3_errstr(E) interface returns the English-language text +** that describes the [result code] E, as UTF-8, or NULL if E is not an +** result code for which a text error message is available. ** ^(Memory to hold the error message string is managed internally ** and must not be freed by the application)^. ** @@ -5887,13 +5889,27 @@ SQLITE_API int sqlite3_create_window_function( ** ** ** [[SQLITE_SUBTYPE]]
SQLITE_SUBTYPE
-** The SQLITE_SUBTYPE flag indicates to SQLite that a function may call +** The SQLITE_SUBTYPE flag indicates to SQLite that a function might call ** [sqlite3_value_subtype()] to inspect the sub-types of its arguments. -** Specifying this flag makes no difference for scalar or aggregate user -** functions. However, if it is not specified for a user-defined window -** function, then any sub-types belonging to arguments passed to the window -** function may be discarded before the window function is called (i.e. -** sqlite3_value_subtype() will always return 0). +** This flag instructs SQLite to omit some corner-case optimizations that +** might disrupt the operation of the [sqlite3_value_subtype()] function, +** causing it to return zero rather than the correct subtype(). +** SQL functions that invokes [sqlite3_value_subtype()] should have this +** property. If the SQLITE_SUBTYPE property is omitted, then the return +** value from [sqlite3_value_subtype()] might sometimes be zero even though +** a non-zero subtype was specified by the function argument expression. +** +** [[SQLITE_RESULT_SUBTYPE]]
SQLITE_RESULT_SUBTYPE
+** The SQLITE_RESULT_SUBTYPE flag indicates to SQLite that a function might call +** [sqlite3_result_subtype()] to cause a sub-type to be associated with its +** result. +** Every function that invokes [sqlite3_result_subtype()] should have this +** property. If it does not, then the call to [sqlite3_result_subtype()] +** might become a no-op if the function is used as term in an +** [expression index]. On the other hand, SQL functions that never invoke +** [sqlite3_result_subtype()] should avoid setting this property, as the +** purpose of this property is to disable certain optimizations that are +** incompatible with subtypes. **
** */ @@ -5901,6 +5917,7 @@ SQLITE_API int sqlite3_create_window_function( #define SQLITE_DIRECTONLY 0x000080000 #define SQLITE_SUBTYPE 0x000100000 #define SQLITE_INNOCUOUS 0x000200000 +#define SQLITE_RESULT_SUBTYPE 0x001000000 /* ** CAPI3REF: Deprecated Functions @@ -6097,6 +6114,12 @@ SQLITE_API int sqlite3_value_encoding(sqlite3_value*); ** information can be used to pass a limited amount of context from ** one SQL function to another. Use the [sqlite3_result_subtype()] ** routine to set the subtype for the return value of an SQL function. +** +** Every [application-defined SQL function] that invoke this interface +** should include the [SQLITE_SUBTYPE] property in the text +** encoding argument when the function is [sqlite3_create_function|registered]. +** If the [SQLITE_SUBTYPE] property is omitted, then sqlite3_value_subtype() +** might return zero instead of the upstream subtype in some corner cases. */ SQLITE_API unsigned int sqlite3_value_subtype(sqlite3_value*); @@ -6227,14 +6250,22 @@ SQLITE_API sqlite3 *sqlite3_context_db_handle(sqlite3_context*); **
  • ^(when sqlite3_set_auxdata() is invoked again on the same ** parameter)^, or **
  • ^(during the original sqlite3_set_auxdata() call when a memory -** allocation error occurs.)^ +** allocation error occurs.)^ +**
  • ^(during the original sqlite3_set_auxdata() call if the function +** is evaluated during query planning instead of during query execution, +** as sometimes happens with [SQLITE_ENABLE_STAT4].)^ ** -** Note the last bullet in particular. The destructor X in +** Note the last two bullets in particular. The destructor X in ** sqlite3_set_auxdata(C,N,P,X) might be called immediately, before the ** sqlite3_set_auxdata() interface even returns. Hence sqlite3_set_auxdata() ** should be called near the end of the function implementation and the ** function implementation should not make any use of P after -** sqlite3_set_auxdata() has been called. +** sqlite3_set_auxdata() has been called. Furthermore, a call to +** sqlite3_get_auxdata() that occurs immediately after a corresponding call +** to sqlite3_set_auxdata() might still return NULL if an out-of-memory +** condition occurred during the sqlite3_set_auxdata() call or if the +** function is being evaluated during query planning rather than during +** query execution. ** ** ^(In practice, auxiliary data is preserved between function calls for ** function parameters that are compile-time constants, including literal @@ -6508,6 +6539,20 @@ SQLITE_API int sqlite3_result_zeroblob64(sqlite3_context*, sqlite3_uint64 n); ** higher order bits are discarded. ** The number of subtype bytes preserved by SQLite might increase ** in future releases of SQLite. +** +** Every [application-defined SQL function] that invokes this interface +** should include the [SQLITE_RESULT_SUBTYPE] property in its +** text encoding argument when the SQL function is +** [sqlite3_create_function|registered]. If the [SQLITE_RESULT_SUBTYPE] +** property is omitted from the function that invokes sqlite3_result_subtype(), +** then in some cases the sqlite3_result_subtype() might fail to set +** the result subtype. +** +** If SQLite is compiled with -DSQLITE_STRICT_SUBTYPE=1, then any +** SQL function that invokes the sqlite3_result_subtype() interface +** and that does not have the SQLITE_RESULT_SUBTYPE property will raise +** an error. Future versions of SQLite might enable -DSQLITE_STRICT_SUBTYPE=1 +** by default. */ SQLITE_API void sqlite3_result_subtype(sqlite3_context*,unsigned int); @@ -8308,9 +8353,11 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*); ** ** ^(Some systems (for example, Windows 95) do not support the operation ** implemented by sqlite3_mutex_try(). On those systems, sqlite3_mutex_try() -** will always return SQLITE_BUSY. The SQLite core only ever uses -** sqlite3_mutex_try() as an optimization so this is acceptable -** behavior.)^ +** will always return SQLITE_BUSY. In most cases the SQLite core only uses +** sqlite3_mutex_try() as an optimization, so this is acceptable +** behavior. The exceptions are unix builds that set the +** SQLITE_ENABLE_SETLK_TIMEOUT build option. In that case a working +** sqlite3_mutex_try() is required.)^ ** ** ^The sqlite3_mutex_leave() routine exits a mutex that was ** previously entered by the same thread. The behavior @@ -8569,6 +8616,7 @@ SQLITE_API int sqlite3_test_control(int op, ...); #define SQLITE_TESTCTRL_ASSERT 12 #define SQLITE_TESTCTRL_ALWAYS 13 #define SQLITE_TESTCTRL_RESERVE 14 /* NOT USED */ +#define SQLITE_TESTCTRL_JSON_SELFCHECK 14 #define SQLITE_TESTCTRL_OPTIMIZATIONS 15 #define SQLITE_TESTCTRL_ISKEYWORD 16 /* NOT USED */ #define SQLITE_TESTCTRL_SCRATCHMALLOC 17 /* NOT USED */ @@ -13082,8 +13130,11 @@ struct Fts5PhraseIter { ** created with the "columnsize=0" option. ** ** xColumnText: -** This function attempts to retrieve the text of column iCol of the -** current document. If successful, (*pz) is set to point to a buffer +** If parameter iCol is less than zero, or greater than or equal to the +** number of columns in the table, SQLITE_RANGE is returned. +** +** Otherwise, this function attempts to retrieve the text of column iCol of +** the current document. If successful, (*pz) is set to point to a buffer ** containing the text in utf-8 encoding, (*pn) is set to the size in bytes ** (not characters) of the buffer and SQLITE_OK is returned. Otherwise, ** if an error occurs, an SQLite error code is returned and the final values @@ -13093,8 +13144,10 @@ struct Fts5PhraseIter { ** Returns the number of phrases in the current query expression. ** ** xPhraseSize: -** Returns the number of tokens in phrase iPhrase of the query. Phrases -** are numbered starting from zero. +** If parameter iCol is less than zero, or greater than or equal to the +** number of phrases in the current query, as returned by xPhraseCount, +** 0 is returned. Otherwise, this function returns the number of tokens in +** phrase iPhrase of the query. Phrases are numbered starting from zero. ** ** xInstCount: ** Set *pnInst to the total number of occurrences of all phrases within @@ -13110,12 +13163,13 @@ struct Fts5PhraseIter { ** Query for the details of phrase match iIdx within the current row. ** Phrase matches are numbered starting from zero, so the iIdx argument ** should be greater than or equal to zero and smaller than the value -** output by xInstCount(). +** output by xInstCount(). If iIdx is less than zero or greater than +** or equal to the value returned by xInstCount(), SQLITE_RANGE is returned. ** -** Usually, output parameter *piPhrase is set to the phrase number, *piCol +** Otherwise, output parameter *piPhrase is set to the phrase number, *piCol ** to the column in which it occurs and *piOff the token offset of the -** first token of the phrase. Returns SQLITE_OK if successful, or an error -** code (i.e. SQLITE_NOMEM) if an error occurs. +** first token of the phrase. SQLITE_OK is returned if successful, or an +** error code (i.e. SQLITE_NOMEM) if an error occurs. ** ** This API can be quite slow if used with an FTS5 table created with the ** "detail=none" or "detail=column" option. @@ -13141,6 +13195,10 @@ struct Fts5PhraseIter { ** Invoking Api.xUserData() returns a copy of the pointer passed as ** the third argument to pUserData. ** +** If parameter iPhrase is less than zero, or greater than or equal to +** the number of phrases in the query, as returned by xPhraseCount(), +** this function returns SQLITE_RANGE. +** ** If the callback function returns any value other than SQLITE_OK, the ** query is abandoned and the xQueryPhrase function returns immediately. ** If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK. @@ -13255,9 +13313,42 @@ struct Fts5PhraseIter { ** ** xPhraseNextColumn() ** See xPhraseFirstColumn above. +** +** xQueryToken(pFts5, iPhrase, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase iPhrase of the current +** query. Before returning, output parameter *ppToken is set to point +** to a buffer containing the requested token, and *pnToken to the +** size of this buffer in bytes. +** +** If iPhrase or iToken are less than zero, or if iPhrase is greater than +** or equal to the number of phrases in the query as reported by +** xPhraseCount(), or if iToken is equal to or greater than the number of +** tokens in the phrase, SQLITE_RANGE is returned and *ppToken and *pnToken + are both zeroed. +** +** The output text is not a copy of the query text that specified the +** token. It is the output of the tokenizer module. For tokendata=1 +** tables, this includes any embedded 0x00 and trailing data. +** +** xInstToken(pFts5, iIdx, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase hit iIdx within the +** current row. If iIdx is less than zero or greater than or equal to the +** value returned by xInstCount(), SQLITE_RANGE is returned. Otherwise, +** output variable (*ppToken) is set to point to a buffer containing the +** matching document token, and (*pnToken) to the size of that buffer in +** bytes. This API is not available if the specified token matches a +** prefix query term. In that case both output variables are always set +** to 0. +** +** The output text is not a copy of the document text that was tokenized. +** It is the output of the tokenizer module. For tokendata=1 tables, this +** includes any embedded 0x00 and trailing data. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. */ struct Fts5ExtensionApi { - int iVersion; /* Currently always set to 2 */ + int iVersion; /* Currently always set to 3 */ void *(*xUserData)(Fts5Context*); @@ -13292,6 +13383,13 @@ struct Fts5ExtensionApi { int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*); void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol); + + /* Below this point are iVersion>=3 only */ + int (*xQueryToken)(Fts5Context*, + int iPhrase, int iToken, + const char **ppToken, int *pnToken + ); + int (*xInstToken)(Fts5Context*, int iIdx, int iToken, const char**, int*); }; /* @@ -13778,7 +13876,7 @@ struct fts5_api { ** max_page_count macro. */ #ifndef SQLITE_MAX_PAGE_COUNT -# define SQLITE_MAX_PAGE_COUNT 1073741823 +# define SQLITE_MAX_PAGE_COUNT 0xfffffffe /* 4294967294 */ #endif /* @@ -13917,6 +14015,19 @@ struct fts5_api { # undef SQLITE_USE_SEH #endif +/* +** Enable SQLITE_DIRECT_OVERFLOW_READ, unless the build explicitly +** disables it using -DSQLITE_DIRECT_OVERFLOW_READ=0 +*/ +#if defined(SQLITE_DIRECT_OVERFLOW_READ) && SQLITE_DIRECT_OVERFLOW_READ+1==1 + /* Disable if -DSQLITE_DIRECT_OVERFLOW_READ=0 */ +# undef SQLITE_DIRECT_OVERFLOW_READ +#else + /* In all other cases, enable */ +# define SQLITE_DIRECT_OVERFLOW_READ 1 +#endif + + /* ** The SQLITE_THREADSAFE macro must be defined as 0, 1, or 2. ** 0 means mutexes are permanently disable and the library is never @@ -15799,7 +15910,7 @@ SQLITE_PRIVATE sqlite3_file *sqlite3PagerJrnlFile(Pager*); SQLITE_PRIVATE const char *sqlite3PagerJournalname(Pager*); SQLITE_PRIVATE void *sqlite3PagerTempSpace(Pager*); SQLITE_PRIVATE int sqlite3PagerIsMemdb(Pager*); -SQLITE_PRIVATE void sqlite3PagerCacheStat(Pager *, int, int, int *); +SQLITE_PRIVATE void sqlite3PagerCacheStat(Pager *, int, int, u64*); SQLITE_PRIVATE void sqlite3PagerClearCache(Pager*); SQLITE_PRIVATE int sqlite3SectorSize(sqlite3_file *); @@ -16386,6 +16497,7 @@ typedef struct VdbeOpList VdbeOpList; #define P4_INT64 (-13) /* P4 is a 64-bit signed integer */ #define P4_INTARRAY (-14) /* P4 is a vector of 32-bit integers */ #define P4_FUNCCTX (-15) /* P4 is a pointer to an sqlite3_context object */ +#define P4_TABLEREF (-16) /* Like P4_TABLE, but reference counted */ /* Error message codes for OP_Halt */ #define P5_ConstraintNotNull 1 @@ -16608,13 +16720,15 @@ typedef struct VdbeOpList VdbeOpList; #define OP_Pagecount 178 #define OP_MaxPgcnt 179 #define OP_ClrSubtype 180 /* synopsis: r[P1].subtype = 0 */ -#define OP_FilterAdd 181 /* synopsis: filter(P1) += key(P3@P4) */ -#define OP_Trace 182 -#define OP_CursorHint 183 -#define OP_ReleaseReg 184 /* synopsis: release r[P1@P2] mask P3 */ -#define OP_Noop 185 -#define OP_Explain 186 -#define OP_Abortable 187 +#define OP_GetSubtype 181 /* synopsis: r[P2] = r[P1].subtype */ +#define OP_SetSubtype 182 /* synopsis: r[P2].subtype = r[P1] */ +#define OP_FilterAdd 183 /* synopsis: filter(P1) += key(P3@P4) */ +#define OP_Trace 184 +#define OP_CursorHint 185 +#define OP_ReleaseReg 186 /* synopsis: release r[P1@P2] mask P3 */ +#define OP_Noop 187 +#define OP_Explain 188 +#define OP_Abortable 189 /* Properties such as "out2" or "jump" that are specified in ** comments following the "case" for each opcode in the vdbe.c @@ -16650,8 +16764,8 @@ typedef struct VdbeOpList VdbeOpList; /* 152 */ 0x00, 0x10, 0x00, 0x00, 0x06, 0x10, 0x00, 0x04,\ /* 160 */ 0x1a, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,\ /* 168 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x40, 0x10, 0x50,\ -/* 176 */ 0x40, 0x00, 0x10, 0x10, 0x02, 0x00, 0x00, 0x00,\ -/* 184 */ 0x00, 0x00, 0x00, 0x00,} +/* 176 */ 0x40, 0x00, 0x10, 0x10, 0x02, 0x12, 0x12, 0x00,\ +/* 184 */ 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,} /* The resolve3P2Values() routine is able to run faster if it knows ** the value of the largest JUMP opcode. The smaller the maximum @@ -17812,14 +17926,15 @@ struct FuncDestructor { #define SQLITE_FUNC_SLOCHNG 0x2000 /* "Slow Change". Value constant during a ** single query - might change over time */ #define SQLITE_FUNC_TEST 0x4000 /* Built-in testing functions */ -/* 0x8000 -- available for reuse */ +#define SQLITE_FUNC_RUNONLY 0x8000 /* Cannot be used by valueFromFunction */ #define SQLITE_FUNC_WINDOW 0x00010000 /* Built-in window-only function */ #define SQLITE_FUNC_INTERNAL 0x00040000 /* For use by NestedParse() only */ #define SQLITE_FUNC_DIRECT 0x00080000 /* Not for use in TRIGGERs or VIEWs */ -#define SQLITE_FUNC_SUBTYPE 0x00100000 /* Result likely to have sub-type */ +/* SQLITE_SUBTYPE 0x00100000 // Consumer of subtypes */ #define SQLITE_FUNC_UNSAFE 0x00200000 /* Function has side effects */ #define SQLITE_FUNC_INLINE 0x00400000 /* Functions implemented in-line */ #define SQLITE_FUNC_BUILTIN 0x00800000 /* This is a built-in function */ +/* SQLITE_RESULT_SUBTYPE 0x01000000 // Generator of subtypes */ #define SQLITE_FUNC_ANYORDER 0x08000000 /* count/min/max aggregate */ /* Identifier numbers for each in-line function */ @@ -17911,10 +18026,11 @@ struct FuncDestructor { #define MFUNCTION(zName, nArg, xPtr, xFunc) \ {nArg, SQLITE_FUNC_BUILTIN|SQLITE_FUNC_CONSTANT|SQLITE_UTF8, \ xPtr, 0, xFunc, 0, 0, 0, #zName, {0} } -#define JFUNCTION(zName, nArg, iArg, xFunc) \ - {nArg, SQLITE_FUNC_BUILTIN|SQLITE_DETERMINISTIC|\ - SQLITE_FUNC_CONSTANT|SQLITE_UTF8, \ - SQLITE_INT_TO_PTR(iArg), 0, xFunc, 0, 0, 0, #zName, {0} } +#define JFUNCTION(zName, nArg, bUseCache, bWS, bRS, bJsonB, iArg, xFunc) \ + {nArg, SQLITE_FUNC_BUILTIN|SQLITE_DETERMINISTIC|SQLITE_FUNC_CONSTANT|\ + SQLITE_UTF8|((bUseCache)*SQLITE_FUNC_RUNONLY)|\ + ((bRS)*SQLITE_SUBTYPE)|((bWS)*SQLITE_RESULT_SUBTYPE), \ + SQLITE_INT_TO_PTR(iArg|((bJsonB)*JSON_BLOB)),0,xFunc,0, 0, 0, #zName, {0} } #define INLINE_FUNC(zName, nArg, iArg, mFlags) \ {nArg, SQLITE_FUNC_BUILTIN|\ SQLITE_UTF8|SQLITE_FUNC_INLINE|SQLITE_FUNC_CONSTANT|(mFlags), \ @@ -18549,6 +18665,7 @@ struct Index { unsigned isCovering:1; /* True if this is a covering index */ unsigned noSkipScan:1; /* Do not try to use skip-scan if true */ unsigned hasStat1:1; /* aiRowLogEst values come from sqlite_stat1 */ + unsigned bLowQual:1; /* sqlite_stat1 says this is a low-quality index */ unsigned bNoQuery:1; /* Do not use this index to optimize queries */ unsigned bAscKeyBug:1; /* True if the bba7b69f9849b5bf bug applies */ unsigned bHasVCol:1; /* Index references one or more VIRTUAL columns */ @@ -18662,6 +18779,7 @@ struct AggInfo { int iOBTab; /* Ephemeral table to implement ORDER BY */ u8 bOBPayload; /* iOBTab has payload columns separate from key */ u8 bOBUnique; /* Enforce uniqueness on iOBTab keys */ + u8 bUseSubtype; /* Transfer subtype info through sorter */ } *aFunc; int nFunc; /* Number of entries in aFunc[] */ u32 selId; /* Select to which this AggInfo belongs */ @@ -19195,6 +19313,7 @@ struct NameContext { int nRef; /* Number of names resolved by this context */ int nNcErr; /* Number of errors encountered while resolving names */ int ncFlags; /* Zero or more NC_* flags defined below */ + u32 nNestedSelect; /* Number of nested selects using this NC */ Select *pWinSelect; /* SELECT statement for any window functions */ }; @@ -19911,6 +20030,9 @@ struct sqlite3_str { ** ** 3. Make a (read-only) copy of a read-only RCStr string using ** sqlite3RCStrRef(). +** +** "String" is in the name, but an RCStr object can also be used to hold +** binary data. */ struct RCStr { u64 nRCRef; /* Number of references */ @@ -19969,6 +20091,9 @@ struct Sqlite3Config { u8 bSmallMalloc; /* Avoid large memory allocations if true */ u8 bExtraSchemaChecks; /* Verify type,name,tbl_name in schema */ u8 bUseLongDouble; /* Make use of long double */ +#ifdef SQLITE_DEBUG + u8 bJsonSelfcheck; /* Double-check JSON parsing */ +#endif int mxStrlen; /* Maximum string length */ int neverCorrupt; /* Database is always well-formed */ int szLookaside; /* Default lookaside buffer size */ @@ -20595,6 +20720,7 @@ SQLITE_PRIVATE void sqlite3ExprOrderByAggregateError(Parse*,Expr*); SQLITE_PRIVATE void sqlite3ExprFunctionUsable(Parse*,const Expr*,const FuncDef*); SQLITE_PRIVATE void sqlite3ExprAssignVarNumber(Parse*, Expr*, u32); SQLITE_PRIVATE void sqlite3ExprDelete(sqlite3*, Expr*); +SQLITE_PRIVATE void sqlite3ExprDeleteGeneric(sqlite3*,void*); SQLITE_PRIVATE void sqlite3ExprDeferredDelete(Parse*, Expr*); SQLITE_PRIVATE void sqlite3ExprUnmapAndDelete(Parse*, Expr*); SQLITE_PRIVATE ExprList *sqlite3ExprListAppend(Parse*,ExprList*,Expr*); @@ -20604,6 +20730,7 @@ SQLITE_PRIVATE void sqlite3ExprListSetSortOrder(ExprList*,int,int); SQLITE_PRIVATE void sqlite3ExprListSetName(Parse*,ExprList*,const Token*,int); SQLITE_PRIVATE void sqlite3ExprListSetSpan(Parse*,ExprList*,const char*,const char*); SQLITE_PRIVATE void sqlite3ExprListDelete(sqlite3*, ExprList*); +SQLITE_PRIVATE void sqlite3ExprListDeleteGeneric(sqlite3*,void*); SQLITE_PRIVATE u32 sqlite3ExprListFlags(const ExprList*); SQLITE_PRIVATE int sqlite3IndexHasDuplicateRootPage(Index*); SQLITE_PRIVATE int sqlite3Init(sqlite3*, char**); @@ -20694,6 +20821,7 @@ SQLITE_PRIVATE int sqlite3DbMaskAllZero(yDbMask); SQLITE_PRIVATE void sqlite3DropTable(Parse*, SrcList*, int, int); SQLITE_PRIVATE void sqlite3CodeDropTable(Parse*, Table*, int, int); SQLITE_PRIVATE void sqlite3DeleteTable(sqlite3*, Table*); +SQLITE_PRIVATE void sqlite3DeleteTableGeneric(sqlite3*, void*); SQLITE_PRIVATE void sqlite3FreeIndex(sqlite3*, Index*); #ifndef SQLITE_OMIT_AUTOINCREMENT SQLITE_PRIVATE void sqlite3AutoincrementBegin(Parse *pParse); @@ -20730,6 +20858,7 @@ SQLITE_PRIVATE int sqlite3Select(Parse*, Select*, SelectDest*); SQLITE_PRIVATE Select *sqlite3SelectNew(Parse*,ExprList*,SrcList*,Expr*,ExprList*, Expr*,ExprList*,u32,Expr*); SQLITE_PRIVATE void sqlite3SelectDelete(sqlite3*, Select*); +SQLITE_PRIVATE void sqlite3SelectDeleteGeneric(sqlite3*,void*); SQLITE_PRIVATE Table *sqlite3SrcListLookup(Parse*, SrcList*); SQLITE_PRIVATE int sqlite3IsReadOnly(Parse*, Table*, Trigger*); SQLITE_PRIVATE void sqlite3OpenTable(Parse*, int iCur, int iDb, Table*, int); @@ -20956,6 +21085,7 @@ SQLITE_PRIVATE int sqlite3Utf16ByteLen(const void *pData, int nChar); #endif SQLITE_PRIVATE int sqlite3Utf8CharLen(const char *pData, int nByte); SQLITE_PRIVATE u32 sqlite3Utf8Read(const u8**); +SQLITE_PRIVATE int sqlite3Utf8ReadLimited(const u8*, int, u32*); SQLITE_PRIVATE LogEst sqlite3LogEst(u64); SQLITE_PRIVATE LogEst sqlite3LogEstAdd(LogEst,LogEst); SQLITE_PRIVATE LogEst sqlite3LogEstFromDouble(double); @@ -21302,6 +21432,7 @@ SQLITE_PRIVATE Cte *sqlite3CteNew(Parse*,Token*,ExprList*,Select*,u8); SQLITE_PRIVATE void sqlite3CteDelete(sqlite3*,Cte*); SQLITE_PRIVATE With *sqlite3WithAdd(Parse*,With*,Cte*); SQLITE_PRIVATE void sqlite3WithDelete(sqlite3*,With*); +SQLITE_PRIVATE void sqlite3WithDeleteGeneric(sqlite3*,void*); SQLITE_PRIVATE With *sqlite3WithPush(Parse*, With*, u8); #else # define sqlite3CteNew(P,T,E,S) ((void*)0) @@ -22679,6 +22810,9 @@ SQLITE_PRIVATE SQLITE_WSD struct Sqlite3Config sqlite3Config = { 0, /* bSmallMalloc */ 1, /* bExtraSchemaChecks */ sizeof(LONGDOUBLE_TYPE)>8, /* bUseLongDouble */ +#ifdef SQLITE_DEBUG + 0, /* bJsonSelfcheck */ +#endif 0x7ffffffe, /* mxStrlen */ 0, /* neverCorrupt */ SQLITE_DEFAULT_LOOKASIDE, /* szLookaside, nLookaside */ @@ -23931,7 +24065,7 @@ SQLITE_API int sqlite3_db_status( case SQLITE_DBSTATUS_CACHE_MISS: case SQLITE_DBSTATUS_CACHE_WRITE:{ int i; - int nRet = 0; + u64 nRet = 0; assert( SQLITE_DBSTATUS_CACHE_MISS==SQLITE_DBSTATUS_CACHE_HIT+1 ); assert( SQLITE_DBSTATUS_CACHE_WRITE==SQLITE_DBSTATUS_CACHE_HIT+2 ); @@ -23944,7 +24078,7 @@ SQLITE_API int sqlite3_db_status( *pHighwater = 0; /* IMP: R-42420-56072 */ /* IMP: R-54100-20147 */ /* IMP: R-29431-39229 */ - *pCurrent = nRet; + *pCurrent = (int)nRet & 0x7fffffff; break; } @@ -25013,6 +25147,12 @@ static int isDate( } computeJD(p); if( p->isError || !validJulianDay(p->iJD) ) return 1; + if( argc==1 && p->validYMD && p->D>28 ){ + /* Make sure a YYYY-MM-DD is normalized. + ** Example: 2023-02-31 -> 2023-03-03 */ + assert( p->validJD ); + p->validYMD = 0; + } return 0; } @@ -29454,7 +29594,7 @@ SQLITE_PRIVATE void sqlite3MemoryBarrier(void){ SQLITE_MEMORY_BARRIER; #elif defined(__GNUC__) __sync_synchronize(); -#elif MSVC_VERSION>=1300 +#elif MSVC_VERSION>=1400 _ReadWriteBarrier(); #elif defined(MemoryBarrier) MemoryBarrier(); @@ -32041,7 +32181,7 @@ SQLITE_API void sqlite3_str_appendf(StrAccum *p, const char *zFormat, ...){ /***************************************************************************** -** Reference counted string storage +** Reference counted string/blob storage *****************************************************************************/ /* @@ -32893,7 +33033,7 @@ SQLITE_PRIVATE void sqlite3TreeViewExpr(TreeView *pView, const Expr *pExpr, u8 m assert( pExpr->x.pList->nExpr==2 ); pY = pExpr->x.pList->a[0].pExpr; pZ = pExpr->x.pList->a[1].pExpr; - sqlite3TreeViewLine(pView, "BETWEEN"); + sqlite3TreeViewLine(pView, "BETWEEN%s", zFlgs); sqlite3TreeViewExpr(pView, pX, 1); sqlite3TreeViewExpr(pView, pY, 1); sqlite3TreeViewExpr(pView, pZ, 0); @@ -34028,7 +34168,38 @@ SQLITE_PRIVATE u32 sqlite3Utf8Read( return c; } - +/* +** Read a single UTF8 character out of buffer z[], but reading no +** more than n characters from the buffer. z[] is not zero-terminated. +** +** Return the number of bytes used to construct the character. +** +** Invalid UTF8 might generate a strange result. No effort is made +** to detect invalid UTF8. +** +** At most 4 bytes will be read out of z[]. The return value will always +** be between 1 and 4. +*/ +SQLITE_PRIVATE int sqlite3Utf8ReadLimited( + const u8 *z, + int n, + u32 *piOut +){ + u32 c; + int i = 1; + assert( n>0 ); + c = z[0]; + if( c>=0xc0 ){ + c = sqlite3Utf8Trans1[c-0xc0]; + if( n>4 ) n = 4; + while( iiBusyTimeout; +#if SQLITE_ENABLE_SETLK_TIMEOUT==1 pFile->iBusyTimeout = *(int*)pArg; +#elif SQLITE_ENABLE_SETLK_TIMEOUT==2 + pFile->iBusyTimeout = !!(*(int*)pArg); +#else +# error "SQLITE_ENABLE_SETLK_TIMEOUT must be set to 1 or 2" +#endif *(int*)pArg = iOld; return SQLITE_OK; } @@ -42102,6 +42281,25 @@ static int unixGetpagesize(void){ ** Either unixShmNode.pShmMutex must be held or unixShmNode.nRef==0 and ** unixMutexHeld() is true when reading or writing any other field ** in this structure. +** +** aLock[SQLITE_SHM_NLOCK]: +** This array records the various locks held by clients on each of the +** SQLITE_SHM_NLOCK slots. If the aLock[] entry is set to 0, then no +** locks are held by the process on this slot. If it is set to -1, then +** some client holds an EXCLUSIVE lock on the locking slot. If the aLock[] +** value is set to a positive value, then it is the number of shared +** locks currently held on the slot. +** +** aMutex[SQLITE_SHM_NLOCK]: +** Normally, when SQLITE_ENABLE_SETLK_TIMEOUT is not defined, mutex +** pShmMutex is used to protect the aLock[] array and the right to +** call fcntl() on unixShmNode.hShm to obtain or release locks. +** +** If SQLITE_ENABLE_SETLK_TIMEOUT is defined though, we use an array +** of mutexes - one for each locking slot. To read or write locking +** slot aLock[iSlot], the caller must hold the corresponding mutex +** aMutex[iSlot]. Similarly, to call fcntl() to obtain or release a +** lock corresponding to slot iSlot, mutex aMutex[iSlot] must be held. */ struct unixShmNode { unixInodeInfo *pInode; /* unixInodeInfo that owns this SHM node */ @@ -42115,10 +42313,11 @@ struct unixShmNode { char **apRegion; /* Array of mapped shared-memory regions */ int nRef; /* Number of unixShm objects pointing to this */ unixShm *pFirst; /* All unixShm objects pointing to this */ +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + sqlite3_mutex *aMutex[SQLITE_SHM_NLOCK]; +#endif int aLock[SQLITE_SHM_NLOCK]; /* # shared locks on slot, -1==excl lock */ #ifdef SQLITE_DEBUG - u8 exclMask; /* Mask of exclusive locks held */ - u8 sharedMask; /* Mask of shared locks held */ u8 nextShmId; /* Next available unixShm.id value */ #endif }; @@ -42201,16 +42400,35 @@ static int unixShmSystemLock( struct flock f; /* The posix advisory locking structure */ int rc = SQLITE_OK; /* Result code form fcntl() */ - /* Access to the unixShmNode object is serialized by the caller */ pShmNode = pFile->pInode->pShmNode; - assert( pShmNode->nRef==0 || sqlite3_mutex_held(pShmNode->pShmMutex) ); - assert( pShmNode->nRef>0 || unixMutexHeld() ); + + /* Assert that the parameters are within expected range and that the + ** correct mutex or mutexes are held. */ + assert( pShmNode->nRef>=0 ); + assert( (ofst==UNIX_SHM_DMS && n==1) + || (ofst>=UNIX_SHM_BASE && ofst+n<=(UNIX_SHM_BASE+SQLITE_SHM_NLOCK)) + ); + if( ofst==UNIX_SHM_DMS ){ + assert( pShmNode->nRef>0 || unixMutexHeld() ); + assert( pShmNode->nRef==0 || sqlite3_mutex_held(pShmNode->pShmMutex) ); + }else{ +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + int ii; + for(ii=ofst-UNIX_SHM_BASE; iiaMutex[ii]) ); + } +#else + assert( sqlite3_mutex_held(pShmNode->pShmMutex) ); + assert( pShmNode->nRef>0 ); +#endif + } /* Shared locks never span more than one byte */ assert( n==1 || lockType!=F_RDLCK ); /* Locks are within range */ assert( n>=1 && n<=SQLITE_SHM_NLOCK ); + assert( ofst>=UNIX_SHM_BASE && ofst<=(UNIX_SHM_DMS+SQLITE_SHM_NLOCK) ); if( pShmNode->hShm>=0 ){ int res; @@ -42221,7 +42439,7 @@ static int unixShmSystemLock( f.l_len = n; res = osSetPosixAdvisoryLock(pShmNode->hShm, &f, pFile); if( res==-1 ){ -#ifdef SQLITE_ENABLE_SETLK_TIMEOUT +#if defined(SQLITE_ENABLE_SETLK_TIMEOUT) && SQLITE_ENABLE_SETLK_TIMEOUT==1 rc = (pFile->iBusyTimeout ? SQLITE_BUSY_TIMEOUT : SQLITE_BUSY); #else rc = SQLITE_BUSY; @@ -42229,39 +42447,28 @@ static int unixShmSystemLock( } } - /* Update the global lock state and do debug tracing */ + /* Do debug tracing */ #ifdef SQLITE_DEBUG - { u16 mask; OSTRACE(("SHM-LOCK ")); - mask = ofst>31 ? 0xffff : (1<<(ofst+n)) - (1<exclMask &= ~mask; - pShmNode->sharedMask &= ~mask; + OSTRACE(("unlock %d..%d ok\n", ofst, ofst+n-1)); }else if( lockType==F_RDLCK ){ - OSTRACE(("read-lock %d ok", ofst)); - pShmNode->exclMask &= ~mask; - pShmNode->sharedMask |= mask; + OSTRACE(("read-lock %d..%d ok\n", ofst, ofst+n-1)); }else{ assert( lockType==F_WRLCK ); - OSTRACE(("write-lock %d ok", ofst)); - pShmNode->exclMask |= mask; - pShmNode->sharedMask &= ~mask; + OSTRACE(("write-lock %d..%d ok\n", ofst, ofst+n-1)); } }else{ if( lockType==F_UNLCK ){ - OSTRACE(("unlock %d failed", ofst)); + OSTRACE(("unlock %d..%d failed\n", ofst, ofst+n-1)); }else if( lockType==F_RDLCK ){ - OSTRACE(("read-lock failed")); + OSTRACE(("read-lock %d..%d failed\n", ofst, ofst+n-1)); }else{ assert( lockType==F_WRLCK ); - OSTRACE(("write-lock %d failed", ofst)); + OSTRACE(("write-lock %d..%d failed\n", ofst, ofst+n-1)); } } - OSTRACE((" - afterwards %03x,%03x\n", - pShmNode->sharedMask, pShmNode->exclMask)); - } #endif return rc; @@ -42298,6 +42505,11 @@ static void unixShmPurge(unixFile *pFd){ int i; assert( p->pInode==pFd->pInode ); sqlite3_mutex_free(p->pShmMutex); +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + for(i=0; iaMutex[i]); + } +#endif for(i=0; inRegion; i+=nShmPerMap){ if( p->hShm>=0 ){ osMunmap(p->apRegion[i], p->szRegion); @@ -42357,7 +42569,20 @@ static int unixLockSharedMemory(unixFile *pDbFd, unixShmNode *pShmNode){ pShmNode->isUnlocked = 1; rc = SQLITE_READONLY_CANTINIT; }else{ +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + /* Do not use a blocking lock here. If the lock cannot be obtained + ** immediately, it means some other connection is truncating the + ** *-shm file. And after it has done so, it will not release its + ** lock, but only downgrade it to a shared lock. So no point in + ** blocking here. The call below to obtain the shared DMS lock may + ** use a blocking lock. */ + int iSaveTimeout = pDbFd->iBusyTimeout; + pDbFd->iBusyTimeout = 0; +#endif rc = unixShmSystemLock(pDbFd, F_WRLCK, UNIX_SHM_DMS, 1); +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + pDbFd->iBusyTimeout = iSaveTimeout; +#endif /* The first connection to attach must truncate the -shm file. We ** truncate to 3 bytes (an arbitrary small number, less than the ** -shm header size) rather than 0 as a system debugging aid, to @@ -42478,6 +42703,18 @@ static int unixOpenSharedMemory(unixFile *pDbFd){ rc = SQLITE_NOMEM_BKPT; goto shm_open_err; } +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + { + int ii; + for(ii=0; iiaMutex[ii] = sqlite3_mutex_alloc(SQLITE_MUTEX_FAST); + if( pShmNode->aMutex[ii]==0 ){ + rc = SQLITE_NOMEM_BKPT; + goto shm_open_err; + } + } + } +#endif } if( pInode->bProcessLock==0 ){ @@ -42699,9 +42936,11 @@ static int unixShmMap( */ #ifdef SQLITE_DEBUG static int assertLockingArrayOk(unixShmNode *pShmNode){ +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + return 1; +#else unixShm *pX; int aLock[SQLITE_SHM_NLOCK]; - assert( sqlite3_mutex_held(pShmNode->pShmMutex) ); memset(aLock, 0, sizeof(aLock)); for(pX=pShmNode->pFirst; pX; pX=pX->pNext){ @@ -42719,13 +42958,14 @@ static int assertLockingArrayOk(unixShmNode *pShmNode){ assert( 0==memcmp(pShmNode->aLock, aLock, sizeof(aLock)) ); return (memcmp(pShmNode->aLock, aLock, sizeof(aLock))==0); +#endif } #endif /* ** Change the lock state for a shared-memory segment. ** -** Note that the relationship between SHAREd and EXCLUSIVE locks is a little +** Note that the relationship between SHARED and EXCLUSIVE locks is a little ** different here than in posix. In xShmLock(), one can go from unlocked ** to shared and back or from unlocked to exclusive and back. But one may ** not go from shared to exclusive or from exclusive to shared. @@ -42740,7 +42980,7 @@ static int unixShmLock( unixShm *p; /* The shared memory being locked */ unixShmNode *pShmNode; /* The underlying file iNode */ int rc = SQLITE_OK; /* Result code */ - u16 mask; /* Mask of locks to take or release */ + u16 mask = (1<<(ofst+n)) - (1<pShm; @@ -42775,88 +43015,151 @@ static int unixShmLock( ** It is not permitted to block on the RECOVER lock. */ #ifdef SQLITE_ENABLE_SETLK_TIMEOUT - assert( (flags & SQLITE_SHM_UNLOCK) || pDbFd->iBusyTimeout==0 || ( - (ofst!=2) /* not RECOVER */ - && (ofst!=1 || (p->exclMask|p->sharedMask)==0) - && (ofst!=0 || (p->exclMask|p->sharedMask)<3) - && (ofst<3 || (p->exclMask|p->sharedMask)<(1<exclMask|p->sharedMask); + assert( (flags & SQLITE_SHM_UNLOCK) || pDbFd->iBusyTimeout==0 || ( + (ofst!=2) /* not RECOVER */ + && (ofst!=1 || lockMask==0 || lockMask==2) + && (ofst!=0 || lockMask<3) + && (ofst<3 || lockMask<(1<1 || mask==(1<pShmMutex); - assert( assertLockingArrayOk(pShmNode) ); - if( flags & SQLITE_SHM_UNLOCK ){ - if( (p->exclMask|p->sharedMask) & mask ){ - int ii; - int bUnlock = 1; + /* Check if there is any work to do. There are three cases: + ** + ** a) An unlock operation where there are locks to unlock, + ** b) An shared lock where the requested lock is not already held + ** c) An exclusive lock where the requested lock is not already held + ** + ** The SQLite core never requests an exclusive lock that it already holds. + ** This is assert()ed below. + */ + assert( flags!=(SQLITE_SHM_EXCLUSIVE|SQLITE_SHM_LOCK) + || 0==(p->exclMask & mask) + ); + if( ((flags & SQLITE_SHM_UNLOCK) && ((p->exclMask|p->sharedMask) & mask)) + || (flags==(SQLITE_SHM_SHARED|SQLITE_SHM_LOCK) && 0==(p->sharedMask & mask)) + || (flags==(SQLITE_SHM_EXCLUSIVE|SQLITE_SHM_LOCK)) + ){ - for(ii=ofst; ii((p->sharedMask & (1<aMutex[iMutex]); + if( rc!=SQLITE_OK ) goto leave_shmnode_mutexes; + }else{ + sqlite3_mutex_enter(pShmNode->aMutex[iMutex]); } + } +#else + sqlite3_mutex_enter(pShmNode->pShmMutex); +#endif - if( bUnlock ){ - rc = unixShmSystemLock(pDbFd, F_UNLCK, ofst+UNIX_SHM_BASE, n); - if( rc==SQLITE_OK ){ - memset(&aLock[ofst], 0, sizeof(int)*n); + if( ALWAYS(rc==SQLITE_OK) ){ + if( flags & SQLITE_SHM_UNLOCK ){ + /* Case (a) - unlock. */ + int bUnlock = 1; + assert( (p->exclMask & p->sharedMask)==0 ); + assert( !(flags & SQLITE_SHM_EXCLUSIVE) || (p->exclMask & mask)==mask ); + assert( !(flags & SQLITE_SHM_SHARED) || (p->sharedMask & mask)==mask ); + + /* If this is a SHARED lock being unlocked, it is possible that other + ** clients within this process are holding the same SHARED lock. In + ** this case, set bUnlock to 0 so that the posix lock is not removed + ** from the file-descriptor below. */ + if( flags & SQLITE_SHM_SHARED ){ + assert( n==1 ); + assert( aLock[ofst]>=1 ); + if( aLock[ofst]>1 ){ + bUnlock = 0; + aLock[ofst]--; + p->sharedMask &= ~mask; + } } - }else if( ALWAYS(p->sharedMask & (1<1 ); - aLock[ofst]--; - } - /* Undo the local locks */ - if( rc==SQLITE_OK ){ - p->exclMask &= ~mask; - p->sharedMask &= ~mask; - } - } - }else if( flags & SQLITE_SHM_SHARED ){ - assert( n==1 ); - assert( (p->exclMask & (1<sharedMask & mask)==0 ){ - if( aLock[ofst]<0 ){ - rc = SQLITE_BUSY; - }else if( aLock[ofst]==0 ){ - rc = unixShmSystemLock(pDbFd, F_RDLCK, ofst+UNIX_SHM_BASE, n); - } + if( bUnlock ){ + rc = unixShmSystemLock(pDbFd, F_UNLCK, ofst+UNIX_SHM_BASE, n); + if( rc==SQLITE_OK ){ + memset(&aLock[ofst], 0, sizeof(int)*n); + p->sharedMask &= ~mask; + p->exclMask &= ~mask; + } + } + }else if( flags & SQLITE_SHM_SHARED ){ + /* Case (b) - a shared lock. */ - /* Get the local shared locks */ - if( rc==SQLITE_OK ){ - p->sharedMask |= mask; - aLock[ofst]++; - } - } - }else{ - /* Make sure no sibling connections hold locks that will block this - ** lock. If any do, return SQLITE_BUSY right away. */ - int ii; - for(ii=ofst; iisharedMask & mask)==0 ); - if( ALWAYS((p->exclMask & (1<sharedMask |= mask; + aLock[ofst]++; + } + }else{ + /* Case (c) - an exclusive lock. */ + int ii; + + assert( flags==(SQLITE_SHM_LOCK|SQLITE_SHM_EXCLUSIVE) ); assert( (p->sharedMask & mask)==0 ); - p->exclMask |= mask; + assert( (p->exclMask & mask)==0 ); + + /* Make sure no sibling connections hold locks that will block this + ** lock. If any do, return SQLITE_BUSY right away. */ for(ii=ofst; iiexclMask |= mask; + for(ii=ofst; ii=ofst; iMutex--){ + sqlite3_mutex_leave(pShmNode->aMutex[iMutex]); } +#else + sqlite3_mutex_leave(pShmNode->pShmMutex); +#endif } - assert( assertLockingArrayOk(pShmNode) ); - sqlite3_mutex_leave(pShmNode->pShmMutex); + OSTRACE(("SHM-LOCK shmid-%d, pid-%d got %03x,%03x\n", p->id, osGetpid(0), p->sharedMask, p->exclMask)); return rc; @@ -57074,7 +57377,7 @@ struct Pager { char *zJournal; /* Name of the journal file */ int (*xBusyHandler)(void*); /* Function to call when busy */ void *pBusyHandlerArg; /* Context argument for xBusyHandler */ - int aStat[4]; /* Total cache hits, misses, writes, spills */ + u32 aStat[4]; /* Total cache hits, misses, writes, spills */ #ifdef SQLITE_TEST int nRead; /* Database pages read */ #endif @@ -57204,9 +57507,8 @@ SQLITE_PRIVATE int sqlite3PagerDirectReadOk(Pager *pPager, Pgno pgno){ #ifndef SQLITE_OMIT_WAL if( pPager->pWal ){ u32 iRead = 0; - int rc; - rc = sqlite3WalFindFrame(pPager->pWal, pgno, &iRead); - return (rc==SQLITE_OK && iRead==0); + (void)sqlite3WalFindFrame(pPager->pWal, pgno, &iRead); + return iRead==0; } #endif return 1; @@ -61448,10 +61750,13 @@ SQLITE_PRIVATE int sqlite3PagerOpen( */ SQLITE_API sqlite3_file *sqlite3_database_file_object(const char *zName){ Pager *pPager; + const char *p; while( zName[-1]!=0 || zName[-2]!=0 || zName[-3]!=0 || zName[-4]!=0 ){ zName--; } - pPager = *(Pager**)(zName - 4 - sizeof(Pager*)); + p = zName - 4 - sizeof(Pager*); + assert( EIGHT_BYTE_ALIGNMENT(p) ); + pPager = *(Pager**)p; return pPager->fd; } @@ -63215,11 +63520,11 @@ SQLITE_PRIVATE int *sqlite3PagerStats(Pager *pPager){ a[3] = pPager->eState==PAGER_OPEN ? -1 : (int) pPager->dbSize; a[4] = pPager->eState; a[5] = pPager->errCode; - a[6] = pPager->aStat[PAGER_STAT_HIT]; - a[7] = pPager->aStat[PAGER_STAT_MISS]; + a[6] = (int)pPager->aStat[PAGER_STAT_HIT] & 0x7fffffff; + a[7] = (int)pPager->aStat[PAGER_STAT_MISS] & 0x7fffffff; a[8] = 0; /* Used to be pPager->nOvfl */ a[9] = pPager->nRead; - a[10] = pPager->aStat[PAGER_STAT_WRITE]; + a[10] = (int)pPager->aStat[PAGER_STAT_WRITE] & 0x7fffffff; return a; } #endif @@ -63235,7 +63540,7 @@ SQLITE_PRIVATE int *sqlite3PagerStats(Pager *pPager){ ** reset parameter is non-zero, the cache hit or miss count is zeroed before ** returning. */ -SQLITE_PRIVATE void sqlite3PagerCacheStat(Pager *pPager, int eStat, int reset, int *pnVal){ +SQLITE_PRIVATE void sqlite3PagerCacheStat(Pager *pPager, int eStat, int reset, u64 *pnVal){ assert( eStat==SQLITE_DBSTATUS_CACHE_HIT || eStat==SQLITE_DBSTATUS_CACHE_MISS @@ -64175,7 +64480,7 @@ SQLITE_PRIVATE int sqlite3PagerWalFramesize(Pager *pPager){ } #endif -#ifdef SQLITE_USE_SEH +#if defined(SQLITE_USE_SEH) && !defined(SQLITE_OMIT_WAL) SQLITE_PRIVATE int sqlite3PagerWalSystemErrno(Pager *pPager){ return sqlite3WalSystemErrno(pPager->pWal); } @@ -66191,6 +66496,19 @@ static int walIteratorInit(Wal *pWal, u32 nBackfill, WalIterator **pp){ } #ifdef SQLITE_ENABLE_SETLK_TIMEOUT + + +/* +** Attempt to enable blocking locks that block for nMs ms. Return 1 if +** blocking locks are successfully enabled, or 0 otherwise. +*/ +static int walEnableBlockingMs(Wal *pWal, int nMs){ + int rc = sqlite3OsFileControl( + pWal->pDbFd, SQLITE_FCNTL_LOCK_TIMEOUT, (void*)&nMs + ); + return (rc==SQLITE_OK); +} + /* ** Attempt to enable blocking locks. Blocking locks are enabled only if (a) ** they are supported by the VFS, and (b) the database handle is configured @@ -66202,11 +66520,7 @@ static int walEnableBlocking(Wal *pWal){ if( pWal->db ){ int tmout = pWal->db->busyTimeout; if( tmout ){ - int rc; - rc = sqlite3OsFileControl( - pWal->pDbFd, SQLITE_FCNTL_LOCK_TIMEOUT, (void*)&tmout - ); - res = (rc==SQLITE_OK); + res = walEnableBlockingMs(pWal, tmout); } } return res; @@ -66255,20 +66569,10 @@ SQLITE_PRIVATE void sqlite3WalDb(Wal *pWal, sqlite3 *db){ pWal->db = db; } -/* -** Take an exclusive WRITE lock. Blocking if so configured. -*/ -static int walLockWriter(Wal *pWal){ - int rc; - walEnableBlocking(pWal); - rc = walLockExclusive(pWal, WAL_WRITE_LOCK, 1); - walDisableBlocking(pWal); - return rc; -} #else # define walEnableBlocking(x) 0 # define walDisableBlocking(x) -# define walLockWriter(pWal) walLockExclusive((pWal), WAL_WRITE_LOCK, 1) +# define walEnableBlockingMs(pWal, ms) 0 # define sqlite3WalDb(pWal, db) #endif /* ifdef SQLITE_ENABLE_SETLK_TIMEOUT */ @@ -66869,7 +67173,9 @@ static int walIndexReadHdr(Wal *pWal, int *pChanged){ } }else{ int bWriteLock = pWal->writeLock; - if( bWriteLock || SQLITE_OK==(rc = walLockWriter(pWal)) ){ + if( bWriteLock + || SQLITE_OK==(rc = walLockExclusive(pWal, WAL_WRITE_LOCK, 1)) + ){ pWal->writeLock = 1; if( SQLITE_OK==(rc = walIndexPage(pWal, 0, &page0)) ){ badHdr = walIndexTryHdr(pWal, pChanged); @@ -66877,7 +67183,8 @@ static int walIndexReadHdr(Wal *pWal, int *pChanged){ /* If the wal-index header is still malformed even while holding ** a WRITE lock, it can only mean that the header is corrupted and ** needs to be reconstructed. So run recovery to do exactly that. - */ + ** Disable blocking locks first. */ + walDisableBlocking(pWal); rc = walIndexRecover(pWal); *pChanged = 1; } @@ -67087,6 +67394,37 @@ static int walBeginShmUnreliable(Wal *pWal, int *pChanged){ return rc; } +/* +** The final argument passed to walTryBeginRead() is of type (int*). The +** caller should invoke walTryBeginRead as follows: +** +** int cnt = 0; +** do { +** rc = walTryBeginRead(..., &cnt); +** }while( rc==WAL_RETRY ); +** +** The final value of "cnt" is of no use to the caller. It is used by +** the implementation of walTryBeginRead() as follows: +** +** + Each time walTryBeginRead() is called, it is incremented. Once +** it reaches WAL_RETRY_PROTOCOL_LIMIT - indicating that walTryBeginRead() +** has many times been invoked and failed with WAL_RETRY - walTryBeginRead() +** returns SQLITE_PROTOCOL. +** +** + If SQLITE_ENABLE_SETLK_TIMEOUT is defined and walTryBeginRead() failed +** because a blocking lock timed out (SQLITE_BUSY_TIMEOUT from the OS +** layer), the WAL_RETRY_BLOCKED_MASK bit is set in "cnt". In this case +** the next invocation of walTryBeginRead() may omit an expected call to +** sqlite3OsSleep(). There has already been a delay when the previous call +** waited on a lock. +*/ +#define WAL_RETRY_PROTOCOL_LIMIT 100 +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT +# define WAL_RETRY_BLOCKED_MASK 0x10000000 +#else +# define WAL_RETRY_BLOCKED_MASK 0 +#endif + /* ** Attempt to start a read transaction. This might fail due to a race or ** other transient condition. When that happens, it returns WAL_RETRY to @@ -67137,13 +67475,16 @@ static int walBeginShmUnreliable(Wal *pWal, int *pChanged){ ** so it takes care to hold an exclusive lock on the corresponding ** WAL_READ_LOCK() while changing values. */ -static int walTryBeginRead(Wal *pWal, int *pChanged, int useWal, int cnt){ +static int walTryBeginRead(Wal *pWal, int *pChanged, int useWal, int *pCnt){ volatile WalCkptInfo *pInfo; /* Checkpoint information in wal-index */ u32 mxReadMark; /* Largest aReadMark[] value */ int mxI; /* Index of largest aReadMark[] value */ int i; /* Loop counter */ int rc = SQLITE_OK; /* Return code */ u32 mxFrame; /* Wal frame to lock to */ +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + int nBlockTmout = 0; +#endif assert( pWal->readLock<0 ); /* Not currently locked */ @@ -67167,14 +67508,34 @@ static int walTryBeginRead(Wal *pWal, int *pChanged, int useWal, int cnt){ ** so that on the 100th (and last) RETRY we delay for 323 milliseconds. ** The total delay time before giving up is less than 10 seconds. */ - if( cnt>5 ){ + (*pCnt)++; + if( *pCnt>5 ){ int nDelay = 1; /* Pause time in microseconds */ - if( cnt>100 ){ + int cnt = (*pCnt & ~WAL_RETRY_BLOCKED_MASK); + if( cnt>WAL_RETRY_PROTOCOL_LIMIT ){ VVA_ONLY( pWal->lockError = 1; ) return SQLITE_PROTOCOL; } - if( cnt>=10 ) nDelay = (cnt-9)*(cnt-9)*39; + if( *pCnt>=10 ) nDelay = (cnt-9)*(cnt-9)*39; +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + /* In SQLITE_ENABLE_SETLK_TIMEOUT builds, configure the file-descriptor + ** to block for locks for approximately nDelay us. This affects three + ** locks: (a) the shared lock taken on the DMS slot in os_unix.c (if + ** using os_unix.c), (b) the WRITER lock taken in walIndexReadHdr() if the + ** first attempted read fails, and (c) the shared lock taken on the + ** read-mark. + ** + ** If the previous call failed due to an SQLITE_BUSY_TIMEOUT error, + ** then sleep for the minimum of 1us. The previous call already provided + ** an extra delay while it was blocking on the lock. + */ + nBlockTmout = (nDelay+998) / 1000; + if( !useWal && walEnableBlockingMs(pWal, nBlockTmout) ){ + if( *pCnt & WAL_RETRY_BLOCKED_MASK ) nDelay = 1; + } +#endif sqlite3OsSleep(pWal->pVfs, nDelay); + *pCnt &= ~WAL_RETRY_BLOCKED_MASK; } if( !useWal ){ @@ -67182,6 +67543,13 @@ static int walTryBeginRead(Wal *pWal, int *pChanged, int useWal, int cnt){ if( pWal->bShmUnreliable==0 ){ rc = walIndexReadHdr(pWal, pChanged); } +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + walDisableBlocking(pWal); + if( rc==SQLITE_BUSY_TIMEOUT ){ + rc = SQLITE_BUSY; + *pCnt |= WAL_RETRY_BLOCKED_MASK; + } +#endif if( rc==SQLITE_BUSY ){ /* If there is not a recovery running in another thread or process ** then convert BUSY errors to WAL_RETRY. If recovery is known to @@ -67296,9 +67664,19 @@ static int walTryBeginRead(Wal *pWal, int *pChanged, int useWal, int cnt){ return rc==SQLITE_BUSY ? WAL_RETRY : SQLITE_READONLY_CANTINIT; } + (void)walEnableBlockingMs(pWal, nBlockTmout); rc = walLockShared(pWal, WAL_READ_LOCK(mxI)); + walDisableBlocking(pWal); if( rc ){ - return rc==SQLITE_BUSY ? WAL_RETRY : rc; +#ifdef SQLITE_ENABLE_SETLK_TIMEOUT + if( rc==SQLITE_BUSY_TIMEOUT ){ + *pCnt |= WAL_RETRY_BLOCKED_MASK; + } +#else + assert( rc!=SQLITE_BUSY_TIMEOUT ); +#endif + assert( (rc&0xFF)!=SQLITE_BUSY||rc==SQLITE_BUSY||rc==SQLITE_BUSY_TIMEOUT ); + return (rc&0xFF)==SQLITE_BUSY ? WAL_RETRY : rc; } /* Now that the read-lock has been obtained, check that neither the ** value in the aReadMark[] array or the contents of the wal-index @@ -67486,7 +67864,7 @@ static int walBeginReadTransaction(Wal *pWal, int *pChanged){ #endif do{ - rc = walTryBeginRead(pWal, pChanged, 0, ++cnt); + rc = walTryBeginRead(pWal, pChanged, 0, &cnt); }while( rc==WAL_RETRY ); testcase( (rc&0xff)==SQLITE_BUSY ); testcase( (rc&0xff)==SQLITE_IOERR ); @@ -67667,6 +68045,7 @@ static int walFindFrame( iRead = iFrame; } if( (nCollide--)==0 ){ + *piRead = 0; return SQLITE_CORRUPT_BKPT; } iKey = walNextHash(iKey); @@ -67970,7 +68349,7 @@ static int walRestartLog(Wal *pWal){ cnt = 0; do{ int notUsed; - rc = walTryBeginRead(pWal, ¬Used, 1, ++cnt); + rc = walTryBeginRead(pWal, ¬Used, 1, &cnt); }while( rc==WAL_RETRY ); assert( (rc&0xff)!=SQLITE_BUSY ); /* BUSY not possible when useWal==1 */ testcase( (rc&0xff)==SQLITE_IOERR ); @@ -68391,10 +68770,9 @@ SQLITE_PRIVATE int sqlite3WalCheckpoint( if( pWal->readOnly ) return SQLITE_READONLY; WALTRACE(("WAL%p: checkpoint begins\n", pWal)); - /* Enable blocking locks, if possible. If blocking locks are successfully - ** enabled, set xBusy2=0 so that the busy-handler is never invoked. */ + /* Enable blocking locks, if possible. */ sqlite3WalDb(pWal, db); - (void)walEnableBlocking(pWal); + if( xBusy2 ) (void)walEnableBlocking(pWal); /* IMPLEMENTATION-OF: R-62028-47212 All calls obtain an exclusive ** "checkpoint" lock on the database file. @@ -68435,9 +68813,14 @@ SQLITE_PRIVATE int sqlite3WalCheckpoint( /* Read the wal-index header. */ SEH_TRY { if( rc==SQLITE_OK ){ + /* For a passive checkpoint, do not re-enable blocking locks after + ** reading the wal-index header. A passive checkpoint should not block + ** or invoke the busy handler. The only lock such a checkpoint may + ** attempt to obtain is a lock on a read-slot, and it should give up + ** immediately and do a partial checkpoint if it cannot obtain it. */ walDisableBlocking(pWal); rc = walIndexReadHdr(pWal, &isChanged); - (void)walEnableBlocking(pWal); + if( eMode2!=SQLITE_CHECKPOINT_PASSIVE ) (void)walEnableBlocking(pWal); if( isChanged && pWal->pDbFd->pMethods->iVersion>=3 ){ sqlite3OsUnfetch(pWal->pDbFd, 0, 0); } @@ -68774,7 +69157,7 @@ SQLITE_PRIVATE sqlite3_file *sqlite3WalFile(Wal *pWal){ ** 22 1 Min embedded payload fraction (must be 32) ** 23 1 Min leaf payload fraction (must be 32) ** 24 4 File change counter -** 28 4 Reserved for future use +** 28 4 The size of the database in pages ** 32 4 First freelist page ** 36 4 Number of freelist pages in the file ** 40 60 15 4-byte meta values passed to higher layers @@ -74901,7 +75284,6 @@ static int accessPayload( assert( aWrite>=pBufStart ); /* due to (6) */ memcpy(aSave, aWrite, 4); rc = sqlite3OsRead(fd, aWrite, a+4, (i64)pBt->pageSize*(nextPage-1)); - if( rc && nextPage>pBt->nPage ) rc = SQLITE_CORRUPT_BKPT; nextPage = get4byte(aWrite); memcpy(aWrite, aSave, 4); }else @@ -83412,7 +83794,7 @@ static int valueFromFunction( #endif assert( pFunc ); if( (pFunc->funcFlags & (SQLITE_FUNC_CONSTANT|SQLITE_FUNC_SLOCHNG))==0 - || (pFunc->funcFlags & SQLITE_FUNC_NEEDCOLL) + || (pFunc->funcFlags & (SQLITE_FUNC_NEEDCOLL|SQLITE_FUNC_RUNONLY))!=0 ){ return SQLITE_OK; } @@ -84136,10 +84518,11 @@ static int growOpArray(Vdbe *v, int nOp){ ** sqlite3CantopenError(lineno) */ static void test_addop_breakpoint(int pc, Op *pOp){ - static int n = 0; + static u64 n = 0; (void)pc; (void)pOp; n++; + if( n==LARGEST_UINT64 ) abort(); /* so that n is used, preventing a warning */ } #endif @@ -85324,6 +85707,10 @@ static void freeP4(sqlite3 *db, int p4type, void *p4){ if( db->pnBytesFreed==0 ) sqlite3VtabUnlock((VTable *)p4); break; } + case P4_TABLEREF: { + if( db->pnBytesFreed==0 ) sqlite3DeleteTable(db, (Table*)p4); + break; + } } } @@ -85451,7 +85838,7 @@ static void SQLITE_NOINLINE vdbeChangeP4Full( int n ){ if( pOp->p4type ){ - freeP4(p->db, pOp->p4type, pOp->p4.p); + assert( pOp->p4type > P4_FREE_IF_LE ); pOp->p4type = 0; pOp->p4.p = 0; } @@ -89574,7 +89961,15 @@ SQLITE_API int sqlite3_clear_bindings(sqlite3_stmt *pStmt){ int rc = SQLITE_OK; Vdbe *p = (Vdbe*)pStmt; #if SQLITE_THREADSAFE - sqlite3_mutex *mutex = ((Vdbe*)pStmt)->db->mutex; + sqlite3_mutex *mutex; +#endif +#ifdef SQLITE_ENABLE_API_ARMOR + if( pStmt==0 ){ + return SQLITE_MISUSE_BKPT; + } +#endif +#if SQLITE_THREADSAFE + mutex = p->db->mutex; #endif sqlite3_mutex_enter(mutex); for(i=0; inVar; i++){ @@ -89953,6 +90348,18 @@ SQLITE_API void sqlite3_result_subtype(sqlite3_context *pCtx, unsigned int eSubt #ifdef SQLITE_ENABLE_API_ARMOR if( pCtx==0 ) return; #endif +#if defined(SQLITE_STRICT_SUBTYPE) && SQLITE_STRICT_SUBTYPE+0!=0 + if( pCtx->pFunc!=0 + && (pCtx->pFunc->funcFlags & SQLITE_RESULT_SUBTYPE)==0 + ){ + char zErr[200]; + sqlite3_snprintf(sizeof(zErr), zErr, + "misuse of sqlite3_result_subtype() by %s()", + pCtx->pFunc->zName); + sqlite3_result_error(pCtx, zErr, -1); + return; + } +#endif /* SQLITE_STRICT_SUBTYPE */ pOut = pCtx->pOut; assert( sqlite3_mutex_held(pOut->db->mutex) ); pOut->eSubtype = eSubtype & 0xff; @@ -90352,9 +90759,8 @@ SQLITE_API int sqlite3_step(sqlite3_stmt *pStmt){ SQLITE_API void *sqlite3_user_data(sqlite3_context *p){ #ifdef SQLITE_ENABLE_API_ARMOR if( p==0 ) return 0; -#else - assert( p && p->pFunc ); #endif + assert( p && p->pFunc ); return p->pFunc->pUserData; } @@ -92271,11 +92677,12 @@ SQLITE_API int sqlite3_found_count = 0; ** sqlite3CantopenError(lineno) */ static void test_trace_breakpoint(int pc, Op *pOp, Vdbe *v){ - static int n = 0; + static u64 n = 0; (void)pc; (void)pOp; (void)v; n++; + if( n==LARGEST_UINT64 ) abort(); /* So that n is used, preventing a warning */ } #endif @@ -94172,7 +94579,7 @@ case OP_AddImm: { /* in1 */ pIn1 = &aMem[pOp->p1]; memAboutToChange(p, pIn1); sqlite3VdbeMemIntegerify(pIn1); - pIn1->u.i += pOp->p2; + *(u64*)&pIn1->u.i += (u64)pOp->p2; break; } @@ -100318,24 +100725,23 @@ case OP_VCheck: { /* out2 */ pOut = &aMem[pOp->p2]; sqlite3VdbeMemSetNull(pOut); /* Innocent until proven guilty */ - assert( pOp->p4type==P4_TABLE ); + assert( pOp->p4type==P4_TABLEREF ); pTab = pOp->p4.pTab; assert( pTab!=0 ); + assert( pTab->nTabRef>0 ); assert( IsVirtual(pTab) ); - assert( pTab->u.vtab.p!=0 ); + if( pTab->u.vtab.p==0 ) break; pVtab = pTab->u.vtab.p->pVtab; assert( pVtab!=0 ); pModule = pVtab->pModule; assert( pModule!=0 ); assert( pModule->iVersion>=4 ); assert( pModule->xIntegrity!=0 ); - pTab->nTabRef++; sqlite3VtabLock(pTab->u.vtab.p); assert( pOp->p1>=0 && pOp->p1nDb ); rc = pModule->xIntegrity(pVtab, db->aDb[pOp->p1].zDbSName, pTab->zName, pOp->p3, &zErr); sqlite3VtabUnlock(pTab->u.vtab.p); - sqlite3DeleteTable(db, pTab); if( rc ){ sqlite3_free(zErr); goto abort_due_to_error; @@ -100460,6 +100866,7 @@ case OP_VColumn: { /* ncycle */ const sqlite3_module *pModule; Mem *pDest; sqlite3_context sContext; + FuncDef nullFunc; VdbeCursor *pCur = p->apCsr[pOp->p1]; assert( pCur!=0 ); @@ -100477,6 +100884,9 @@ case OP_VColumn: { /* ncycle */ memset(&sContext, 0, sizeof(sContext)); sContext.pOut = pDest; sContext.enc = encoding; + nullFunc.pUserData = 0; + nullFunc.funcFlags = SQLITE_RESULT_SUBTYPE; + sContext.pFunc = &nullFunc; assert( pOp->p5==OPFLAG_NOCHNG || pOp->p5==0 ); if( pOp->p5 & OPFLAG_NOCHNG ){ sqlite3VdbeMemSetNull(pDest); @@ -100809,6 +101219,42 @@ case OP_ClrSubtype: { /* in1 */ break; } +/* Opcode: GetSubtype P1 P2 * * * +** Synopsis: r[P2] = r[P1].subtype +** +** Extract the subtype value from register P1 and write that subtype +** into register P2. If P1 has no subtype, then P1 gets a NULL. +*/ +case OP_GetSubtype: { /* in1 out2 */ + pIn1 = &aMem[pOp->p1]; + pOut = &aMem[pOp->p2]; + if( pIn1->flags & MEM_Subtype ){ + sqlite3VdbeMemSetInt64(pOut, pIn1->eSubtype); + }else{ + sqlite3VdbeMemSetNull(pOut); + } + break; +} + +/* Opcode: SetSubtype P1 P2 * * * +** Synopsis: r[P2].subtype = r[P1] +** +** Set the subtype value of register P2 to the integer from register P1. +** If P1 is NULL, clear the subtype from p2. +*/ +case OP_SetSubtype: { /* in1 out2 */ + pIn1 = &aMem[pOp->p1]; + pOut = &aMem[pOp->p2]; + if( pIn1->flags & MEM_Null ){ + pOut->flags &= ~MEM_Subtype; + }else{ + assert( pIn1->flags & MEM_Int ); + pOut->flags |= MEM_Subtype; + pOut->eSubtype = (u8)(pIn1->u.i & 0xff); + } + break; +} + /* Opcode: FilterAdd P1 * P3 P4 * ** Synopsis: filter(P1) += key(P3@P4) ** @@ -105857,6 +106303,7 @@ SQLITE_PRIVATE Bitmask sqlite3ExprColUsed(Expr *pExpr){ assert( ExprUseYTab(pExpr) ); pExTab = pExpr->y.pTab; assert( pExTab!=0 ); + assert( n < pExTab->nCol ); if( (pExTab->tabFlags & TF_HasGenerated)!=0 && (pExTab->aCol[n].colFlags & COLFLAG_GENERATED)!=0 ){ @@ -106433,6 +106880,7 @@ static int lookupName( sqlite3RecordErrorOffsetOfExpr(pParse->db, pExpr); pParse->checkSchema = 1; pTopNC->nNcErr++; + eNewExprOp = TK_NULL; } assert( pFJMatch==0 ); @@ -106459,7 +106907,7 @@ static int lookupName( ** If a generated column is referenced, set bits for every column ** of the table. */ - if( pExpr->iColumn>=0 && pMatch!=0 ){ + if( pExpr->iColumn>=0 && cnt==1 && pMatch!=0 ){ pMatch->colUsed |= sqlite3ExprColUsed(pExpr); } @@ -106924,11 +107372,12 @@ static int resolveExprStep(Walker *pWalker, Expr *pExpr){ while( pNC2 && sqlite3ReferencesSrcList(pParse, pExpr, pNC2->pSrcList)==0 ){ - pExpr->op2++; + pExpr->op2 += (1 + pNC2->nNestedSelect); pNC2 = pNC2->pNext; } assert( pDef!=0 || IN_RENAME_OBJECT ); if( pNC2 && pDef ){ + pExpr->op2 += pNC2->nNestedSelect; assert( SQLITE_FUNC_MINMAX==NC_MinMaxAgg ); assert( SQLITE_FUNC_ANYORDER==NC_OrderAgg ); testcase( (pDef->funcFlags & SQLITE_FUNC_MINMAX)!=0 ); @@ -107487,6 +107936,7 @@ static int resolveSelectStep(Walker *pWalker, Select *p){ /* Recursively resolve names in all subqueries in the FROM clause */ + if( pOuterNC ) pOuterNC->nNestedSelect++; for(i=0; ipSrc->nSrc; i++){ SrcItem *pItem = &p->pSrc->a[i]; if( pItem->pSelect && (pItem->pSelect->selFlags & SF_Resolved)==0 ){ @@ -107511,6 +107961,9 @@ static int resolveSelectStep(Walker *pWalker, Select *p){ } } } + if( pOuterNC && ALWAYS(pOuterNC->nNestedSelect>0) ){ + pOuterNC->nNestedSelect--; + } /* Set up the local name-context to pass to sqlite3ResolveExprNames() to ** resolve the result-set expression list. @@ -109098,9 +109551,7 @@ SQLITE_PRIVATE void sqlite3ExprAddFunctionOrderBy( assert( ExprUseXList(pExpr) ); if( pExpr->x.pList==0 || NEVER(pExpr->x.pList->nExpr==0) ){ /* Ignore ORDER BY on zero-argument aggregates */ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprListDelete, - pOrderBy); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, pOrderBy); return; } if( IsWindowFunc(pExpr) ){ @@ -109281,6 +109732,9 @@ static SQLITE_NOINLINE void sqlite3ExprDeleteNN(sqlite3 *db, Expr *p){ SQLITE_PRIVATE void sqlite3ExprDelete(sqlite3 *db, Expr *p){ if( p ) sqlite3ExprDeleteNN(db, p); } +SQLITE_PRIVATE void sqlite3ExprDeleteGeneric(sqlite3 *db, void *p){ + if( ALWAYS(p) ) sqlite3ExprDeleteNN(db, (Expr*)p); +} /* ** Clear both elements of an OnOrUsing object @@ -109306,9 +109760,7 @@ SQLITE_PRIVATE void sqlite3ClearOnOrUsing(sqlite3 *db, OnOrUsing *p){ ** pExpr to the pParse->pConstExpr list with a register number of 0. */ SQLITE_PRIVATE void sqlite3ExprDeferredDelete(Parse *pParse, Expr *pExpr){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprDelete, - pExpr); + sqlite3ParserAddCleanup(pParse, sqlite3ExprDeleteGeneric, pExpr); } /* Invoke sqlite3RenameExprUnmap() and sqlite3ExprDelete() on the @@ -110114,6 +110566,9 @@ static SQLITE_NOINLINE void exprListDeleteNN(sqlite3 *db, ExprList *pList){ SQLITE_PRIVATE void sqlite3ExprListDelete(sqlite3 *db, ExprList *pList){ if( pList ) exprListDeleteNN(db, pList); } +SQLITE_PRIVATE void sqlite3ExprListDeleteGeneric(sqlite3 *db, void *pList){ + if( ALWAYS(pList) ) exprListDeleteNN(db, (ExprList*)pList); +} /* ** Return the bitwise-OR of all Expr.flags fields in the given @@ -110613,9 +111068,10 @@ SQLITE_PRIVATE int sqlite3ExprCanBeNull(const Expr *p){ case TK_COLUMN: assert( ExprUseYTab(p) ); return ExprHasProperty(p, EP_CanBeNull) || - p->y.pTab==0 || /* Reference to column of index on expression */ + NEVER(p->y.pTab==0) || /* Reference to column of index on expr */ (p->iColumn>=0 && p->y.pTab->aCol!=0 /* Possible due to prior error */ + && ALWAYS(p->iColumny.pTab->nCol) && p->y.pTab->aCol[p->iColumn].notNull==0); default: return 1; @@ -113197,8 +113653,10 @@ SQLITE_PRIVATE void sqlite3ExprCode(Parse *pParse, Expr *pExpr, int target){ inReg = sqlite3ExprCodeTarget(pParse, pExpr, target); if( inReg!=target ){ u8 op; - if( ALWAYS(pExpr) - && (ExprHasProperty(pExpr,EP_Subquery) || pExpr->op==TK_REGISTER) + Expr *pX = sqlite3ExprSkipCollateAndLikely(pExpr); + testcase( pX!=pExpr ); + if( ALWAYS(pX) + && (ExprHasProperty(pX,EP_Subquery) || pX->op==TK_REGISTER) ){ op = OP_Copy; }else{ @@ -113918,8 +114376,8 @@ SQLITE_PRIVATE int sqlite3ExprListCompare(const ExprList *pA, const ExprList *pB */ SQLITE_PRIVATE int sqlite3ExprCompareSkip(Expr *pA,Expr *pB, int iTab){ return sqlite3ExprCompare(0, - sqlite3ExprSkipCollateAndLikely(pA), - sqlite3ExprSkipCollateAndLikely(pB), + sqlite3ExprSkipCollate(pA), + sqlite3ExprSkipCollate(pB), iTab); } @@ -114644,13 +115102,14 @@ static int analyzeAggregate(Walker *pWalker, Expr *pExpr){ case TK_AGG_FUNCTION: { if( (pNC->ncFlags & NC_InAggFunc)==0 && pWalker->walkerDepth==pExpr->op2 + && pExpr->pAggInfo==0 ){ /* Check to see if pExpr is a duplicate of another aggregate ** function that is already in the pAggInfo structure */ struct AggInfo_func *pItem = pAggInfo->aFunc; for(i=0; inFunc; i++, pItem++){ - if( pItem->pFExpr==pExpr ) break; + if( NEVER(pItem->pFExpr==pExpr) ) break; if( sqlite3ExprCompare(0, pItem->pFExpr, pExpr, -1)==0 ){ break; } @@ -114693,6 +115152,8 @@ static int analyzeAggregate(Walker *pWalker, Expr *pExpr){ }else{ pItem->bOBPayload = 1; } + pItem->bUseSubtype = + (pItem->pFunc->funcFlags & SQLITE_SUBTYPE)!=0; }else{ pItem->iOBTab = -1; } @@ -117459,9 +117920,9 @@ static void openStatTable( typedef struct StatAccum StatAccum; typedef struct StatSample StatSample; struct StatSample { - tRowcnt *anEq; /* sqlite_stat4.nEq */ tRowcnt *anDLt; /* sqlite_stat4.nDLt */ #ifdef SQLITE_ENABLE_STAT4 + tRowcnt *anEq; /* sqlite_stat4.nEq */ tRowcnt *anLt; /* sqlite_stat4.nLt */ union { i64 iRowid; /* Rowid in main table of the key */ @@ -117619,9 +118080,9 @@ static void statInit( /* Allocate the space required for the StatAccum object */ n = sizeof(*p) - + sizeof(tRowcnt)*nColUp /* StatAccum.anEq */ - + sizeof(tRowcnt)*nColUp; /* StatAccum.anDLt */ + + sizeof(tRowcnt)*nColUp; /* StatAccum.anDLt */ #ifdef SQLITE_ENABLE_STAT4 + n += sizeof(tRowcnt)*nColUp; /* StatAccum.anEq */ if( mxSample ){ n += sizeof(tRowcnt)*nColUp /* StatAccum.anLt */ + sizeof(StatSample)*(nCol+mxSample) /* StatAccum.aBest[], a[] */ @@ -117642,9 +118103,9 @@ static void statInit( p->nKeyCol = nKeyCol; p->nSkipAhead = 0; p->current.anDLt = (tRowcnt*)&p[1]; - p->current.anEq = &p->current.anDLt[nColUp]; #ifdef SQLITE_ENABLE_STAT4 + p->current.anEq = &p->current.anDLt[nColUp]; p->mxSample = p->nLimit==0 ? mxSample : 0; if( mxSample ){ u8 *pSpace; /* Allocated space not yet assigned */ @@ -117911,7 +118372,9 @@ static void statPush( if( p->nRow==0 ){ /* This is the first call to this function. Do initialization. */ +#ifdef SQLITE_ENABLE_STAT4 for(i=0; inCol; i++) p->current.anEq[i] = 1; +#endif }else{ /* Second and subsequent calls get processed here */ #ifdef SQLITE_ENABLE_STAT4 @@ -117920,15 +118383,17 @@ static void statPush( /* Update anDLt[], anLt[] and anEq[] to reflect the values that apply ** to the current row of the index. */ +#ifdef SQLITE_ENABLE_STAT4 for(i=0; icurrent.anEq[i]++; } +#endif for(i=iChng; inCol; i++){ p->current.anDLt[i]++; #ifdef SQLITE_ENABLE_STAT4 if( p->mxSample ) p->current.anLt[i] += p->current.anEq[i]; -#endif p->current.anEq[i] = 1; +#endif } } @@ -118062,7 +118527,9 @@ static void statGet( u64 iVal = (p->nRow + nDistinct - 1) / nDistinct; if( iVal==2 && p->nRow*10 <= nDistinct*11 ) iVal = 1; sqlite3_str_appendf(&sStat, " %llu", iVal); +#ifdef SQLITE_ENABLE_STAT4 assert( p->current.anEq[i] ); +#endif } sqlite3ResultStrAccum(context, &sStat); } @@ -118751,6 +119218,16 @@ static void decodeIntArray( while( z[0]!=0 && z[0]!=' ' ) z++; while( z[0]==' ' ) z++; } + + /* Set the bLowQual flag if the peak number of rows obtained + ** from a full equality match is so large that a full table scan + ** seems likely to be faster than using the index. + */ + if( aLog[0] > 66 /* Index has more than 100 rows */ + && aLog[0] <= aLog[nOut-1] /* And only a single value seen */ + ){ + pIndex->bLowQual = 1; + } } } @@ -120797,7 +121274,7 @@ SQLITE_PRIVATE void sqlite3ColumnSetExpr( */ SQLITE_PRIVATE Expr *sqlite3ColumnExpr(Table *pTab, Column *pCol){ if( pCol->iDflt==0 ) return 0; - if( NEVER(!IsOrdinaryTable(pTab)) ) return 0; + if( !IsOrdinaryTable(pTab) ) return 0; if( NEVER(pTab->u.tab.pDfltList==0) ) return 0; if( NEVER(pTab->u.tab.pDfltList->nExpriDflt) ) return 0; return pTab->u.tab.pDfltList->a[pCol->iDflt-1].pExpr; @@ -120950,6 +121427,9 @@ SQLITE_PRIVATE void sqlite3DeleteTable(sqlite3 *db, Table *pTable){ if( db->pnBytesFreed==0 && (--pTable->nTabRef)>0 ) return; deleteTable(db, pTable); } +SQLITE_PRIVATE void sqlite3DeleteTableGeneric(sqlite3 *db, void *pTable){ + sqlite3DeleteTable(db, (Table*)pTable); +} /* @@ -121487,7 +121967,8 @@ SQLITE_PRIVATE void sqlite3ColumnPropertiesFromName(Table *pTab, Column *pCol){ /* ** Clean up the data structures associated with the RETURNING clause. */ -static void sqlite3DeleteReturning(sqlite3 *db, Returning *pRet){ +static void sqlite3DeleteReturning(sqlite3 *db, void *pArg){ + Returning *pRet = (Returning*)pArg; Hash *pHash; pHash = &(db->aDb[1].pSchema->trigHash); sqlite3HashInsert(pHash, pRet->zName, 0); @@ -121529,8 +122010,7 @@ SQLITE_PRIVATE void sqlite3AddReturning(Parse *pParse, ExprList *pList){ pParse->u1.pReturning = pRet; pRet->pParse = pParse; pRet->pReturnEL = pList; - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3DeleteReturning, pRet); + sqlite3ParserAddCleanup(pParse, sqlite3DeleteReturning, pRet); testcase( pParse->earlyCleanup ); if( db->mallocFailed ) return; sqlite3_snprintf(sizeof(pRet->zName), pRet->zName, @@ -121729,7 +122209,8 @@ SQLITE_PRIVATE char sqlite3AffinityType(const char *zIn, Column *pCol){ assert( zIn!=0 ); while( zIn[0] ){ - h = (h<<8) + sqlite3UpperToLower[(*zIn)&0xff]; + u8 x = *(u8*)zIn; + h = (h<<8) + sqlite3UpperToLower[x]; zIn++; if( h==(('c'<<24)+('h'<<16)+('a'<<8)+'r') ){ /* CHAR */ aff = SQLITE_AFF_TEXT; @@ -125594,7 +126075,7 @@ SQLITE_PRIVATE void sqlite3Reindex(Parse *pParse, Token *pName1, Token *pName2){ if( iDb<0 ) return; z = sqlite3NameFromToken(db, pObjName); if( z==0 ) return; - zDb = db->aDb[iDb].zDbSName; + zDb = pName2->n ? db->aDb[iDb].zDbSName : 0; pTab = sqlite3FindTable(db, z, zDb); if( pTab ){ reindexTable(pParse, pTab, 0); @@ -125604,6 +126085,7 @@ SQLITE_PRIVATE void sqlite3Reindex(Parse *pParse, Token *pName1, Token *pName2){ pIndex = sqlite3FindIndex(db, z, zDb); sqlite3DbFree(db, z); if( pIndex ){ + iDb = sqlite3SchemaToIndex(db, pIndex->pTable->pSchema); sqlite3BeginWriteOperation(pParse, 0, iDb); sqlite3RefillIndex(pParse, pIndex, -1); return; @@ -125769,6 +126251,9 @@ SQLITE_PRIVATE void sqlite3WithDelete(sqlite3 *db, With *pWith){ sqlite3DbFree(db, pWith); } } +SQLITE_PRIVATE void sqlite3WithDeleteGeneric(sqlite3 *db, void *pWith){ + sqlite3WithDelete(db, (With*)pWith); +} #endif /* !defined(SQLITE_OMIT_CTE) */ /************** End of build.c ***********************************************/ @@ -138994,7 +139479,8 @@ SQLITE_PRIVATE void sqlite3Pragma( if( pVTab->pModule->iVersion<4 ) continue; if( pVTab->pModule->xIntegrity==0 ) continue; sqlite3VdbeAddOp3(v, OP_VCheck, i, 3, isQuick); - sqlite3VdbeAppendP4(v, pTab, P4_TABLE); + pTab->nTabRef++; + sqlite3VdbeAppendP4(v, pTab, P4_TABLEREF); a1 = sqlite3VdbeAddOp1(v, OP_IsNull, 3); VdbeCoverage(v); integrityCheckResultRow(v); sqlite3VdbeJumpHere(v, a1); @@ -141021,6 +141507,7 @@ static int sqlite3LockAndPrepare( assert( (rc&db->errMask)==rc ); db->busyHandler.nBusy = 0; sqlite3_mutex_leave(db->mutex); + assert( rc==SQLITE_OK || (*ppStmt)==0 ); return rc; } @@ -141418,6 +141905,9 @@ SQLITE_PRIVATE Select *sqlite3SelectNew( SQLITE_PRIVATE void sqlite3SelectDelete(sqlite3 *db, Select *p){ if( OK_IF_ALWAYS_TRUE(p) ) clearSelect(db, p, 1); } +SQLITE_PRIVATE void sqlite3SelectDeleteGeneric(sqlite3 *db, void *p){ + if( ALWAYS(p) ) clearSelect(db, (Select*)p, 1); +} /* ** Return a pointer to the right-most SELECT statement in a compound. @@ -143553,7 +144043,8 @@ SQLITE_PRIVATE void sqlite3SubqueryColumnTypes( NameContext sNC; assert( pSelect!=0 ); - assert( (pSelect->selFlags & SF_Resolved)!=0 ); + testcase( (pSelect->selFlags & SF_Resolved)==0 ); + assert( (pSelect->selFlags & SF_Resolved)!=0 || IN_RENAME_OBJECT ); assert( pTab->nCol==pSelect->pEList->nExpr || pParse->nErr>0 ); assert( aff==SQLITE_AFF_NONE || aff==SQLITE_AFF_BLOB ); if( db->mallocFailed || IN_RENAME_OBJECT ) return; @@ -144437,9 +144928,7 @@ static int multiSelect( pDest->iSdst = dest.iSdst; pDest->nSdst = dest.nSdst; if( pDelete ){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3SelectDelete, - pDelete); + sqlite3ParserAddCleanup(pParse, sqlite3SelectDeleteGeneric, pDelete); } return rc; } @@ -144990,8 +145479,7 @@ static int multiSelectOrderBy( /* Make arrangements to free the 2nd and subsequent arms of the compound ** after the parse has finished */ if( pSplit->pPrior ){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3SelectDelete, pSplit->pPrior); + sqlite3ParserAddCleanup(pParse, sqlite3SelectDeleteGeneric, pSplit->pPrior); } pSplit->pPrior = pPrior; pPrior->pNext = pSplit; @@ -145812,9 +146300,7 @@ static int flattenSubquery( Table *pTabToDel = pSubitem->pTab; if( pTabToDel->nTabRef==1 ){ Parse *pToplevel = sqlite3ParseToplevel(pParse); - sqlite3ParserAddCleanup(pToplevel, - (void(*)(sqlite3*,void*))sqlite3DeleteTable, - pTabToDel); + sqlite3ParserAddCleanup(pToplevel, sqlite3DeleteTableGeneric, pTabToDel); testcase( pToplevel->earlyCleanup ); }else{ pTabToDel->nTabRef--; @@ -146861,8 +147347,7 @@ static struct Cte *searchWith( SQLITE_PRIVATE With *sqlite3WithPush(Parse *pParse, With *pWith, u8 bFree){ if( pWith ){ if( bFree ){ - pWith = (With*)sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3WithDelete, + pWith = (With*)sqlite3ParserAddCleanup(pParse, sqlite3WithDeleteGeneric, pWith); if( pWith==0 ) return 0; } @@ -147606,10 +148091,11 @@ static void selectAddSubqueryTypeInfo(Walker *pWalker, Select *p){ SrcList *pTabList; SrcItem *pFrom; - assert( p->selFlags & SF_Resolved ); if( p->selFlags & SF_HasTypeInfo ) return; p->selFlags |= SF_HasTypeInfo; pParse = pWalker->pParse; + testcase( (p->selFlags & SF_Resolved)==0 ); + assert( (p->selFlags & SF_Resolved) || IN_RENAME_OBJECT ); pTabList = p->pSrc; for(i=0, pFrom=pTabList->a; inSrc; i++, pFrom++){ Table *pTab = pFrom->pTab; @@ -147894,6 +148380,7 @@ static void resetAccumulator(Parse *pParse, AggInfo *pAggInfo){ assert( pFunc->pFExpr->pLeft!=0 ); assert( pFunc->pFExpr->pLeft->op==TK_ORDER ); assert( ExprUseXList(pFunc->pFExpr->pLeft) ); + assert( pFunc->pFunc!=0 ); pOBList = pFunc->pFExpr->pLeft->x.pList; if( !pFunc->bOBUnique ){ nExtra++; /* One extra column for the OP_Sequence */ @@ -147903,6 +148390,9 @@ static void resetAccumulator(Parse *pParse, AggInfo *pAggInfo){ assert( ExprUseXList(pFunc->pFExpr) ); nExtra += pFunc->pFExpr->x.pList->nExpr; } + if( pFunc->bUseSubtype ){ + nExtra += pFunc->pFExpr->x.pList->nExpr; + } pKeyInfo = sqlite3KeyInfoFromExprList(pParse, pOBList, 0, nExtra); if( !pFunc->bOBUnique && pParse->nErr==0 ){ pKeyInfo->nKeyField++; @@ -147929,9 +148419,9 @@ static void finalizeAggFunctions(Parse *pParse, AggInfo *pAggInfo){ assert( ExprUseXList(pF->pFExpr) ); pList = pF->pFExpr->x.pList; if( pF->iOBTab>=0 ){ - /* For an ORDER BY aggregate, calls to OP_AggStep where deferred and - ** all content was stored in emphermal table pF->iOBTab. Extract that - ** content now (in ORDER BY order) and make all calls to OP_AggStep + /* For an ORDER BY aggregate, calls to OP_AggStep were deferred. Inputs + ** were stored in emphermal table pF->iOBTab. Here, we extract those + ** inputs (in ORDER BY order) and make all calls to OP_AggStep ** before doing the OP_AggFinal call. */ int iTop; /* Start of loop for extracting columns */ int nArg; /* Number of columns to extract */ @@ -147939,6 +148429,7 @@ static void finalizeAggFunctions(Parse *pParse, AggInfo *pAggInfo){ int regAgg; /* Extract into this array */ int j; /* Loop counter */ + assert( pF->pFunc!=0 ); nArg = pList->nExpr; regAgg = sqlite3GetTempRange(pParse, nArg); @@ -147955,6 +148446,15 @@ static void finalizeAggFunctions(Parse *pParse, AggInfo *pAggInfo){ for(j=nArg-1; j>=0; j--){ sqlite3VdbeAddOp3(v, OP_Column, pF->iOBTab, nKey+j, regAgg+j); } + if( pF->bUseSubtype ){ + int regSubtype = sqlite3GetTempReg(pParse); + int iBaseCol = nKey + nArg + (pF->bOBPayload==0 && pF->bOBUnique==0); + for(j=nArg-1; j>=0; j--){ + sqlite3VdbeAddOp3(v, OP_Column, pF->iOBTab, iBaseCol+j, regSubtype); + sqlite3VdbeAddOp2(v, OP_SetSubtype, regSubtype, regAgg+j); + } + sqlite3ReleaseTempReg(pParse, regSubtype); + } sqlite3VdbeAddOp3(v, OP_AggStep, 0, regAgg, AggInfoFuncReg(pAggInfo,i)); sqlite3VdbeAppendP4(v, pF->pFunc, P4_FUNCDEF); sqlite3VdbeChangeP5(v, (u8)nArg); @@ -148009,6 +148509,7 @@ static void updateAccumulator( ExprList *pList; assert( ExprUseXList(pF->pFExpr) ); assert( !IsWindowFunc(pF->pFExpr) ); + assert( pF->pFunc!=0 ); pList = pF->pFExpr->x.pList; if( ExprHasProperty(pF->pFExpr, EP_WinFunc) ){ Expr *pFilter = pF->pFExpr->y.pWin->pFilter; @@ -148053,6 +148554,9 @@ static void updateAccumulator( if( pF->bOBPayload ){ regAggSz += nArg; } + if( pF->bUseSubtype ){ + regAggSz += nArg; + } regAggSz++; /* One extra register to hold result of MakeRecord */ regAgg = sqlite3GetTempRange(pParse, regAggSz); regDistinct = regAgg; @@ -148065,6 +148569,14 @@ static void updateAccumulator( if( pF->bOBPayload ){ regDistinct = regAgg+jj; sqlite3ExprCodeExprList(pParse, pList, regDistinct, 0, SQLITE_ECEL_DUP); + jj += nArg; + } + if( pF->bUseSubtype ){ + int kk; + int regBase = pF->bOBPayload ? regDistinct : regAgg; + for(kk=0; kknExpr; @@ -148269,7 +148781,8 @@ static SrcItem *isSelfJoinView( /* ** Deallocate a single AggInfo object */ -static void agginfoFree(sqlite3 *db, AggInfo *p){ +static void agginfoFree(sqlite3 *db, void *pArg){ + AggInfo *p = (AggInfo*)pArg; sqlite3DbFree(db, p->aCol); sqlite3DbFree(db, p->aFunc); sqlite3DbFreeNN(db, p); @@ -148343,7 +148856,7 @@ static int countOfViewOptimization(Parse *pParse, Select *p){ pSub->selFlags |= SF_Aggregate; pSub->selFlags &= ~SF_Compound; pSub->nSelectRow = 0; - sqlite3ExprListDelete(db, pSub->pEList); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, pSub->pEList); pTerm = pPrior ? sqlite3ExprDup(db, pCount, 0) : pCount; pSub->pEList = sqlite3ExprListAppend(pParse, 0, pTerm); pTerm = sqlite3PExpr(pParse, TK_SELECT, 0, 0); @@ -148523,9 +149036,8 @@ SQLITE_PRIVATE int sqlite3Select( sqlite3TreeViewExprList(0, p->pOrderBy, 0, "ORDERBY"); } #endif - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprListDelete, - p->pOrderBy); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, + p->pOrderBy); testcase( pParse->earlyCleanup ); p->pOrderBy = 0; } @@ -148631,6 +149143,7 @@ SQLITE_PRIVATE int sqlite3Select( TREETRACE(0x1000,pParse,p, ("LEFT-JOIN simplifies to JOIN on term %d\n",i)); pItem->fg.jointype &= ~(JT_LEFT|JT_OUTER); + unsetJoinExpr(p->pWhere, pItem->iCursor, 0); } } if( pItem->fg.jointype & JT_LTORJ ){ @@ -148645,17 +149158,15 @@ SQLITE_PRIVATE int sqlite3Select( TREETRACE(0x1000,pParse,p, ("RIGHT-JOIN simplifies to JOIN on term %d\n",j)); pI2->fg.jointype &= ~(JT_RIGHT|JT_OUTER); + unsetJoinExpr(p->pWhere, pI2->iCursor, 1); } } } - for(j=pTabList->nSrc-1; j>=i; j--){ + for(j=pTabList->nSrc-1; j>=0; j--){ pTabList->a[j].fg.jointype &= ~JT_LTORJ; if( pTabList->a[j].fg.jointype & JT_RIGHT ) break; } } - assert( pItem->iCursor>=0 ); - unsetJoinExpr(p->pWhere, pItem->iCursor, - pTabList->a[0].fg.jointype & JT_LTORJ); } /* No further action if this term of the FROM clause is not a subquery */ @@ -148718,9 +149229,8 @@ SQLITE_PRIVATE int sqlite3Select( ){ TREETRACE(0x800,pParse,p, ("omit superfluous ORDER BY on %r FROM-clause subquery\n",i+1)); - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))sqlite3ExprListDelete, - pSub->pOrderBy); + sqlite3ParserAddCleanup(pParse, sqlite3ExprListDeleteGeneric, + pSub->pOrderBy); pSub->pOrderBy = 0; } @@ -149249,8 +149759,7 @@ SQLITE_PRIVATE int sqlite3Select( */ pAggInfo = sqlite3DbMallocZero(db, sizeof(*pAggInfo) ); if( pAggInfo ){ - sqlite3ParserAddCleanup(pParse, - (void(*)(sqlite3*,void*))agginfoFree, pAggInfo); + sqlite3ParserAddCleanup(pParse, agginfoFree, pAggInfo); testcase( pParse->earlyCleanup ); } if( db->mallocFailed ){ @@ -153899,7 +154408,6 @@ SQLITE_PRIVATE void sqlite3VtabUnlockList(sqlite3 *db){ if( p ){ db->pDisconnect = 0; - sqlite3ExpirePreparedStatements(db, 0); do { VTable *pNext = p->pNext; sqlite3VtabUnlock(p); @@ -155465,7 +155973,7 @@ SQLITE_PRIVATE Bitmask sqlite3WhereGetMask(WhereMaskSet*,int); #ifdef WHERETRACE_ENABLED SQLITE_PRIVATE void sqlite3WhereClausePrint(WhereClause *pWC); SQLITE_PRIVATE void sqlite3WhereTermPrint(WhereTerm *pTerm, int iTerm); -SQLITE_PRIVATE void sqlite3WhereLoopPrint(WhereLoop *p, WhereClause *pWC); +SQLITE_PRIVATE void sqlite3WhereLoopPrint(const WhereLoop *p, const WhereClause *pWC); #endif SQLITE_PRIVATE WhereTerm *sqlite3WhereFindTerm( WhereClause *pWC, /* The WHERE clause to be searched */ @@ -160927,12 +161435,22 @@ static void translateColumnToCopy( for(; iStartp1!=iTabCur ) continue; if( pOp->opcode==OP_Column ){ +#ifdef SQLITE_DEBUG + if( pParse->db->flags & SQLITE_VdbeAddopTrace ){ + printf("TRANSLATE OP_Column to OP_Copy at %d\n", iStart); + } +#endif pOp->opcode = OP_Copy; pOp->p1 = pOp->p2 + iRegister; pOp->p2 = pOp->p3; pOp->p3 = 0; pOp->p5 = 2; /* Cause the MEM_Subtype flag to be cleared */ }else if( pOp->opcode==OP_Rowid ){ +#ifdef SQLITE_DEBUG + if( pParse->db->flags & SQLITE_VdbeAddopTrace ){ + printf("TRANSLATE OP_Rowid to OP_Sequence at %d\n", iStart); + } +#endif pOp->opcode = OP_Sequence; pOp->p1 = iAutoidxCur; #ifdef SQLITE_ALLOW_ROWID_IN_VIEW @@ -162259,7 +162777,8 @@ static int whereRangeScanEst( ** sample, then assume they are 4x more selective. This brings ** the estimated selectivity more in line with what it would be ** if estimated without the use of STAT4 tables. */ - if( iLwrIdx==iUprIdx ) nNew -= 20; assert( 20==sqlite3LogEst(4) ); + if( iLwrIdx==iUprIdx ){ nNew -= 20; } + assert( 20==sqlite3LogEst(4) ); }else{ nNew = 10; assert( 10==sqlite3LogEst(2) ); } @@ -162483,17 +163002,34 @@ SQLITE_PRIVATE void sqlite3WhereClausePrint(WhereClause *pWC){ #ifdef WHERETRACE_ENABLED /* ** Print a WhereLoop object for debugging purposes -*/ -SQLITE_PRIVATE void sqlite3WhereLoopPrint(WhereLoop *p, WhereClause *pWC){ - WhereInfo *pWInfo = pWC->pWInfo; - int nb = 1+(pWInfo->pTabList->nSrc+3)/4; - SrcItem *pItem = pWInfo->pTabList->a + p->iTab; - Table *pTab = pItem->pTab; - Bitmask mAll = (((Bitmask)1)<<(nb*4)) - 1; - sqlite3DebugPrintf("%c%2d.%0*llx.%0*llx", p->cId, - p->iTab, nb, p->maskSelf, nb, p->prereq & mAll); - sqlite3DebugPrintf(" %12s", - pItem->zAlias ? pItem->zAlias : pTab->zName); +** +** Format example: +** +** .--- Position in WHERE clause rSetup, rRun, nOut ---. +** | | +** | .--- selfMask nTerm ------. | +** | | | | +** | | .-- prereq Idx wsFlags----. | | +** | | | Name | | | +** | | | __|__ nEq ---. ___|__ | __|__ +** | / \ / \ / \ | / \ / \ / \ +** 1.002.001 t2.t2xy 2 f 010241 N 2 cost 0,56,31 +*/ +SQLITE_PRIVATE void sqlite3WhereLoopPrint(const WhereLoop *p, const WhereClause *pWC){ + if( pWC ){ + WhereInfo *pWInfo = pWC->pWInfo; + int nb = 1+(pWInfo->pTabList->nSrc+3)/4; + SrcItem *pItem = pWInfo->pTabList->a + p->iTab; + Table *pTab = pItem->pTab; + Bitmask mAll = (((Bitmask)1)<<(nb*4)) - 1; + sqlite3DebugPrintf("%c%2d.%0*llx.%0*llx", p->cId, + p->iTab, nb, p->maskSelf, nb, p->prereq & mAll); + sqlite3DebugPrintf(" %12s", + pItem->zAlias ? pItem->zAlias : pTab->zName); + }else{ + sqlite3DebugPrintf("%c%2d.%03llx.%03llx %c%d", + p->cId, p->iTab, p->maskSelf, p->prereq & 0xfff, p->cId, p->iTab); + } if( (p->wsFlags & WHERE_VIRTUALTABLE)==0 ){ const char *zName; if( p->u.btree.pIndex && (zName = p->u.btree.pIndex->zName)!=0 ){ @@ -162530,6 +163066,15 @@ SQLITE_PRIVATE void sqlite3WhereLoopPrint(WhereLoop *p, WhereClause *pWC){ } } } +SQLITE_PRIVATE void sqlite3ShowWhereLoop(const WhereLoop *p){ + if( p ) sqlite3WhereLoopPrint(p, 0); +} +SQLITE_PRIVATE void sqlite3ShowWhereLoopList(const WhereLoop *p){ + while( p ){ + sqlite3ShowWhereLoop(p); + p = p->pNextLoop; + } +} #endif /* @@ -162642,46 +163187,60 @@ static void whereInfoFree(sqlite3 *db, WhereInfo *pWInfo){ } /* -** Return TRUE if all of the following are true: +** Return TRUE if X is a proper subset of Y but is of equal or less cost. +** In other words, return true if all constraints of X are also part of Y +** and Y has additional constraints that might speed the search that X lacks +** but the cost of running X is not more than the cost of running Y. +** +** In other words, return true if the cost relationwship between X and Y +** is inverted and needs to be adjusted. +** +** Case 1: ** -** (1) X has the same or lower cost, or returns the same or fewer rows, -** than Y. -** (2) X uses fewer WHERE clause terms than Y -** (3) Every WHERE clause term used by X is also used by Y -** (4) X skips at least as many columns as Y -** (5) If X is a covering index, than Y is too +** (1a) X and Y use the same index. +** (1b) X has fewer == terms than Y +** (1c) Neither X nor Y use skip-scan +** (1d) X does not have a a greater cost than Y ** -** Conditions (2) and (3) mean that X is a "proper subset" of Y. -** If X is a proper subset of Y then Y is a better choice and ought -** to have a lower cost. This routine returns TRUE when that cost -** relationship is inverted and needs to be adjusted. Constraint (4) -** was added because if X uses skip-scan less than Y it still might -** deserve a lower cost even if it is a proper subset of Y. Constraint (5) -** was added because a covering index probably deserves to have a lower cost -** than a non-covering index even if it is a proper subset. +** Case 2: +** +** (2a) X has the same or lower cost, or returns the same or fewer rows, +** than Y. +** (2b) X uses fewer WHERE clause terms than Y +** (2c) Every WHERE clause term used by X is also used by Y +** (2d) X skips at least as many columns as Y +** (2e) If X is a covering index, than Y is too */ static int whereLoopCheaperProperSubset( const WhereLoop *pX, /* First WhereLoop to compare */ const WhereLoop *pY /* Compare against this WhereLoop */ ){ int i, j; + if( pX->rRun>pY->rRun && pX->nOut>pY->nOut ) return 0; /* (1d) and (2a) */ + assert( (pX->wsFlags & WHERE_VIRTUALTABLE)==0 ); + assert( (pY->wsFlags & WHERE_VIRTUALTABLE)==0 ); + if( pX->u.btree.nEq < pY->u.btree.nEq /* (1b) */ + && pX->u.btree.pIndex==pY->u.btree.pIndex /* (1a) */ + && pX->nSkip==0 && pY->nSkip==0 /* (1c) */ + ){ + return 1; /* Case 1 is true */ + } if( pX->nLTerm-pX->nSkip >= pY->nLTerm-pY->nSkip ){ - return 0; /* X is not a subset of Y */ + return 0; /* (2b) */ } - if( pX->rRun>pY->rRun && pX->nOut>pY->nOut ) return 0; - if( pY->nSkip > pX->nSkip ) return 0; + if( pY->nSkip > pX->nSkip ) return 0; /* (2d) */ for(i=pX->nLTerm-1; i>=0; i--){ if( pX->aLTerm[i]==0 ) continue; for(j=pY->nLTerm-1; j>=0; j--){ if( pY->aLTerm[j]==pX->aLTerm[i] ) break; } - if( j<0 ) return 0; /* X not a subset of Y since term X[i] not used by Y */ + if( j<0 ) return 0; /* (2c) */ } if( (pX->wsFlags&WHERE_IDX_ONLY)!=0 && (pY->wsFlags&WHERE_IDX_ONLY)==0 ){ - return 0; /* Constraint (5) */ + return 0; /* (2e) */ } - return 1; /* All conditions meet */ + return 1; /* Case 2 is true */ } /* @@ -163171,7 +163730,10 @@ static int whereLoopAddBtreeIndex( assert( pNew->u.btree.nBtm==0 ); opMask = WO_EQ|WO_IN|WO_GT|WO_GE|WO_LT|WO_LE|WO_ISNULL|WO_IS; } - if( pProbe->bUnordered ) opMask &= ~(WO_GT|WO_GE|WO_LT|WO_LE); + if( pProbe->bUnordered || pProbe->bLowQual ){ + if( pProbe->bUnordered ) opMask &= ~(WO_GT|WO_GE|WO_LT|WO_LE); + if( pProbe->bLowQual ) opMask &= ~(WO_EQ|WO_IN|WO_IS); + } assert( pNew->u.btree.nEqnColumn ); assert( pNew->u.btree.nEqnKeyCol @@ -166059,6 +166621,20 @@ static SQLITE_NOINLINE void whereAddIndexedExpr( continue; } if( sqlite3ExprIsConstant(pExpr) ) continue; + if( pExpr->op==TK_FUNCTION ){ + /* Functions that might set a subtype should not be replaced by the + ** value taken from an expression index since the index omits the + ** subtype. https://sqlite.org/forum/forumpost/68d284c86b082c3e */ + int n; + FuncDef *pDef; + sqlite3 *db = pParse->db; + assert( ExprUseXList(pExpr) ); + n = pExpr->x.pList ? pExpr->x.pList->nExpr : 0; + pDef = sqlite3FindFunction(db, pExpr->u.zToken, n, ENC(db), 0); + if( pDef==0 || (pDef->funcFlags & SQLITE_RESULT_SUBTYPE)!=0 ){ + continue; + } + } p = sqlite3DbMallocRaw(pParse->db, sizeof(IndexedExpr)); if( p==0 ) break; p->pIENext = pParse->pIdxEpr; @@ -166262,7 +166838,10 @@ SQLITE_PRIVATE WhereInfo *sqlite3WhereBegin( ** field (type Bitmask) it must be aligned on an 8-byte boundary on ** some architectures. Hence the ROUND8() below. */ - nByteWInfo = ROUND8P(sizeof(WhereInfo)+(nTabList-1)*sizeof(WhereLevel)); + nByteWInfo = ROUND8P(sizeof(WhereInfo)); + if( nTabList>1 ){ + nByteWInfo = ROUND8P(nByteWInfo + (nTabList-1)*sizeof(WhereLevel)); + } pWInfo = sqlite3DbMallocRawNN(db, nByteWInfo + sizeof(WhereLoop)); if( db->mallocFailed ){ sqlite3DbFree(db, pWInfo); @@ -166824,6 +167403,11 @@ SQLITE_PRIVATE WhereInfo *sqlite3WhereBegin( pParse->nQueryLoop = pWInfo->savedNQueryLoop; whereInfoFree(db, pWInfo); } +#ifdef WHERETRACE_ENABLED + /* Prevent harmless compiler warnings about debugging routines + ** being declared but never used */ + sqlite3ShowWhereLoopList(0); +#endif /* WHERETRACE_ENABLED */ return 0; } @@ -168241,7 +168825,7 @@ SQLITE_PRIVATE int sqlite3WindowRewrite(Parse *pParse, Select *p){ assert( ExprUseXList(pWin->pOwner) ); assert( pWin->pWFunc!=0 ); pArgs = pWin->pOwner->x.pList; - if( pWin->pWFunc->funcFlags & SQLITE_FUNC_SUBTYPE ){ + if( pWin->pWFunc->funcFlags & SQLITE_SUBTYPE ){ selectWindowRewriteEList(pParse, pMWin, pSrc, pArgs, pTab, &pSublist); pWin->iArgCol = (pSublist ? pSublist->nExpr : 0); pWin->bExprArgs = 1; @@ -179415,7 +179999,7 @@ SQLITE_PRIVATE int sqlite3CreateFunc( assert( SQLITE_FUNC_CONSTANT==SQLITE_DETERMINISTIC ); assert( SQLITE_FUNC_DIRECT==SQLITE_DIRECTONLY ); extraFlags = enc & (SQLITE_DETERMINISTIC|SQLITE_DIRECTONLY| - SQLITE_SUBTYPE|SQLITE_INNOCUOUS); + SQLITE_SUBTYPE|SQLITE_INNOCUOUS|SQLITE_RESULT_SUBTYPE); enc &= (SQLITE_FUNC_ENCMASK|SQLITE_ANY); /* The SQLITE_INNOCUOUS flag is the same bit as SQLITE_FUNC_UNSAFE. But @@ -182160,6 +182744,28 @@ SQLITE_API int sqlite3_test_control(int op, ...){ break; } #endif + + /* sqlite3_test_control(SQLITE_TESTCTRL_JSON_SELFCHECK, &onOff); + ** + ** Activate or deactivate validation of JSONB that is generated from + ** text. Off by default, as the validation is slow. Validation is + ** only available if compiled using SQLITE_DEBUG. + ** + ** If onOff is initially 1, then turn it on. If onOff is initially + ** off, turn it off. If onOff is initially -1, then change onOff + ** to be the current setting. + */ + case SQLITE_TESTCTRL_JSON_SELFCHECK: { +#if defined(SQLITE_DEBUG) && !defined(SQLITE_OMIT_WSD) + int *pOnOff = va_arg(ap, int*); + if( *pOnOff<0 ){ + *pOnOff = sqlite3Config.bJsonSelfcheck; + }else{ + sqlite3Config.bJsonSelfcheck = (u8)((*pOnOff)&0xff); + } +#endif + break; + } } va_end(ap); #endif /* SQLITE_UNTESTABLE */ @@ -202574,24 +203180,145 @@ SQLITE_PRIVATE int sqlite3FtsUnicodeFold(int c, int eRemoveDiacritic){ ** ****************************************************************************** ** -** This SQLite JSON functions. +** SQLite JSON functions. ** ** This file began as an extension in ext/misc/json1.c in 2015. That ** extension proved so useful that it has now been moved into the core. ** -** For the time being, all JSON is stored as pure text. (We might add -** a JSONB type in the future which stores a binary encoding of JSON in -** a BLOB, but there is no support for JSONB in the current implementation. -** This implementation parses JSON text at 250 MB/s, so it is hard to see -** how JSONB might improve on that.) +** The original design stored all JSON as pure text, canonical RFC-8259. +** Support for JSON-5 extensions was added with version 3.42.0 (2023-05-16). +** All generated JSON text still conforms strictly to RFC-8259, but text +** with JSON-5 extensions is accepted as input. +** +** Beginning with version 3.45.0 (circa 2024-01-01), these routines also +** accept BLOB values that have JSON encoded using a binary representation +** called "JSONB". The name JSONB comes from PostgreSQL, however the on-disk +** format SQLite JSONB is completely different and incompatible with +** PostgreSQL JSONB. +** +** Decoding and interpreting JSONB is still O(N) where N is the size of +** the input, the same as text JSON. However, the constant of proportionality +** for JSONB is much smaller due to faster parsing. The size of each +** element in JSONB is encoded in its header, so there is no need to search +** for delimiters using persnickety syntax rules. JSONB seems to be about +** 3x faster than text JSON as a result. JSONB is also tends to be slightly +** smaller than text JSON, by 5% or 10%, but there are corner cases where +** JSONB can be slightly larger. So you are not far mistaken to say that +** a JSONB blob is the same size as the equivalent RFC-8259 text. +** +** +** THE JSONB ENCODING: +** +** Every JSON element is encoded in JSONB as a header and a payload. +** The header is between 1 and 9 bytes in size. The payload is zero +** or more bytes. +** +** The lower 4 bits of the first byte of the header determines the +** element type: +** +** 0: NULL +** 1: TRUE +** 2: FALSE +** 3: INT -- RFC-8259 integer literal +** 4: INT5 -- JSON5 integer literal +** 5: FLOAT -- RFC-8259 floating point literal +** 6: FLOAT5 -- JSON5 floating point literal +** 7: TEXT -- Text literal acceptable to both SQL and JSON +** 8: TEXTJ -- Text containing RFC-8259 escapes +** 9: TEXT5 -- Text containing JSON5 and/or RFC-8259 escapes +** 10: TEXTRAW -- Text containing unescaped syntax characters +** 11: ARRAY +** 12: OBJECT +** +** The other three possible values (13-15) are reserved for future +** enhancements. +** +** The upper 4 bits of the first byte determine the size of the header +** and sometimes also the size of the payload. If X is the first byte +** of the element and if X>>4 is between 0 and 11, then the payload +** will be that many bytes in size and the header is exactly one byte +** in size. Other four values for X>>4 (12-15) indicate that the header +** is more than one byte in size and that the payload size is determined +** by the remainder of the header, interpreted as a unsigned big-endian +** integer. +** +** Value of X>>4 Size integer Total header size +** ------------- -------------------- ----------------- +** 12 1 byte (0-255) 2 +** 13 2 byte (0-65535) 3 +** 14 4 byte (0-4294967295) 5 +** 15 8 byte (0-1.8e19) 9 +** +** The payload size need not be expressed in its minimal form. For example, +** if the payload size is 10, the size can be expressed in any of 5 different +** ways: (1) (X>>4)==10, (2) (X>>4)==12 following by on 0x0a byte, +** (3) (X>>4)==13 followed by 0x00 and 0x0a, (4) (X>>4)==14 followed by +** 0x00 0x00 0x00 0x0a, or (5) (X>>4)==15 followed by 7 bytes of 0x00 and +** a single byte of 0x0a. The shorter forms are preferred, of course, but +** sometimes when generating JSONB, the payload size is not known in advance +** and it is convenient to reserve sufficient header space to cover the +** largest possible payload size and then come back later and patch up +** the size when it becomes known, resulting in a non-minimal encoding. +** +** The value (X>>4)==15 is not actually used in the current implementation +** (as SQLite is currently unable handle BLOBs larger than about 2GB) +** but is included in the design to allow for future enhancements. +** +** The payload follows the header. NULL, TRUE, and FALSE have no payload and +** their payload size must always be zero. The payload for INT, INT5, +** FLOAT, FLOAT5, TEXT, TEXTJ, TEXT5, and TEXTROW is text. Note that the +** "..." or '...' delimiters are omitted from the various text encodings. +** The payload for ARRAY and OBJECT is a list of additional elements that +** are the content for the array or object. The payload for an OBJECT +** must be an even number of elements. The first element of each pair is +** the label and must be of type TEXT, TEXTJ, TEXT5, or TEXTRAW. +** +** A valid JSONB blob consists of a single element, as described above. +** Usually this will be an ARRAY or OBJECT element which has many more +** elements as its content. But the overall blob is just a single element. +** +** Input validation for JSONB blobs simply checks that the element type +** code is between 0 and 12 and that the total size of the element +** (header plus payload) is the same as the size of the BLOB. If those +** checks are true, the BLOB is assumed to be JSONB and processing continues. +** Errors are only raised if some other miscoding is discovered during +** processing. +** +** Additional information can be found in the doc/jsonb.md file of the +** canonical SQLite source tree. */ #ifndef SQLITE_OMIT_JSON /* #include "sqliteInt.h" */ +/* JSONB element types +*/ +#define JSONB_NULL 0 /* "null" */ +#define JSONB_TRUE 1 /* "true" */ +#define JSONB_FALSE 2 /* "false" */ +#define JSONB_INT 3 /* integer acceptable to JSON and SQL */ +#define JSONB_INT5 4 /* integer in 0x000 notation */ +#define JSONB_FLOAT 5 /* float acceptable to JSON and SQL */ +#define JSONB_FLOAT5 6 /* float with JSON5 extensions */ +#define JSONB_TEXT 7 /* Text compatible with both JSON and SQL */ +#define JSONB_TEXTJ 8 /* Text with JSON escapes */ +#define JSONB_TEXT5 9 /* Text with JSON-5 escape */ +#define JSONB_TEXTRAW 10 /* SQL text that needs escaping for JSON */ +#define JSONB_ARRAY 11 /* An array */ +#define JSONB_OBJECT 12 /* An object */ + +/* Human-readable names for the JSONB values. The index for each +** string must correspond to the JSONB_* integer above. +*/ +static const char * const jsonbType[] = { + "null", "true", "false", "integer", "integer", + "real", "real", "text", "text", "text", + "text", "array", "object", "", "", "", "" +}; + /* ** Growing our own isspace() routine this way is twice as fast as ** the library isspace() function, resulting in a 7% overall performance -** increase for the parser. (Ubuntu14.10 gcc 4.8.4 x64 with -Os). +** increase for the text-JSON parser. (Ubuntu14.10 gcc 4.8.4 x64 with -Os). */ static const char jsonIsSpace[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, @@ -202612,11 +203339,19 @@ static const char jsonIsSpace[] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, }; -#define fast_isspace(x) (jsonIsSpace[(unsigned char)x]) +#define jsonIsspace(x) (jsonIsSpace[(unsigned char)x]) /* -** Characters that are special to JSON. Control charaters, -** '"' and '\\'. +** The set of all space characters recognized by jsonIsspace(). +** Useful as the second argument to strspn(). +*/ +static const char jsonSpaces[] = "\011\012\015\040"; + +/* +** Characters that are special to JSON. Control characters, +** '"' and '\\' and '\''. Actually, '\'' is not special to +** canonical JSON, but it is special in JSON-5, so we include +** it in the set of special characters. */ static const char jsonIsOk[256] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, @@ -202638,22 +203373,49 @@ static const char jsonIsOk[256] = { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }; - -#if !defined(SQLITE_DEBUG) && !defined(SQLITE_COVERAGE_TEST) -# define VVA(X) -#else -# define VVA(X) X -#endif - /* Objects */ +typedef struct JsonCache JsonCache; typedef struct JsonString JsonString; -typedef struct JsonNode JsonNode; typedef struct JsonParse JsonParse; -typedef struct JsonCleanup JsonCleanup; + +/* +** Magic number used for the JSON parse cache in sqlite3_get_auxdata() +*/ +#define JSON_CACHE_ID (-429938) /* Cache entry */ +#define JSON_CACHE_SIZE 4 /* Max number of cache entries */ + +/* +** jsonUnescapeOneChar() returns this invalid code point if it encounters +** a syntax error. +*/ +#define JSON_INVALID_CHAR 0x99999 + +/* A cache mapping JSON text into JSONB blobs. +** +** Each cache entry is a JsonParse object with the following restrictions: +** +** * The bReadOnly flag must be set +** +** * The aBlob[] array must be owned by the JsonParse object. In other +** words, nBlobAlloc must be non-zero. +** +** * eEdit and delta must be zero. +** +** * zJson must be an RCStr. In other words bJsonIsRCStr must be true. +*/ +struct JsonCache { + sqlite3 *db; /* Database connection */ + int nUsed; /* Number of active entries in the cache */ + JsonParse *a[JSON_CACHE_SIZE]; /* One line for each cache entry */ +}; /* An instance of this object represents a JSON string ** under construction. Really, this is a generic string accumulator ** that can be and is used to create strings other than JSON. +** +** If the generated string is longer than will fit into the zSpace[] buffer, +** then it will be an RCStr string. This aids with caching of large +** JSON strings. */ struct JsonString { sqlite3_context *pCtx; /* Function context - put error messages here */ @@ -202661,121 +203423,75 @@ struct JsonString { u64 nAlloc; /* Bytes of storage available in zBuf[] */ u64 nUsed; /* Bytes of zBuf[] currently used */ u8 bStatic; /* True if zBuf is static space */ - u8 bErr; /* True if an error has been encountered */ + u8 eErr; /* True if an error has been encountered */ char zSpace[100]; /* Initial static space */ }; -/* A deferred cleanup task. A list of JsonCleanup objects might be -** run when the JsonParse object is destroyed. -*/ -struct JsonCleanup { - JsonCleanup *pJCNext; /* Next in a list */ - void (*xOp)(void*); /* Routine to run */ - void *pArg; /* Argument to xOp() */ -}; +/* Allowed values for JsonString.eErr */ +#define JSTRING_OOM 0x01 /* Out of memory */ +#define JSTRING_MALFORMED 0x02 /* Malformed JSONB */ +#define JSTRING_ERR 0x04 /* Error already sent to sqlite3_result */ -/* JSON type values +/* The "subtype" set for text JSON values passed through using +** sqlite3_result_subtype() and sqlite3_value_subtype(). */ -#define JSON_SUBST 0 /* Special edit node. Uses u.iPrev */ -#define JSON_NULL 1 -#define JSON_TRUE 2 -#define JSON_FALSE 3 -#define JSON_INT 4 -#define JSON_REAL 5 -#define JSON_STRING 6 -#define JSON_ARRAY 7 -#define JSON_OBJECT 8 - -/* The "subtype" set for JSON values */ #define JSON_SUBTYPE 74 /* Ascii for "J" */ /* -** Names of the various JSON types: -*/ -static const char * const jsonType[] = { - "subst", - "null", "true", "false", "integer", "real", "text", "array", "object" -}; - -/* Bit values for the JsonNode.jnFlag field -*/ -#define JNODE_RAW 0x01 /* Content is raw, not JSON encoded */ -#define JNODE_ESCAPE 0x02 /* Content is text with \ escapes */ -#define JNODE_REMOVE 0x04 /* Do not output */ -#define JNODE_REPLACE 0x08 /* Target of a JSON_SUBST node */ -#define JNODE_APPEND 0x10 /* More ARRAY/OBJECT entries at u.iAppend */ -#define JNODE_LABEL 0x20 /* Is a label of an object */ -#define JNODE_JSON5 0x40 /* Node contains JSON5 enhancements */ - - -/* A single node of parsed JSON. An array of these nodes describes -** a parse of JSON + edits. -** -** Use the json_parse() SQL function (available when compiled with -** -DSQLITE_DEBUG) to see a dump of complete JsonParse objects, including -** a complete listing and decoding of the array of JsonNodes. +** Bit values for the flags passed into various SQL function implementations +** via the sqlite3_user_data() value. */ -struct JsonNode { - u8 eType; /* One of the JSON_ type values */ - u8 jnFlags; /* JNODE flags */ - u8 eU; /* Which union element to use */ - u32 n; /* Bytes of content for INT, REAL or STRING - ** Number of sub-nodes for ARRAY and OBJECT - ** Node that SUBST applies to */ - union { - const char *zJContent; /* 1: Content for INT, REAL, and STRING */ - u32 iAppend; /* 2: More terms for ARRAY and OBJECT */ - u32 iKey; /* 3: Key for ARRAY objects in json_tree() */ - u32 iPrev; /* 4: Previous SUBST node, or 0 */ - } u; -}; +#define JSON_JSON 0x01 /* Result is always JSON */ +#define JSON_SQL 0x02 /* Result is always SQL */ +#define JSON_ABPATH 0x03 /* Allow abbreviated JSON path specs */ +#define JSON_ISSET 0x04 /* json_set(), not json_insert() */ +#define JSON_BLOB 0x08 /* Use the BLOB output format */ -/* A parsed and possibly edited JSON string. Lifecycle: +/* A parsed JSON value. Lifecycle: ** -** 1. JSON comes in and is parsed into an array aNode[]. The original -** JSON text is stored in zJson. +** 1. JSON comes in and is parsed into a JSONB value in aBlob. The +** original text is stored in zJson. This step is skipped if the +** input is JSONB instead of text JSON. ** -** 2. Zero or more changes are made (via json_remove() or json_replace() -** or similar) to the aNode[] array. +** 2. The aBlob[] array is searched using the JSON path notation, if needed. ** -** 3. A new, edited and mimified JSON string is generated from aNode -** and stored in zAlt. The JsonParse object always owns zAlt. +** 3. Zero or more changes are made to aBlob[] (via json_remove() or +** json_replace() or json_patch() or similar). ** -** Step 1 always happens. Step 2 and 3 may or may not happen, depending -** on the operation. -** -** aNode[].u.zJContent entries typically point into zJson. Hence zJson -** must remain valid for the lifespan of the parse. For edits, -** aNode[].u.zJContent might point to malloced space other than zJson. -** Entries in pClup are responsible for freeing that extra malloced space. -** -** When walking the parse tree in aNode[], edits are ignored if useMod is -** false. +** 4. New JSON text is generated from the aBlob[] for output. This step +** is skipped if the function is one of the jsonb_* functions that +** returns JSONB instead of text JSON. */ struct JsonParse { - u32 nNode; /* Number of slots of aNode[] used */ - u32 nAlloc; /* Number of slots of aNode[] allocated */ - JsonNode *aNode; /* Array of nodes containing the parse */ - char *zJson; /* Original JSON string (before edits) */ - char *zAlt; /* Revised and/or mimified JSON */ - u32 *aUp; /* Index of parent of each node */ - JsonCleanup *pClup;/* Cleanup operations prior to freeing this object */ + u8 *aBlob; /* JSONB representation of JSON value */ + u32 nBlob; /* Bytes of aBlob[] actually used */ + u32 nBlobAlloc; /* Bytes allocated to aBlob[]. 0 if aBlob is external */ + char *zJson; /* Json text used for parsing */ + sqlite3 *db; /* The database connection to which this object belongs */ + int nJson; /* Length of the zJson string in bytes */ + u32 nJPRef; /* Number of references to this object */ + u32 iErr; /* Error location in zJson[] */ u16 iDepth; /* Nesting depth */ u8 nErr; /* Number of errors seen */ u8 oom; /* Set to true if out of memory */ u8 bJsonIsRCStr; /* True if zJson is an RCStr */ u8 hasNonstd; /* True if input uses non-standard features like JSON5 */ - u8 useMod; /* Actually use the edits contain inside aNode */ - u8 hasMod; /* aNode contains edits from the original zJson */ - u32 nJPRef; /* Number of references to this object */ - int nJson; /* Length of the zJson string in bytes */ - int nAlt; /* Length of alternative JSON string zAlt, in bytes */ - u32 iErr; /* Error location in zJson[] */ - u32 iSubst; /* Last JSON_SUBST entry in aNode[] */ - u32 iHold; /* Age of this entry in the cache for LRU replacement */ + u8 bReadOnly; /* Do not modify. */ + /* Search and edit information. See jsonLookupStep() */ + u8 eEdit; /* Edit operation to apply */ + int delta; /* Size change due to the edit */ + u32 nIns; /* Number of bytes to insert */ + u32 iLabel; /* Location of label if search landed on an object value */ + u8 *aIns; /* Content to be inserted */ }; +/* Allowed values for JsonParse.eEdit */ +#define JEDIT_DEL 1 /* Delete if exists */ +#define JEDIT_REPL 2 /* Overwrite if exists */ +#define JEDIT_INS 3 /* Insert if not exists */ +#define JEDIT_SET 4 /* Insert or overwrite */ + /* ** Maximum nesting depth of JSON for this implementation. ** @@ -202783,15 +203499,151 @@ struct JsonParse { ** descent parser. A depth of 1000 is far deeper than any sane JSON ** should go. Historical note: This limit was 2000 prior to version 3.42.0 */ -#define JSON_MAX_DEPTH 1000 +#ifndef SQLITE_JSON_MAX_DEPTH +# define JSON_MAX_DEPTH 1000 +#else +# define JSON_MAX_DEPTH SQLITE_JSON_MAX_DEPTH +#endif + +/* +** Allowed values for the flgs argument to jsonParseFuncArg(); +*/ +#define JSON_EDITABLE 0x01 /* Generate a writable JsonParse object */ +#define JSON_KEEPERROR 0x02 /* Return non-NULL even if there is an error */ + +/************************************************************************** +** Forward references +**************************************************************************/ +static void jsonReturnStringAsBlob(JsonString*); +static int jsonFuncArgMightBeBinary(sqlite3_value *pJson); +static u32 jsonTranslateBlobToText(const JsonParse*,u32,JsonString*); +static void jsonReturnParse(sqlite3_context*,JsonParse*); +static JsonParse *jsonParseFuncArg(sqlite3_context*,sqlite3_value*,u32); +static void jsonParseFree(JsonParse*); +static u32 jsonbPayloadSize(const JsonParse*, u32, u32*); +static u32 jsonUnescapeOneChar(const char*, u32, u32*); + +/************************************************************************** +** Utility routines for dealing with JsonCache objects +**************************************************************************/ + +/* +** Free a JsonCache object. +*/ +static void jsonCacheDelete(JsonCache *p){ + int i; + for(i=0; inUsed; i++){ + jsonParseFree(p->a[i]); + } + sqlite3DbFree(p->db, p); +} +static void jsonCacheDeleteGeneric(void *p){ + jsonCacheDelete((JsonCache*)p); +} + +/* +** Insert a new entry into the cache. If the cache is full, expel +** the least recently used entry. Return SQLITE_OK on success or a +** result code otherwise. +** +** Cache entries are stored in age order, oldest first. +*/ +static int jsonCacheInsert( + sqlite3_context *ctx, /* The SQL statement context holding the cache */ + JsonParse *pParse /* The parse object to be added to the cache */ +){ + JsonCache *p; + + assert( pParse->zJson!=0 ); + assert( pParse->bJsonIsRCStr ); + assert( pParse->delta==0 ); + p = sqlite3_get_auxdata(ctx, JSON_CACHE_ID); + if( p==0 ){ + sqlite3 *db = sqlite3_context_db_handle(ctx); + p = sqlite3DbMallocZero(db, sizeof(*p)); + if( p==0 ) return SQLITE_NOMEM; + p->db = db; + sqlite3_set_auxdata(ctx, JSON_CACHE_ID, p, jsonCacheDeleteGeneric); + p = sqlite3_get_auxdata(ctx, JSON_CACHE_ID); + if( p==0 ) return SQLITE_NOMEM; + } + if( p->nUsed >= JSON_CACHE_SIZE ){ + jsonParseFree(p->a[0]); + memmove(p->a, &p->a[1], (JSON_CACHE_SIZE-1)*sizeof(p->a[0])); + p->nUsed = JSON_CACHE_SIZE-1; + } + assert( pParse->nBlobAlloc>0 ); + pParse->eEdit = 0; + pParse->nJPRef++; + pParse->bReadOnly = 1; + p->a[p->nUsed] = pParse; + p->nUsed++; + return SQLITE_OK; +} + +/* +** Search for a cached translation the json text supplied by pArg. Return +** the JsonParse object if found. Return NULL if not found. +** +** When a match if found, the matching entry is moved to become the +** most-recently used entry if it isn't so already. +** +** The JsonParse object returned still belongs to the Cache and might +** be deleted at any moment. If the caller whants the JsonParse to +** linger, it needs to increment the nPJRef reference counter. +*/ +static JsonParse *jsonCacheSearch( + sqlite3_context *ctx, /* The SQL statement context holding the cache */ + sqlite3_value *pArg /* Function argument containing SQL text */ +){ + JsonCache *p; + int i; + const char *zJson; + int nJson; + + if( sqlite3_value_type(pArg)!=SQLITE_TEXT ){ + return 0; + } + zJson = (const char*)sqlite3_value_text(pArg); + if( zJson==0 ) return 0; + nJson = sqlite3_value_bytes(pArg); + + p = sqlite3_get_auxdata(ctx, JSON_CACHE_ID); + if( p==0 ){ + return 0; + } + for(i=0; inUsed; i++){ + if( p->a[i]->zJson==zJson ) break; + } + if( i>=p->nUsed ){ + for(i=0; inUsed; i++){ + if( p->a[i]->nJson!=nJson ) continue; + if( memcmp(p->a[i]->zJson, zJson, nJson)==0 ) break; + } + } + if( inUsed ){ + if( inUsed-1 ){ + /* Make the matching entry the most recently used entry */ + JsonParse *tmp = p->a[i]; + memmove(&p->a[i], &p->a[i+1], (p->nUsed-i-1)*sizeof(tmp)); + p->a[p->nUsed-1] = tmp; + i = p->nUsed - 1; + } + assert( p->a[i]->delta==0 ); + return p->a[i]; + }else{ + return 0; + } +} /************************************************************************** ** Utility routines for dealing with JsonString objects **************************************************************************/ -/* Set the JsonString object to an empty string +/* Turn uninitialized bulk memory into a valid JsonString object +** holding a zero-length string. */ -static void jsonZero(JsonString *p){ +static void jsonStringZero(JsonString *p){ p->zBuf = p->zSpace; p->nAlloc = sizeof(p->zSpace); p->nUsed = 0; @@ -202800,39 +203652,39 @@ static void jsonZero(JsonString *p){ /* Initialize the JsonString object */ -static void jsonInit(JsonString *p, sqlite3_context *pCtx){ +static void jsonStringInit(JsonString *p, sqlite3_context *pCtx){ p->pCtx = pCtx; - p->bErr = 0; - jsonZero(p); + p->eErr = 0; + jsonStringZero(p); } /* Free all allocated memory and reset the JsonString object back to its ** initial state. */ -static void jsonReset(JsonString *p){ +static void jsonStringReset(JsonString *p){ if( !p->bStatic ) sqlite3RCStrUnref(p->zBuf); - jsonZero(p); + jsonStringZero(p); } /* Report an out-of-memory (OOM) condition */ -static void jsonOom(JsonString *p){ - p->bErr = 1; - sqlite3_result_error_nomem(p->pCtx); - jsonReset(p); +static void jsonStringOom(JsonString *p){ + p->eErr |= JSTRING_OOM; + if( p->pCtx ) sqlite3_result_error_nomem(p->pCtx); + jsonStringReset(p); } /* Enlarge pJson->zBuf so that it can hold at least N more bytes. ** Return zero on success. Return non-zero on an OOM error */ -static int jsonGrow(JsonString *p, u32 N){ +static int jsonStringGrow(JsonString *p, u32 N){ u64 nTotal = NnAlloc ? p->nAlloc*2 : p->nAlloc+N+10; char *zNew; if( p->bStatic ){ - if( p->bErr ) return 1; + if( p->eErr ) return 1; zNew = sqlite3RCStrNew(nTotal); if( zNew==0 ){ - jsonOom(p); + jsonStringOom(p); return SQLITE_NOMEM; } memcpy(zNew, p->zBuf, (size_t)p->nUsed); @@ -202841,8 +203693,8 @@ static int jsonGrow(JsonString *p, u32 N){ }else{ p->zBuf = sqlite3RCStrResize(p->zBuf, nTotal); if( p->zBuf==0 ){ - p->bErr = 1; - jsonZero(p); + p->eErr |= JSTRING_OOM; + jsonStringZero(p); return SQLITE_NOMEM; } } @@ -202852,20 +203704,20 @@ static int jsonGrow(JsonString *p, u32 N){ /* Append N bytes from zIn onto the end of the JsonString string. */ -static SQLITE_NOINLINE void jsonAppendExpand( +static SQLITE_NOINLINE void jsonStringExpandAndAppend( JsonString *p, const char *zIn, u32 N ){ assert( N>0 ); - if( jsonGrow(p,N) ) return; + if( jsonStringGrow(p,N) ) return; memcpy(p->zBuf+p->nUsed, zIn, N); p->nUsed += N; } static void jsonAppendRaw(JsonString *p, const char *zIn, u32 N){ if( N==0 ) return; if( N+p->nUsed >= p->nAlloc ){ - jsonAppendExpand(p,zIn,N); + jsonStringExpandAndAppend(p,zIn,N); }else{ memcpy(p->zBuf+p->nUsed, zIn, N); p->nUsed += N; @@ -202874,7 +203726,7 @@ static void jsonAppendRaw(JsonString *p, const char *zIn, u32 N){ static void jsonAppendRawNZ(JsonString *p, const char *zIn, u32 N){ assert( N>0 ); if( N+p->nUsed >= p->nAlloc ){ - jsonAppendExpand(p,zIn,N); + jsonStringExpandAndAppend(p,zIn,N); }else{ memcpy(p->zBuf+p->nUsed, zIn, N); p->nUsed += N; @@ -202886,7 +203738,7 @@ static void jsonAppendRawNZ(JsonString *p, const char *zIn, u32 N){ */ static void jsonPrintf(int N, JsonString *p, const char *zFormat, ...){ va_list ap; - if( (p->nUsed + N >= p->nAlloc) && jsonGrow(p, N) ) return; + if( (p->nUsed + N >= p->nAlloc) && jsonStringGrow(p, N) ) return; va_start(ap, zFormat); sqlite3_vsnprintf(N, p->zBuf+p->nUsed, zFormat, ap); va_end(ap); @@ -202896,7 +203748,7 @@ static void jsonPrintf(int N, JsonString *p, const char *zFormat, ...){ /* Append a single character */ static SQLITE_NOINLINE void jsonAppendCharExpand(JsonString *p, char c){ - if( jsonGrow(p,1) ) return; + if( jsonStringGrow(p,1) ) return; p->zBuf[p->nUsed++] = c; } static void jsonAppendChar(JsonString *p, char c){ @@ -202907,24 +203759,17 @@ static void jsonAppendChar(JsonString *p, char c){ } } -/* Try to force the string to be a zero-terminated RCStr string. +/* Make sure there is a zero terminator on p->zBuf[] ** ** Return true on success. Return false if an OOM prevents this ** from happening. */ -static int jsonForceRCStr(JsonString *p){ +static int jsonStringTerminate(JsonString *p){ jsonAppendChar(p, 0); - if( p->bErr ) return 0; - p->nUsed--; - if( p->bStatic==0 ) return 1; - p->nAlloc = 0; - p->nUsed++; - jsonGrow(p, p->nUsed); p->nUsed--; - return p->bStatic==0; + return p->eErr==0; } - /* Append a comma separator to the output buffer, if the previous ** character is not '[' or '{'. */ @@ -202937,21 +203782,66 @@ static void jsonAppendSeparator(JsonString *p){ } /* Append the N-byte string in zIn to the end of the JsonString string -** under construction. Enclose the string in "..." and escape -** any double-quotes or backslash characters contained within the +** under construction. Enclose the string in double-quotes ("...") and +** escape any double-quotes or backslash characters contained within the ** string. +** +** This routine is a high-runner. There is a measurable performance +** increase associated with unwinding the jsonIsOk[] loop. */ static void jsonAppendString(JsonString *p, const char *zIn, u32 N){ - u32 i; - if( zIn==0 || ((N+p->nUsed+2 >= p->nAlloc) && jsonGrow(p,N+2)!=0) ) return; + u32 k; + u8 c; + const u8 *z = (const u8*)zIn; + if( z==0 ) return; + if( (N+p->nUsed+2 >= p->nAlloc) && jsonStringGrow(p,N+2)!=0 ) return; p->zBuf[p->nUsed++] = '"'; - for(i=0; izBuf[p->nUsed++] = c; - }else if( c=='"' || c=='\\' ){ + while( 1 /*exit-by-break*/ ){ + k = 0; + /* The following while() is the 4-way unwound equivalent of + ** + ** while( k=N ){ + while( k=N ){ + if( k>0 ){ + memcpy(&p->zBuf[p->nUsed], z, k); + p->nUsed += k; + } + break; + } + if( k>0 ){ + memcpy(&p->zBuf[p->nUsed], z, k); + p->nUsed += k; + z += k; + N -= k; + } + c = z[0]; + if( c=='"' || c=='\\' ){ json_simple_escape: - if( (p->nUsed+N+3-i > p->nAlloc) && jsonGrow(p,N+3-i)!=0 ) return; + if( (p->nUsed+N+3 > p->nAlloc) && jsonStringGrow(p,N+3)!=0 ) return; p->zBuf[p->nUsed++] = '\\'; p->zBuf[p->nUsed++] = c; }else if( c=='\'' ){ @@ -202972,7 +203862,7 @@ static void jsonAppendString(JsonString *p, const char *zIn, u32 N){ c = aSpecial[c]; goto json_simple_escape; } - if( (p->nUsed+N+7+i > p->nAlloc) && jsonGrow(p,N+7-i)!=0 ) return; + if( (p->nUsed+N+7 > p->nAlloc) && jsonStringGrow(p,N+7)!=0 ) return; p->zBuf[p->nUsed++] = '\\'; p->zBuf[p->nUsed++] = 'u'; p->zBuf[p->nUsed++] = '0'; @@ -202980,140 +203870,18 @@ static void jsonAppendString(JsonString *p, const char *zIn, u32 N){ p->zBuf[p->nUsed++] = "0123456789abcdef"[c>>4]; p->zBuf[p->nUsed++] = "0123456789abcdef"[c&0xf]; } + z++; + N--; } p->zBuf[p->nUsed++] = '"'; assert( p->nUsednAlloc ); } /* -** The zIn[0..N] string is a JSON5 string literal. Append to p a translation -** of the string literal that standard JSON and that omits all JSON5 -** features. -*/ -static void jsonAppendNormalizedString(JsonString *p, const char *zIn, u32 N){ - u32 i; - jsonAppendChar(p, '"'); - zIn++; - N -= 2; - while( N>0 ){ - for(i=0; i0 ){ - jsonAppendRawNZ(p, zIn, i); - zIn += i; - N -= i; - if( N==0 ) break; - } - assert( zIn[0]=='\\' ); - switch( (u8)zIn[1] ){ - case '\'': - jsonAppendChar(p, '\''); - break; - case 'v': - jsonAppendRawNZ(p, "\\u0009", 6); - break; - case 'x': - jsonAppendRawNZ(p, "\\u00", 4); - jsonAppendRawNZ(p, &zIn[2], 2); - zIn += 2; - N -= 2; - break; - case '0': - jsonAppendRawNZ(p, "\\u0000", 6); - break; - case '\r': - if( zIn[2]=='\n' ){ - zIn++; - N--; - } - break; - case '\n': - break; - case 0xe2: - assert( N>=4 ); - assert( 0x80==(u8)zIn[2] ); - assert( 0xa8==(u8)zIn[3] || 0xa9==(u8)zIn[3] ); - zIn += 2; - N -= 2; - break; - default: - jsonAppendRawNZ(p, zIn, 2); - break; - } - zIn += 2; - N -= 2; - } - jsonAppendChar(p, '"'); -} - -/* -** The zIn[0..N] string is a JSON5 integer literal. Append to p a translation -** of the string literal that standard JSON and that omits all JSON5 -** features. -*/ -static void jsonAppendNormalizedInt(JsonString *p, const char *zIn, u32 N){ - if( zIn[0]=='+' ){ - zIn++; - N--; - }else if( zIn[0]=='-' ){ - jsonAppendChar(p, '-'); - zIn++; - N--; - } - if( zIn[0]=='0' && (zIn[1]=='x' || zIn[1]=='X') ){ - sqlite3_int64 i = 0; - int rc = sqlite3DecOrHexToI64(zIn, &i); - if( rc<=1 ){ - jsonPrintf(100,p,"%lld",i); - }else{ - assert( rc==2 ); - jsonAppendRawNZ(p, "9.0e999", 7); - } - return; - } - assert( N>0 ); - jsonAppendRawNZ(p, zIn, N); -} - -/* -** The zIn[0..N] string is a JSON5 real literal. Append to p a translation -** of the string literal that standard JSON and that omits all JSON5 -** features. -*/ -static void jsonAppendNormalizedReal(JsonString *p, const char *zIn, u32 N){ - u32 i; - if( zIn[0]=='+' ){ - zIn++; - N--; - }else if( zIn[0]=='-' ){ - jsonAppendChar(p, '-'); - zIn++; - N--; - } - if( zIn[0]=='.' ){ - jsonAppendChar(p, '0'); - } - for(i=0; i0 ){ - jsonAppendRawNZ(p, zIn, N); - } -} - - - -/* -** Append a function parameter value to the JSON string under -** construction. +** Append an sqlite3_value (such as a function parameter) to the JSON +** string under construction in p. */ -static void jsonAppendValue( +static void jsonAppendSqlValue( JsonString *p, /* Append to this JSON string */ sqlite3_value *pValue /* Value to append */ ){ @@ -203143,290 +203911,127 @@ static void jsonAppendValue( break; } default: { - if( p->bErr==0 ){ + if( jsonFuncArgMightBeBinary(pValue) ){ + JsonParse px; + memset(&px, 0, sizeof(px)); + px.aBlob = (u8*)sqlite3_value_blob(pValue); + px.nBlob = sqlite3_value_bytes(pValue); + jsonTranslateBlobToText(&px, 0, p); + }else if( p->eErr==0 ){ sqlite3_result_error(p->pCtx, "JSON cannot hold BLOB values", -1); - p->bErr = 2; - jsonReset(p); + p->eErr = JSTRING_ERR; + jsonStringReset(p); } break; } } } - -/* Make the JSON in p the result of the SQL function. +/* Make the text in p (which is probably a generated JSON text string) +** the result of the SQL function. +** +** The JsonString is reset. ** -** The JSON string is reset. +** If pParse and ctx are both non-NULL, then the SQL string in p is +** loaded into the zJson field of the pParse object as a RCStr and the +** pParse is added to the cache. */ -static void jsonResult(JsonString *p){ - if( p->bErr==0 ){ - if( p->bStatic ){ +static void jsonReturnString( + JsonString *p, /* String to return */ + JsonParse *pParse, /* JSONB source or NULL */ + sqlite3_context *ctx /* Where to cache */ +){ + assert( (pParse!=0)==(ctx!=0) ); + assert( ctx==0 || ctx==p->pCtx ); + if( p->eErr==0 ){ + int flags = SQLITE_PTR_TO_INT(sqlite3_user_data(p->pCtx)); + if( flags & JSON_BLOB ){ + jsonReturnStringAsBlob(p); + }else if( p->bStatic ){ sqlite3_result_text64(p->pCtx, p->zBuf, p->nUsed, SQLITE_TRANSIENT, SQLITE_UTF8); - }else if( jsonForceRCStr(p) ){ - sqlite3RCStrRef(p->zBuf); - sqlite3_result_text64(p->pCtx, p->zBuf, p->nUsed, + }else if( jsonStringTerminate(p) ){ + if( pParse && pParse->bJsonIsRCStr==0 && pParse->nBlobAlloc>0 ){ + int rc; + pParse->zJson = sqlite3RCStrRef(p->zBuf); + pParse->nJson = p->nUsed; + pParse->bJsonIsRCStr = 1; + rc = jsonCacheInsert(ctx, pParse); + if( rc==SQLITE_NOMEM ){ + sqlite3_result_error_nomem(ctx); + jsonStringReset(p); + return; + } + } + sqlite3_result_text64(p->pCtx, sqlite3RCStrRef(p->zBuf), p->nUsed, sqlite3RCStrUnref, SQLITE_UTF8); + }else{ + sqlite3_result_error_nomem(p->pCtx); } - } - if( p->bErr==1 ){ + }else if( p->eErr & JSTRING_OOM ){ sqlite3_result_error_nomem(p->pCtx); + }else if( p->eErr & JSTRING_MALFORMED ){ + sqlite3_result_error(p->pCtx, "malformed JSON", -1); } - jsonReset(p); + jsonStringReset(p); } /************************************************************************** -** Utility routines for dealing with JsonNode and JsonParse objects +** Utility routines for dealing with JsonParse objects **************************************************************************/ -/* -** Return the number of consecutive JsonNode slots need to represent -** the parsed JSON at pNode. The minimum answer is 1. For ARRAY and -** OBJECT types, the number might be larger. -** -** Appended elements are not counted. The value returned is the number -** by which the JsonNode counter should increment in order to go to the -** next peer value. -*/ -static u32 jsonNodeSize(JsonNode *pNode){ - return pNode->eType>=JSON_ARRAY ? pNode->n+1 : 1; -} - /* ** Reclaim all memory allocated by a JsonParse object. But do not ** delete the JsonParse object itself. */ static void jsonParseReset(JsonParse *pParse){ - while( pParse->pClup ){ - JsonCleanup *pTask = pParse->pClup; - pParse->pClup = pTask->pJCNext; - pTask->xOp(pTask->pArg); - sqlite3_free(pTask); - } assert( pParse->nJPRef<=1 ); - if( pParse->aNode ){ - sqlite3_free(pParse->aNode); - pParse->aNode = 0; - } - pParse->nNode = 0; - pParse->nAlloc = 0; - if( pParse->aUp ){ - sqlite3_free(pParse->aUp); - pParse->aUp = 0; - } if( pParse->bJsonIsRCStr ){ sqlite3RCStrUnref(pParse->zJson); pParse->zJson = 0; + pParse->nJson = 0; pParse->bJsonIsRCStr = 0; } - if( pParse->zAlt ){ - sqlite3RCStrUnref(pParse->zAlt); - pParse->zAlt = 0; + if( pParse->nBlobAlloc ){ + sqlite3DbFree(pParse->db, pParse->aBlob); + pParse->aBlob = 0; + pParse->nBlob = 0; + pParse->nBlobAlloc = 0; } } /* -** Free a JsonParse object that was obtained from sqlite3_malloc(). -** -** Note that destroying JsonParse might call sqlite3RCStrUnref() to -** destroy the zJson value. The RCStr object might recursively invoke -** JsonParse to destroy this pParse object again. Take care to ensure -** that this recursive destructor sequence terminates harmlessly. +** Decrement the reference count on the JsonParse object. When the +** count reaches zero, free the object. */ static void jsonParseFree(JsonParse *pParse){ - if( pParse->nJPRef>1 ){ - pParse->nJPRef--; - }else{ - jsonParseReset(pParse); - sqlite3_free(pParse); - } -} - -/* -** Add a cleanup task to the JsonParse object. -** -** If an OOM occurs, the cleanup operation happens immediately -** and this function returns SQLITE_NOMEM. -*/ -static int jsonParseAddCleanup( - JsonParse *pParse, /* Add the cleanup task to this parser */ - void(*xOp)(void*), /* The cleanup task */ - void *pArg /* Argument to the cleanup */ -){ - JsonCleanup *pTask = sqlite3_malloc64( sizeof(*pTask) ); - if( pTask==0 ){ - pParse->oom = 1; - xOp(pArg); - return SQLITE_ERROR; - } - pTask->pJCNext = pParse->pClup; - pParse->pClup = pTask; - pTask->xOp = xOp; - pTask->pArg = pArg; - return SQLITE_OK; -} - -/* -** Convert the JsonNode pNode into a pure JSON string and -** append to pOut. Subsubstructure is also included. Return -** the number of JsonNode objects that are encoded. -*/ -static void jsonRenderNode( - JsonParse *pParse, /* the complete parse of the JSON */ - JsonNode *pNode, /* The node to render */ - JsonString *pOut /* Write JSON here */ -){ - assert( pNode!=0 ); - while( (pNode->jnFlags & JNODE_REPLACE)!=0 && pParse->useMod ){ - u32 idx = (u32)(pNode - pParse->aNode); - u32 i = pParse->iSubst; - while( 1 /*exit-by-break*/ ){ - assert( inNode ); - assert( pParse->aNode[i].eType==JSON_SUBST ); - assert( pParse->aNode[i].eU==4 ); - assert( pParse->aNode[i].u.iPrevaNode[i].n==idx ){ - pNode = &pParse->aNode[i+1]; - break; - } - i = pParse->aNode[i].u.iPrev; - } - } - switch( pNode->eType ){ - default: { - assert( pNode->eType==JSON_NULL ); - jsonAppendRawNZ(pOut, "null", 4); - break; - } - case JSON_TRUE: { - jsonAppendRawNZ(pOut, "true", 4); - break; - } - case JSON_FALSE: { - jsonAppendRawNZ(pOut, "false", 5); - break; - } - case JSON_STRING: { - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_RAW ){ - if( pNode->jnFlags & JNODE_LABEL ){ - jsonAppendChar(pOut, '"'); - jsonAppendRaw(pOut, pNode->u.zJContent, pNode->n); - jsonAppendChar(pOut, '"'); - }else{ - jsonAppendString(pOut, pNode->u.zJContent, pNode->n); - } - }else if( pNode->jnFlags & JNODE_JSON5 ){ - jsonAppendNormalizedString(pOut, pNode->u.zJContent, pNode->n); - }else{ - assert( pNode->n>0 ); - jsonAppendRawNZ(pOut, pNode->u.zJContent, pNode->n); - } - break; - } - case JSON_REAL: { - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_JSON5 ){ - jsonAppendNormalizedReal(pOut, pNode->u.zJContent, pNode->n); - }else{ - assert( pNode->n>0 ); - jsonAppendRawNZ(pOut, pNode->u.zJContent, pNode->n); - } - break; - } - case JSON_INT: { - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_JSON5 ){ - jsonAppendNormalizedInt(pOut, pNode->u.zJContent, pNode->n); - }else{ - assert( pNode->n>0 ); - jsonAppendRawNZ(pOut, pNode->u.zJContent, pNode->n); - } - break; - } - case JSON_ARRAY: { - u32 j = 1; - jsonAppendChar(pOut, '['); - for(;;){ - while( j<=pNode->n ){ - if( (pNode[j].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ){ - jsonAppendSeparator(pOut); - jsonRenderNode(pParse, &pNode[j], pOut); - } - j += jsonNodeSize(&pNode[j]); - } - if( (pNode->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pNode->eU==2 ); - pNode = &pParse->aNode[pNode->u.iAppend]; - j = 1; - } - jsonAppendChar(pOut, ']'); - break; - } - case JSON_OBJECT: { - u32 j = 1; - jsonAppendChar(pOut, '{'); - for(;;){ - while( j<=pNode->n ){ - if( (pNode[j+1].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ){ - jsonAppendSeparator(pOut); - jsonRenderNode(pParse, &pNode[j], pOut); - jsonAppendChar(pOut, ':'); - jsonRenderNode(pParse, &pNode[j+1], pOut); - } - j += 1 + jsonNodeSize(&pNode[j+1]); - } - if( (pNode->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pNode->eU==2 ); - pNode = &pParse->aNode[pNode->u.iAppend]; - j = 1; - } - jsonAppendChar(pOut, '}'); - break; + if( pParse ){ + if( pParse->nJPRef>1 ){ + pParse->nJPRef--; + }else{ + jsonParseReset(pParse); + sqlite3DbFree(pParse->db, pParse); } } } -/* -** Return a JsonNode and all its descendants as a JSON string. -*/ -static void jsonReturnJson( - JsonParse *pParse, /* The complete JSON */ - JsonNode *pNode, /* Node to return */ - sqlite3_context *pCtx, /* Return value for this function */ - int bGenerateAlt /* Also store the rendered text in zAlt */ -){ - JsonString s; - if( pParse->oom ){ - sqlite3_result_error_nomem(pCtx); - return; - } - if( pParse->nErr==0 ){ - jsonInit(&s, pCtx); - jsonRenderNode(pParse, pNode, &s); - if( bGenerateAlt && pParse->zAlt==0 && jsonForceRCStr(&s) ){ - pParse->zAlt = sqlite3RCStrRef(s.zBuf); - pParse->nAlt = s.nUsed; - } - jsonResult(&s); - sqlite3_result_subtype(pCtx, JSON_SUBTYPE); - } -} +/************************************************************************** +** Utility routines for the JSON text parser +**************************************************************************/ /* ** Translate a single byte of Hex into an integer. -** This routine only works if h really is a valid hexadecimal -** character: 0..9a..fA..F +** This routine only gives a correct answer if h really is a valid hexadecimal +** character: 0..9a..fA..F. But unlike sqlite3HexToInt(), it does not +** assert() if the digit is not hex. */ static u8 jsonHexToInt(int h){ - assert( (h>='0' && h<='9') || (h>='a' && h<='f') || (h>='A' && h<='F') ); +#ifdef SQLITE_ASCII + h += 9*(1&(h>>6)); +#endif #ifdef SQLITE_EBCDIC h += 9*(1&~(h>>4)); -#else - h += 9*(1&(h>>6)); #endif return (u8)(h & 0xf); } @@ -203436,10 +204041,6 @@ static u8 jsonHexToInt(int h){ */ static u32 jsonHexToInt4(const char *z){ u32 v; - assert( sqlite3Isxdigit(z[0]) ); - assert( sqlite3Isxdigit(z[1]) ); - assert( sqlite3Isxdigit(z[2]) ); - assert( sqlite3Isxdigit(z[3]) ); v = (jsonHexToInt(z[0])<<12) + (jsonHexToInt(z[1])<<8) + (jsonHexToInt(z[2])<<4) @@ -203447,281 +204048,6 @@ static u32 jsonHexToInt4(const char *z){ return v; } -/* -** Make the JsonNode the return value of the function. -*/ -static void jsonReturn( - JsonParse *pParse, /* Complete JSON parse tree */ - JsonNode *pNode, /* Node to return */ - sqlite3_context *pCtx /* Return value for this function */ -){ - switch( pNode->eType ){ - default: { - assert( pNode->eType==JSON_NULL ); - sqlite3_result_null(pCtx); - break; - } - case JSON_TRUE: { - sqlite3_result_int(pCtx, 1); - break; - } - case JSON_FALSE: { - sqlite3_result_int(pCtx, 0); - break; - } - case JSON_INT: { - sqlite3_int64 i = 0; - int rc; - int bNeg = 0; - const char *z; - - assert( pNode->eU==1 ); - z = pNode->u.zJContent; - if( z[0]=='-' ){ z++; bNeg = 1; } - else if( z[0]=='+' ){ z++; } - rc = sqlite3DecOrHexToI64(z, &i); - if( rc<=1 ){ - sqlite3_result_int64(pCtx, bNeg ? -i : i); - }else if( rc==3 && bNeg ){ - sqlite3_result_int64(pCtx, SMALLEST_INT64); - }else{ - goto to_double; - } - break; - } - case JSON_REAL: { - double r; - const char *z; - assert( pNode->eU==1 ); - to_double: - z = pNode->u.zJContent; - sqlite3AtoF(z, &r, sqlite3Strlen30(z), SQLITE_UTF8); - sqlite3_result_double(pCtx, r); - break; - } - case JSON_STRING: { - if( pNode->jnFlags & JNODE_RAW ){ - assert( pNode->eU==1 ); - sqlite3_result_text(pCtx, pNode->u.zJContent, pNode->n, - SQLITE_TRANSIENT); - }else if( (pNode->jnFlags & JNODE_ESCAPE)==0 ){ - /* JSON formatted without any backslash-escapes */ - assert( pNode->eU==1 ); - sqlite3_result_text(pCtx, pNode->u.zJContent+1, pNode->n-2, - SQLITE_TRANSIENT); - }else{ - /* Translate JSON formatted string into raw text */ - u32 i; - u32 n = pNode->n; - const char *z; - char *zOut; - u32 j; - u32 nOut = n; - assert( pNode->eU==1 ); - z = pNode->u.zJContent; - zOut = sqlite3_malloc( nOut+1 ); - if( zOut==0 ){ - sqlite3_result_error_nomem(pCtx); - break; - } - for(i=1, j=0; i>6)); - zOut[j++] = 0x80 | (v&0x3f); - }else{ - u32 vlo; - if( (v&0xfc00)==0xd800 - && i>18); - zOut[j++] = 0x80 | ((v>>12)&0x3f); - zOut[j++] = 0x80 | ((v>>6)&0x3f); - zOut[j++] = 0x80 | (v&0x3f); - }else{ - zOut[j++] = 0xe0 | (v>>12); - zOut[j++] = 0x80 | ((v>>6)&0x3f); - zOut[j++] = 0x80 | (v&0x3f); - } - } - continue; - }else if( c=='b' ){ - c = '\b'; - }else if( c=='f' ){ - c = '\f'; - }else if( c=='n' ){ - c = '\n'; - }else if( c=='r' ){ - c = '\r'; - }else if( c=='t' ){ - c = '\t'; - }else if( c=='v' ){ - c = '\v'; - }else if( c=='\'' || c=='"' || c=='/' || c=='\\' ){ - /* pass through unchanged */ - }else if( c=='0' ){ - c = 0; - }else if( c=='x' ){ - c = (jsonHexToInt(z[i+1])<<4) | jsonHexToInt(z[i+2]); - i += 2; - }else if( c=='\r' && z[i+1]=='\n' ){ - i++; - continue; - }else if( 0xe2==(u8)c ){ - assert( 0x80==(u8)z[i+1] ); - assert( 0xa8==(u8)z[i+2] || 0xa9==(u8)z[i+2] ); - i += 2; - continue; - }else{ - continue; - } - } /* end if( c=='\\' ) */ - zOut[j++] = c; - } /* end for() */ - zOut[j] = 0; - sqlite3_result_text(pCtx, zOut, j, sqlite3_free); - } - break; - } - case JSON_ARRAY: - case JSON_OBJECT: { - jsonReturnJson(pParse, pNode, pCtx, 0); - break; - } - } -} - -/* Forward reference */ -static int jsonParseAddNode(JsonParse*,u32,u32,const char*); - -/* -** A macro to hint to the compiler that a function should not be -** inlined. -*/ -#if defined(__GNUC__) -# define JSON_NOINLINE __attribute__((noinline)) -#elif defined(_MSC_VER) && _MSC_VER>=1310 -# define JSON_NOINLINE __declspec(noinline) -#else -# define JSON_NOINLINE -#endif - - -/* -** Add a single node to pParse->aNode after first expanding the -** size of the aNode array. Return the index of the new node. -** -** If an OOM error occurs, set pParse->oom and return -1. -*/ -static JSON_NOINLINE int jsonParseAddNodeExpand( - JsonParse *pParse, /* Append the node to this object */ - u32 eType, /* Node type */ - u32 n, /* Content size or sub-node count */ - const char *zContent /* Content */ -){ - u32 nNew; - JsonNode *pNew; - assert( pParse->nNode>=pParse->nAlloc ); - if( pParse->oom ) return -1; - nNew = pParse->nAlloc*2 + 10; - pNew = sqlite3_realloc64(pParse->aNode, sizeof(JsonNode)*nNew); - if( pNew==0 ){ - pParse->oom = 1; - return -1; - } - pParse->nAlloc = sqlite3_msize(pNew)/sizeof(JsonNode); - pParse->aNode = pNew; - assert( pParse->nNodenAlloc ); - return jsonParseAddNode(pParse, eType, n, zContent); -} - -/* -** Create a new JsonNode instance based on the arguments and append that -** instance to the JsonParse. Return the index in pParse->aNode[] of the -** new node, or -1 if a memory allocation fails. -*/ -static int jsonParseAddNode( - JsonParse *pParse, /* Append the node to this object */ - u32 eType, /* Node type */ - u32 n, /* Content size or sub-node count */ - const char *zContent /* Content */ -){ - JsonNode *p; - assert( pParse->aNode!=0 || pParse->nNode>=pParse->nAlloc ); - if( pParse->nNode>=pParse->nAlloc ){ - return jsonParseAddNodeExpand(pParse, eType, n, zContent); - } - assert( pParse->aNode!=0 ); - p = &pParse->aNode[pParse->nNode]; - assert( p!=0 ); - p->eType = (u8)(eType & 0xff); - p->jnFlags = (u8)(eType >> 8); - VVA( p->eU = zContent ? 1 : 0 ); - p->n = n; - p->u.zJContent = zContent; - return pParse->nNode++; -} - -/* -** Add an array of new nodes to the current pParse->aNode array. -** Return the index of the first node added. -** -** If an OOM error occurs, set pParse->oom. -*/ -static void jsonParseAddNodeArray( - JsonParse *pParse, /* Append the node to this object */ - JsonNode *aNode, /* Array of nodes to add */ - u32 nNode /* Number of elements in aNew */ -){ - assert( aNode!=0 ); - assert( nNode>=1 ); - if( pParse->nNode + nNode > pParse->nAlloc ){ - u32 nNew = pParse->nNode + nNode; - JsonNode *aNew = sqlite3_realloc64(pParse->aNode, nNew*sizeof(JsonNode)); - if( aNew==0 ){ - pParse->oom = 1; - return; - } - pParse->nAlloc = sqlite3_msize(aNew)/sizeof(JsonNode); - pParse->aNode = aNew; - } - memcpy(&pParse->aNode[pParse->nNode], aNode, nNode*sizeof(JsonNode)); - pParse->nNode += nNode; -} - -/* -** Add a new JSON_SUBST node. The node immediately following -** this new node will be the substitute content for iNode. -*/ -static int jsonParseAddSubstNode( - JsonParse *pParse, /* Add the JSON_SUBST here */ - u32 iNode /* References this node */ -){ - int idx = jsonParseAddNode(pParse, JSON_SUBST, iNode, 0); - if( pParse->oom ) return -1; - pParse->aNode[iNode].jnFlags |= JNODE_REPLACE; - pParse->aNode[idx].eU = 4; - pParse->aNode[idx].u.iPrev = pParse->iSubst; - pParse->iSubst = idx; - pParse->hasMod = 1; - pParse->useMod = 1; - return idx; -} - /* ** Return true if z[] begins with 2 (or more) hexadecimal digits */ @@ -203875,63 +204201,500 @@ static const struct NanInfName { char *zMatch; char *zRepl; } aNanInfName[] = { - { 'i', 'I', 3, JSON_REAL, 7, "inf", "9.0e999" }, - { 'i', 'I', 8, JSON_REAL, 7, "infinity", "9.0e999" }, - { 'n', 'N', 3, JSON_NULL, 4, "NaN", "null" }, - { 'q', 'Q', 4, JSON_NULL, 4, "QNaN", "null" }, - { 's', 'S', 4, JSON_NULL, 4, "SNaN", "null" }, + { 'i', 'I', 3, JSONB_FLOAT, 7, "inf", "9.0e999" }, + { 'i', 'I', 8, JSONB_FLOAT, 7, "infinity", "9.0e999" }, + { 'n', 'N', 3, JSONB_NULL, 4, "NaN", "null" }, + { 'q', 'Q', 4, JSONB_NULL, 4, "QNaN", "null" }, + { 's', 'S', 4, JSONB_NULL, 4, "SNaN", "null" }, }; + /* -** Parse a single JSON value which begins at pParse->zJson[i]. Return the -** index of the first character past the end of the value parsed. +** Report the wrong number of arguments for json_insert(), json_replace() +** or json_set(). +*/ +static void jsonWrongNumArgs( + sqlite3_context *pCtx, + const char *zFuncName +){ + char *zMsg = sqlite3_mprintf("json_%s() needs an odd number of arguments", + zFuncName); + sqlite3_result_error(pCtx, zMsg, -1); + sqlite3_free(zMsg); +} + +/**************************************************************************** +** Utility routines for dealing with the binary BLOB representation of JSON +****************************************************************************/ + +/* +** Expand pParse->aBlob so that it holds at least N bytes. ** -** Special return values: +** Return the number of errors. +*/ +static int jsonBlobExpand(JsonParse *pParse, u32 N){ + u8 *aNew; + u32 t; + assert( N>pParse->nBlobAlloc ); + if( pParse->nBlobAlloc==0 ){ + t = 100; + }else{ + t = pParse->nBlobAlloc*2; + } + if( tdb, pParse->aBlob, t); + if( aNew==0 ){ pParse->oom = 1; return 1; } + pParse->aBlob = aNew; + pParse->nBlobAlloc = t; + return 0; +} + +/* +** If pParse->aBlob is not previously editable (because it is taken +** from sqlite3_value_blob(), as indicated by the fact that +** pParse->nBlobAlloc==0 and pParse->nBlob>0) then make it editable +** by making a copy into space obtained from malloc. +** +** Return true on success. Return false on OOM. +*/ +static int jsonBlobMakeEditable(JsonParse *pParse, u32 nExtra){ + u8 *aOld; + u32 nSize; + assert( !pParse->bReadOnly ); + if( pParse->oom ) return 0; + if( pParse->nBlobAlloc>0 ) return 1; + aOld = pParse->aBlob; + nSize = pParse->nBlob + nExtra; + pParse->aBlob = 0; + if( jsonBlobExpand(pParse, nSize) ){ + return 0; + } + assert( pParse->nBlobAlloc >= pParse->nBlob + nExtra ); + memcpy(pParse->aBlob, aOld, pParse->nBlob); + return 1; +} + +/* Expand pParse->aBlob and append one bytes. +*/ +static SQLITE_NOINLINE void jsonBlobExpandAndAppendOneByte( + JsonParse *pParse, + u8 c +){ + jsonBlobExpand(pParse, pParse->nBlob+1); + if( pParse->oom==0 ){ + assert( pParse->nBlob+1<=pParse->nBlobAlloc ); + pParse->aBlob[pParse->nBlob++] = c; + } +} + +/* Append a single character. +*/ +static void jsonBlobAppendOneByte(JsonParse *pParse, u8 c){ + if( pParse->nBlob >= pParse->nBlobAlloc ){ + jsonBlobExpandAndAppendOneByte(pParse, c); + }else{ + pParse->aBlob[pParse->nBlob++] = c; + } +} + +/* Slow version of jsonBlobAppendNode() that first resizes the +** pParse->aBlob structure. +*/ +static void jsonBlobAppendNode(JsonParse*,u8,u32,const void*); +static SQLITE_NOINLINE void jsonBlobExpandAndAppendNode( + JsonParse *pParse, + u8 eType, + u32 szPayload, + const void *aPayload +){ + if( jsonBlobExpand(pParse, pParse->nBlob+szPayload+9) ) return; + jsonBlobAppendNode(pParse, eType, szPayload, aPayload); +} + + +/* Append an node type byte together with the payload size and +** possibly also the payload. +** +** If aPayload is not NULL, then it is a pointer to the payload which +** is also appended. If aPayload is NULL, the pParse->aBlob[] array +** is resized (if necessary) so that it is big enough to hold the +** payload, but the payload is not appended and pParse->nBlob is left +** pointing to where the first byte of payload will eventually be. +*/ +static void jsonBlobAppendNode( + JsonParse *pParse, /* The JsonParse object under construction */ + u8 eType, /* Node type. One of JSONB_* */ + u32 szPayload, /* Number of bytes of payload */ + const void *aPayload /* The payload. Might be NULL */ +){ + u8 *a; + if( pParse->nBlob+szPayload+9 > pParse->nBlobAlloc ){ + jsonBlobExpandAndAppendNode(pParse,eType,szPayload,aPayload); + return; + } + assert( pParse->aBlob!=0 ); + a = &pParse->aBlob[pParse->nBlob]; + if( szPayload<=11 ){ + a[0] = eType | (szPayload<<4); + pParse->nBlob += 1; + }else if( szPayload<=0xff ){ + a[0] = eType | 0xc0; + a[1] = szPayload & 0xff; + pParse->nBlob += 2; + }else if( szPayload<=0xffff ){ + a[0] = eType | 0xd0; + a[1] = (szPayload >> 8) & 0xff; + a[2] = szPayload & 0xff; + pParse->nBlob += 3; + }else{ + a[0] = eType | 0xe0; + a[1] = (szPayload >> 24) & 0xff; + a[2] = (szPayload >> 16) & 0xff; + a[3] = (szPayload >> 8) & 0xff; + a[4] = szPayload & 0xff; + pParse->nBlob += 5; + } + if( aPayload ){ + pParse->nBlob += szPayload; + memcpy(&pParse->aBlob[pParse->nBlob-szPayload], aPayload, szPayload); + } +} + +/* Change the payload size for the node at index i to be szPayload. +*/ +static int jsonBlobChangePayloadSize( + JsonParse *pParse, + u32 i, + u32 szPayload +){ + u8 *a; + u8 szType; + u8 nExtra; + u8 nNeeded; + int delta; + if( pParse->oom ) return 0; + a = &pParse->aBlob[i]; + szType = a[0]>>4; + if( szType<=11 ){ + nExtra = 0; + }else if( szType==12 ){ + nExtra = 1; + }else if( szType==13 ){ + nExtra = 2; + }else{ + nExtra = 4; + } + if( szPayload<=11 ){ + nNeeded = 0; + }else if( szPayload<=0xff ){ + nNeeded = 1; + }else if( szPayload<=0xffff ){ + nNeeded = 2; + }else{ + nNeeded = 4; + } + delta = nNeeded - nExtra; + if( delta ){ + u32 newSize = pParse->nBlob + delta; + if( delta>0 ){ + if( newSize>pParse->nBlobAlloc && jsonBlobExpand(pParse, newSize) ){ + return 0; /* OOM error. Error state recorded in pParse->oom. */ + } + a = &pParse->aBlob[i]; + memmove(&a[1+delta], &a[1], pParse->nBlob - (i+1)); + }else{ + memmove(&a[1], &a[1-delta], pParse->nBlob - (i+1-delta)); + } + pParse->nBlob = newSize; + } + if( nNeeded==0 ){ + a[0] = (a[0] & 0x0f) | (szPayload<<4); + }else if( nNeeded==1 ){ + a[0] = (a[0] & 0x0f) | 0xc0; + a[1] = szPayload & 0xff; + }else if( nNeeded==2 ){ + a[0] = (a[0] & 0x0f) | 0xd0; + a[1] = (szPayload >> 8) & 0xff; + a[2] = szPayload & 0xff; + }else{ + a[0] = (a[0] & 0x0f) | 0xe0; + a[1] = (szPayload >> 24) & 0xff; + a[2] = (szPayload >> 16) & 0xff; + a[3] = (szPayload >> 8) & 0xff; + a[4] = szPayload & 0xff; + } + return delta; +} + +/* +** If z[0] is 'u' and is followed by exactly 4 hexadecimal character, +** then set *pOp to JSONB_TEXTJ and return true. If not, do not make +** any changes to *pOp and return false. +*/ +static int jsonIs4HexB(const char *z, int *pOp){ + if( z[0]!='u' ) return 0; + if( !jsonIs4Hex(&z[1]) ) return 0; + *pOp = JSONB_TEXTJ; + return 1; +} + +/* +** Check a single element of the JSONB in pParse for validity. +** +** The element to be checked starts at offset i and must end at on the +** last byte before iEnd. +** +** Return 0 if everything is correct. Return the 1-based byte offset of the +** error if a problem is detected. (In other words, if the error is at offset +** 0, return 1). +*/ +static u32 jsonbValidityCheck( + const JsonParse *pParse, /* Input JSONB. Only aBlob and nBlob are used */ + u32 i, /* Start of element as pParse->aBlob[i] */ + u32 iEnd, /* One more than the last byte of the element */ + u32 iDepth /* Current nesting depth */ +){ + u32 n, sz, j, k; + const u8 *z; + u8 x; + if( iDepth>JSON_MAX_DEPTH ) return i+1; + sz = 0; + n = jsonbPayloadSize(pParse, i, &sz); + if( NEVER(n==0) ) return i+1; /* Checked by caller */ + if( NEVER(i+n+sz!=iEnd) ) return i+1; /* Checked by caller */ + z = pParse->aBlob; + x = z[i] & 0x0f; + switch( x ){ + case JSONB_NULL: + case JSONB_TRUE: + case JSONB_FALSE: { + return n+sz==1 ? 0 : i+1; + } + case JSONB_INT: { + if( sz<1 ) return i+1; + j = i+n; + if( z[j]=='-' ){ + j++; + if( sz<2 ) return i+1; + } + k = i+n+sz; + while( jk ) return j+1; + if( z[j+1]!='.' && z[j+1]!='e' && z[j+1]!='E' ) return j+1; + j++; + } + for(; j0 ) return j+1; + if( x==JSONB_FLOAT && (j==k-1 || !sqlite3Isdigit(z[j+1])) ){ + return j+1; + } + seen = 1; + continue; + } + if( z[j]=='e' || z[j]=='E' ){ + if( seen==2 ) return j+1; + if( j==k-1 ) return j+1; + if( z[j+1]=='+' || z[j+1]=='-' ){ + j++; + if( j==k-1 ) return j+1; + } + seen = 2; + continue; + } + return j+1; + } + if( seen==0 ) return i+1; + return 0; + } + case JSONB_TEXT: { + j = i+n; + k = j+sz; + while( j=k ){ + return j+1; + }else if( strchr("\"\\/bfnrt",z[j+1])!=0 ){ + j++; + }else if( z[j+1]=='u' ){ + if( j+5>=k ) return j+1; + if( !jsonIs4Hex((const char*)&z[j+2]) ) return j+1; + j++; + }else if( x!=JSONB_TEXT5 ){ + return j+1; + }else{ + u32 c = 0; + u32 szC = jsonUnescapeOneChar((const char*)&z[j], k-j, &c); + if( c==JSON_INVALID_CHAR ) return j+1; + j += szC - 1; + } + } + j++; + } + return 0; + } + case JSONB_TEXTRAW: { + return 0; + } + case JSONB_ARRAY: { + u32 sub; + j = i+n; + k = j+sz; + while( jk ) return j+1; + sub = jsonbValidityCheck(pParse, j, j+n+sz, iDepth+1); + if( sub ) return sub; + j += n + sz; + } + assert( j==k ); + return 0; + } + case JSONB_OBJECT: { + u32 cnt = 0; + u32 sub; + j = i+n; + k = j+sz; + while( jk ) return j+1; + if( (cnt & 1)==0 ){ + x = z[j] & 0x0f; + if( xJSONB_TEXTRAW ) return j+1; + } + sub = jsonbValidityCheck(pParse, j, j+n+sz, iDepth+1); + if( sub ) return sub; + cnt++; + j += n + sz; + } + assert( j==k ); + if( (cnt & 1)!=0 ) return j+1; + return 0; + } + default: { + return i+1; + } + } +} + +/* +** Translate a single element of JSON text at pParse->zJson[i] into +** its equivalent binary JSONB representation. Append the translation into +** pParse->aBlob[] beginning at pParse->nBlob. The size of +** pParse->aBlob[] is increased as necessary. +** +** Return the index of the first character past the end of the element parsed, +** or one of the following special result codes: ** ** 0 End of input -** -1 Syntax error -** -2 '}' seen -** -3 ']' seen -** -4 ',' seen -** -5 ':' seen +** -1 Syntax error or OOM +** -2 '}' seen \ +** -3 ']' seen \___ For these returns, pParse->iErr is set to +** -4 ',' seen / the index in zJson[] of the seen character +** -5 ':' seen / */ -static int jsonParseValue(JsonParse *pParse, u32 i){ +static int jsonTranslateTextToBlob(JsonParse *pParse, u32 i){ char c; u32 j; - int iThis; + u32 iThis, iStart; int x; - JsonNode *pNode; + u8 t; const char *z = pParse->zJson; json_parse_restart: switch( (u8)z[i] ){ case '{': { /* Parse object */ - iThis = jsonParseAddNode(pParse, JSON_OBJECT, 0, 0); - if( iThis<0 ) return -1; + iThis = pParse->nBlob; + jsonBlobAppendNode(pParse, JSONB_OBJECT, pParse->nJson-i, 0); if( ++pParse->iDepth > JSON_MAX_DEPTH ){ pParse->iErr = i; return -1; } + iStart = pParse->nBlob; for(j=i+1;;j++){ - u32 nNode = pParse->nNode; - x = jsonParseValue(pParse, j); + u32 iBlob = pParse->nBlob; + x = jsonTranslateTextToBlob(pParse, j); if( x<=0 ){ + int op; if( x==(-2) ){ j = pParse->iErr; - if( pParse->nNode!=(u32)iThis+1 ) pParse->hasNonstd = 1; + if( pParse->nBlob!=(u32)iStart ) pParse->hasNonstd = 1; break; } j += json5Whitespace(&z[j]); + op = JSONB_TEXT; if( sqlite3JsonId1(z[j]) - || (z[j]=='\\' && z[j+1]=='u' && jsonIs4Hex(&z[j+2])) + || (z[j]=='\\' && jsonIs4HexB(&z[j+1], &op)) ){ int k = j+1; while( (sqlite3JsonId2(z[k]) && json5Whitespace(&z[k])==0) - || (z[k]=='\\' && z[k+1]=='u' && jsonIs4Hex(&z[k+2])) + || (z[k]=='\\' && jsonIs4HexB(&z[k+1], &op)) ){ k++; } - jsonParseAddNode(pParse, JSON_STRING | (JNODE_RAW<<8), k-j, &z[j]); + assert( iBlob==pParse->nBlob ); + jsonBlobAppendNode(pParse, op, k-j, &z[j]); pParse->hasNonstd = 1; x = k; }else{ @@ -203940,24 +204703,24 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ } } if( pParse->oom ) return -1; - pNode = &pParse->aNode[nNode]; - if( pNode->eType!=JSON_STRING ){ + t = pParse->aBlob[iBlob] & 0x0f; + if( tJSONB_TEXTRAW ){ pParse->iErr = j; return -1; } - pNode->jnFlags |= JNODE_LABEL; j = x; if( z[j]==':' ){ j++; }else{ - if( fast_isspace(z[j]) ){ - do{ j++; }while( fast_isspace(z[j]) ); + if( jsonIsspace(z[j]) ){ + /* strspn() is not helpful here */ + do{ j++; }while( jsonIsspace(z[j]) ); if( z[j]==':' ){ j++; goto parse_object_value; } } - x = jsonParseValue(pParse, j); + x = jsonTranslateTextToBlob(pParse, j); if( x!=(-5) ){ if( x!=(-1) ) pParse->iErr = j; return -1; @@ -203965,7 +204728,7 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ j = pParse->iErr+1; } parse_object_value: - x = jsonParseValue(pParse, j); + x = jsonTranslateTextToBlob(pParse, j); if( x<=0 ){ if( x!=(-1) ) pParse->iErr = j; return -1; @@ -203976,15 +204739,15 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ }else if( z[j]=='}' ){ break; }else{ - if( fast_isspace(z[j]) ){ - do{ j++; }while( fast_isspace(z[j]) ); + if( jsonIsspace(z[j]) ){ + j += 1 + (u32)strspn(&z[j+1], jsonSpaces); if( z[j]==',' ){ continue; }else if( z[j]=='}' ){ break; } } - x = jsonParseValue(pParse, j); + x = jsonTranslateTextToBlob(pParse, j); if( x==(-4) ){ j = pParse->iErr; continue; @@ -203997,25 +204760,26 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ pParse->iErr = j; return -1; } - pParse->aNode[iThis].n = pParse->nNode - (u32)iThis - 1; + jsonBlobChangePayloadSize(pParse, iThis, pParse->nBlob - iStart); pParse->iDepth--; return j+1; } case '[': { /* Parse array */ - iThis = jsonParseAddNode(pParse, JSON_ARRAY, 0, 0); - if( iThis<0 ) return -1; + iThis = pParse->nBlob; + jsonBlobAppendNode(pParse, JSONB_ARRAY, pParse->nJson - i, 0); + iStart = pParse->nBlob; + if( pParse->oom ) return -1; if( ++pParse->iDepth > JSON_MAX_DEPTH ){ pParse->iErr = i; return -1; } - memset(&pParse->aNode[iThis].u, 0, sizeof(pParse->aNode[iThis].u)); for(j=i+1;;j++){ - x = jsonParseValue(pParse, j); + x = jsonTranslateTextToBlob(pParse, j); if( x<=0 ){ if( x==(-3) ){ j = pParse->iErr; - if( pParse->nNode!=(u32)iThis+1 ) pParse->hasNonstd = 1; + if( pParse->nBlob!=iStart ) pParse->hasNonstd = 1; break; } if( x!=(-1) ) pParse->iErr = j; @@ -204027,15 +204791,15 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ }else if( z[j]==']' ){ break; }else{ - if( fast_isspace(z[j]) ){ - do{ j++; }while( fast_isspace(z[j]) ); + if( jsonIsspace(z[j]) ){ + j += 1 + (u32)strspn(&z[j+1], jsonSpaces); if( z[j]==',' ){ continue; }else if( z[j]==']' ){ break; } } - x = jsonParseValue(pParse, j); + x = jsonTranslateTextToBlob(pParse, j); if( x==(-4) ){ j = pParse->iErr; continue; @@ -204048,23 +204812,33 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ pParse->iErr = j; return -1; } - pParse->aNode[iThis].n = pParse->nNode - (u32)iThis - 1; + jsonBlobChangePayloadSize(pParse, iThis, pParse->nBlob - iStart); pParse->iDepth--; return j+1; } case '\'': { - u8 jnFlags; + u8 opcode; char cDelim; pParse->hasNonstd = 1; - jnFlags = JNODE_JSON5; + opcode = JSONB_TEXT; goto parse_string; case '"': /* Parse string */ - jnFlags = 0; + opcode = JSONB_TEXT; parse_string: cDelim = z[i]; - for(j=i+1; 1; j++){ - if( jsonIsOk[(unsigned char)z[j]] ) continue; + j = i+1; + while( 1 /*exit-by-break*/ ){ + if( jsonIsOk[(u8)z[j]] ){ + if( !jsonIsOk[(u8)z[j+1]] ){ + j += 1; + }else if( !jsonIsOk[(u8)z[j+2]] ){ + j += 2; + }else{ + j += 3; + continue; + } + } c = z[j]; if( c==cDelim ){ break; @@ -204073,16 +204847,16 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ if( c=='"' || c=='\\' || c=='/' || c=='b' || c=='f' || c=='n' || c=='r' || c=='t' || (c=='u' && jsonIs4Hex(&z[j+1])) ){ - jnFlags |= JNODE_ESCAPE; + if( opcode==JSONB_TEXT ) opcode = JSONB_TEXTJ; }else if( c=='\'' || c=='0' || c=='v' || c=='\n' || (0xe2==(u8)c && 0x80==(u8)z[j+1] && (0xa8==(u8)z[j+2] || 0xa9==(u8)z[j+2])) || (c=='x' && jsonIs2Hex(&z[j+1])) ){ - jnFlags |= (JNODE_ESCAPE|JNODE_JSON5); + opcode = JSONB_TEXT5; pParse->hasNonstd = 1; }else if( c=='\r' ){ if( z[j+1]=='\n' ) j++; - jnFlags |= (JNODE_ESCAPE|JNODE_JSON5); + opcode = JSONB_TEXT5; pParse->hasNonstd = 1; }else{ pParse->iErr = j; @@ -204092,14 +204866,17 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ /* Control characters are not allowed in strings */ pParse->iErr = j; return -1; + }else if( c=='"' ){ + opcode = JSONB_TEXT5; } + j++; } - jsonParseAddNode(pParse, JSON_STRING | (jnFlags<<8), j+1-i, &z[i]); + jsonBlobAppendNode(pParse, opcode, j-1-i, &z[i+1]); return j+1; } case 't': { if( strncmp(z+i,"true",4)==0 && !sqlite3Isalnum(z[i+4]) ){ - jsonParseAddNode(pParse, JSON_TRUE, 0, 0); + jsonBlobAppendOneByte(pParse, JSONB_TRUE); return i+4; } pParse->iErr = i; @@ -204107,23 +204884,22 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ } case 'f': { if( strncmp(z+i,"false",5)==0 && !sqlite3Isalnum(z[i+5]) ){ - jsonParseAddNode(pParse, JSON_FALSE, 0, 0); + jsonBlobAppendOneByte(pParse, JSONB_FALSE); return i+5; } pParse->iErr = i; return -1; } case '+': { - u8 seenDP, seenE, jnFlags; + u8 seenE; pParse->hasNonstd = 1; - jnFlags = JNODE_JSON5; + t = 0x00; /* Bit 0x01: JSON5. Bit 0x02: FLOAT */ goto parse_number; case '.': if( sqlite3Isdigit(z[i+1]) ){ pParse->hasNonstd = 1; - jnFlags = JNODE_JSON5; + t = 0x03; /* Bit 0x01: JSON5. Bit 0x02: FLOAT */ seenE = 0; - seenDP = JSON_REAL; goto parse_number_2; } pParse->iErr = i; @@ -204140,9 +204916,8 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ case '8': case '9': /* Parse number */ - jnFlags = 0; + t = 0x00; /* Bit 0x01: JSON5. Bit 0x02: FLOAT */ parse_number: - seenDP = JSON_INT; seenE = 0; assert( '-' < '0' ); assert( '+' < '0' ); @@ -204152,9 +204927,9 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ if( c<='0' ){ if( c=='0' ){ if( (z[i+1]=='x' || z[i+1]=='X') && sqlite3Isxdigit(z[i+2]) ){ - assert( seenDP==JSON_INT ); + assert( t==0x00 ); pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t = 0x01; for(j=i+3; sqlite3Isxdigit(z[j]); j++){} goto parse_number_finish; }else if( sqlite3Isdigit(z[i+1]) ){ @@ -204171,15 +204946,15 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ ){ pParse->hasNonstd = 1; if( z[i]=='-' ){ - jsonParseAddNode(pParse, JSON_REAL, 8, "-9.0e999"); + jsonBlobAppendNode(pParse, JSONB_FLOAT, 6, "-9e999"); }else{ - jsonParseAddNode(pParse, JSON_REAL, 7, "9.0e999"); + jsonBlobAppendNode(pParse, JSONB_FLOAT, 5, "9e999"); } return i + (sqlite3StrNICmp(&z[i+4],"inity",5)==0 ? 9 : 4); } if( z[i+1]=='.' ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; goto parse_number_2; } pParse->iErr = i; @@ -204191,30 +204966,31 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ return -1; }else if( (z[i+2]=='x' || z[i+2]=='X') && sqlite3Isxdigit(z[i+3]) ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; for(j=i+4; sqlite3Isxdigit(z[j]); j++){} goto parse_number_finish; } } } } + parse_number_2: for(j=i+1;; j++){ c = z[j]; if( sqlite3Isdigit(c) ) continue; if( c=='.' ){ - if( seenDP==JSON_REAL ){ + if( (t & 0x02)!=0 ){ pParse->iErr = j; return -1; } - seenDP = JSON_REAL; + t |= 0x02; continue; } if( c=='e' || c=='E' ){ if( z[j-1]<'0' ){ if( ALWAYS(z[j-1]=='.') && ALWAYS(j-2>=i) && sqlite3Isdigit(z[j-2]) ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; }else{ pParse->iErr = j; return -1; @@ -204224,7 +205000,7 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ pParse->iErr = j; return -1; } - seenDP = JSON_REAL; + t |= 0x02; seenE = 1; c = z[j+1]; if( c=='+' || c=='-' ){ @@ -204242,14 +205018,18 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ if( z[j-1]<'0' ){ if( ALWAYS(z[j-1]=='.') && ALWAYS(j-2>=i) && sqlite3Isdigit(z[j-2]) ){ pParse->hasNonstd = 1; - jnFlags |= JNODE_JSON5; + t |= 0x01; }else{ pParse->iErr = j; return -1; } } parse_number_finish: - jsonParseAddNode(pParse, seenDP | (jnFlags<<8), j - i, &z[i]); + assert( JSONB_INT+0x01==JSONB_INT5 ); + assert( JSONB_FLOAT+0x01==JSONB_FLOAT5 ); + assert( JSONB_INT+0x02==JSONB_FLOAT ); + if( z[i]=='+' ) i++; + jsonBlobAppendNode(pParse, JSONB_INT+t, j-i, &z[i]); return j; } case '}': { @@ -204275,9 +205055,7 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ case 0x0a: case 0x0d: case 0x20: { - do{ - i++; - }while( fast_isspace(z[i]) ); + i += 1 + (u32)strspn(&z[i+1], jsonSpaces); goto json_parse_restart; } case 0x0b: @@ -204299,7 +205077,7 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ } case 'n': { if( strncmp(z+i,"null",4)==0 && !sqlite3Isalnum(z[i+4]) ){ - jsonParseAddNode(pParse, JSON_NULL, 0, 0); + jsonBlobAppendOneByte(pParse, JSONB_NULL); return i+4; } /* fall-through into the default case that checks for NaN */ @@ -204315,8 +205093,11 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ continue; } if( sqlite3Isalnum(z[i+nn]) ) continue; - jsonParseAddNode(pParse, aNanInfName[k].eType, - aNanInfName[k].nRepl, aNanInfName[k].zRepl); + if( aNanInfName[k].eType==JSONB_FLOAT ){ + jsonBlobAppendNode(pParse, JSONB_FLOAT, 5, "9e999"); + }else{ + jsonBlobAppendOneByte(pParse, JSONB_NULL); + } pParse->hasNonstd = 1; return i + nn; } @@ -204326,6 +205107,7 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ } /* End switch(z[i]) */ } + /* ** Parse a complete JSON string. Return 0 on success or non-zero if there ** are any errors. If an error occurs, free all memory held by pParse, @@ -204334,20 +205116,26 @@ static int jsonParseValue(JsonParse *pParse, u32 i){ ** pParse must be initialized to an empty parse object prior to calling ** this routine. */ -static int jsonParse( +static int jsonConvertTextToBlob( JsonParse *pParse, /* Initialize and fill this JsonParse object */ sqlite3_context *pCtx /* Report errors here */ ){ int i; const char *zJson = pParse->zJson; - i = jsonParseValue(pParse, 0); + i = jsonTranslateTextToBlob(pParse, 0); if( pParse->oom ) i = -1; if( i>0 ){ +#ifdef SQLITE_DEBUG assert( pParse->iDepth==0 ); - while( fast_isspace(zJson[i]) ) i++; + if( sqlite3Config.bJsonSelfcheck ){ + assert( jsonbValidityCheck(pParse, 0, pParse->nBlob, 0)==0 ); + } +#endif + while( jsonIsspace(zJson[i]) ) i++; if( zJson[i] ){ i += json5Whitespace(&zJson[i]); if( zJson[i] ){ + if( pCtx ) sqlite3_result_error(pCtx, "malformed JSON", -1); jsonParseReset(pParse); return 1; } @@ -204368,248 +205156,710 @@ static int jsonParse( return 0; } +/* +** The input string pStr is a well-formed JSON text string. Convert +** this into the JSONB format and make it the return value of the +** SQL function. +*/ +static void jsonReturnStringAsBlob(JsonString *pStr){ + JsonParse px; + memset(&px, 0, sizeof(px)); + jsonStringTerminate(pStr); + px.zJson = pStr->zBuf; + px.nJson = pStr->nUsed; + px.db = sqlite3_context_db_handle(pStr->pCtx); + (void)jsonTranslateTextToBlob(&px, 0); + if( px.oom ){ + sqlite3DbFree(px.db, px.aBlob); + sqlite3_result_error_nomem(pStr->pCtx); + }else{ + assert( px.nBlobAlloc>0 ); + assert( !px.bReadOnly ); + sqlite3_result_blob(pStr->pCtx, px.aBlob, px.nBlob, SQLITE_DYNAMIC); + } +} -/* Mark node i of pParse as being a child of iParent. Call recursively -** to fill in all the descendants of node i. +/* The byte at index i is a node type-code. This routine +** determines the payload size for that node and writes that +** payload size in to *pSz. It returns the offset from i to the +** beginning of the payload. Return 0 on error. */ -static void jsonParseFillInParentage(JsonParse *pParse, u32 i, u32 iParent){ - JsonNode *pNode = &pParse->aNode[i]; - u32 j; - pParse->aUp[i] = iParent; - switch( pNode->eType ){ - case JSON_ARRAY: { - for(j=1; j<=pNode->n; j += jsonNodeSize(pNode+j)){ - jsonParseFillInParentage(pParse, i+j, i); +static u32 jsonbPayloadSize(const JsonParse *pParse, u32 i, u32 *pSz){ + u8 x; + u32 sz; + u32 n; + if( NEVER(i>pParse->nBlob) ){ + *pSz = 0; + return 0; + } + x = pParse->aBlob[i]>>4; + if( x<=11 ){ + sz = x; + n = 1; + }else if( x==12 ){ + if( i+1>=pParse->nBlob ){ + *pSz = 0; + return 0; + } + sz = pParse->aBlob[i+1]; + n = 2; + }else if( x==13 ){ + if( i+2>=pParse->nBlob ){ + *pSz = 0; + return 0; + } + sz = (pParse->aBlob[i+1]<<8) + pParse->aBlob[i+2]; + n = 3; + }else if( x==14 ){ + if( i+4>=pParse->nBlob ){ + *pSz = 0; + return 0; + } + sz = ((u32)pParse->aBlob[i+1]<<24) + (pParse->aBlob[i+2]<<16) + + (pParse->aBlob[i+3]<<8) + pParse->aBlob[i+4]; + n = 5; + }else{ + if( i+8>=pParse->nBlob + || pParse->aBlob[i+1]!=0 + || pParse->aBlob[i+2]!=0 + || pParse->aBlob[i+3]!=0 + || pParse->aBlob[i+4]!=0 + ){ + *pSz = 0; + return 0; + } + sz = (pParse->aBlob[i+5]<<24) + (pParse->aBlob[i+6]<<16) + + (pParse->aBlob[i+7]<<8) + pParse->aBlob[i+8]; + n = 9; + } + if( i+sz+n > pParse->nBlob + && i+sz+n > pParse->nBlob-pParse->delta + ){ + sz = 0; + n = 0; + } + *pSz = sz; + return n; +} + + +/* +** Translate the binary JSONB representation of JSON beginning at +** pParse->aBlob[i] into a JSON text string. Append the JSON +** text onto the end of pOut. Return the index in pParse->aBlob[] +** of the first byte past the end of the element that is translated. +** +** If an error is detected in the BLOB input, the pOut->eErr flag +** might get set to JSTRING_MALFORMED. But not all BLOB input errors +** are detected. So a malformed JSONB input might either result +** in an error, or in incorrect JSON. +** +** The pOut->eErr JSTRING_OOM flag is set on a OOM. +*/ +static u32 jsonTranslateBlobToText( + const JsonParse *pParse, /* the complete parse of the JSON */ + u32 i, /* Start rendering at this index */ + JsonString *pOut /* Write JSON here */ +){ + u32 sz, n, j, iEnd; + + n = jsonbPayloadSize(pParse, i, &sz); + if( n==0 ){ + pOut->eErr |= JSTRING_MALFORMED; + return pParse->nBlob+1; + } + switch( pParse->aBlob[i] & 0x0f ){ + case JSONB_NULL: { + jsonAppendRawNZ(pOut, "null", 4); + return i+1; + } + case JSONB_TRUE: { + jsonAppendRawNZ(pOut, "true", 4); + return i+1; + } + case JSONB_FALSE: { + jsonAppendRawNZ(pOut, "false", 5); + return i+1; + } + case JSONB_INT: + case JSONB_FLOAT: { + jsonAppendRaw(pOut, (const char*)&pParse->aBlob[i+n], sz); + break; + } + case JSONB_INT5: { /* Integer literal in hexadecimal notation */ + u32 k = 2; + sqlite3_uint64 u = 0; + const char *zIn = (const char*)&pParse->aBlob[i+n]; + int bOverflow = 0; + if( zIn[0]=='-' ){ + jsonAppendChar(pOut, '-'); + k++; + }else if( zIn[0]=='+' ){ + k++; + } + for(; keErr |= JSTRING_MALFORMED; + break; + }else if( (u>>60)!=0 ){ + bOverflow = 1; + }else{ + u = u*16 + sqlite3HexToInt(zIn[k]); + } } + jsonPrintf(100,pOut,bOverflow?"9.0e999":"%llu", u); break; } - case JSON_OBJECT: { - for(j=1; j<=pNode->n; j += jsonNodeSize(pNode+j+1)+1){ - pParse->aUp[i+j] = i; - jsonParseFillInParentage(pParse, i+j+1, i); + case JSONB_FLOAT5: { /* Float literal missing digits beside "." */ + u32 k = 0; + const char *zIn = (const char*)&pParse->aBlob[i+n]; + if( zIn[0]=='-' ){ + jsonAppendChar(pOut, '-'); + k++; } + if( zIn[k]=='.' ){ + jsonAppendChar(pOut, '0'); + } + for(; kaBlob[i+n], sz); + jsonAppendChar(pOut, '"'); break; } + case JSONB_TEXT5: { + const char *zIn; + u32 k; + u32 sz2 = sz; + zIn = (const char*)&pParse->aBlob[i+n]; + jsonAppendChar(pOut, '"'); + while( sz2>0 ){ + for(k=0; k0 ){ + jsonAppendRawNZ(pOut, zIn, k); + if( k>=sz2 ){ + break; + } + zIn += k; + sz2 -= k; + } + if( zIn[0]=='"' ){ + jsonAppendRawNZ(pOut, "\\\"", 2); + zIn++; + sz2--; + continue; + } + assert( zIn[0]=='\\' ); + assert( sz2>=1 ); + if( sz2<2 ){ + pOut->eErr |= JSTRING_MALFORMED; + break; + } + switch( (u8)zIn[1] ){ + case '\'': + jsonAppendChar(pOut, '\''); + break; + case 'v': + jsonAppendRawNZ(pOut, "\\u0009", 6); + break; + case 'x': + if( sz2<4 ){ + pOut->eErr |= JSTRING_MALFORMED; + sz2 = 2; + break; + } + jsonAppendRawNZ(pOut, "\\u00", 4); + jsonAppendRawNZ(pOut, &zIn[2], 2); + zIn += 2; + sz2 -= 2; + break; + case '0': + jsonAppendRawNZ(pOut, "\\u0000", 6); + break; + case '\r': + if( sz2>2 && zIn[2]=='\n' ){ + zIn++; + sz2--; + } + break; + case '\n': + break; + case 0xe2: + /* '\' followed by either U+2028 or U+2029 is ignored as + ** whitespace. Not that in UTF8, U+2028 is 0xe2 0x80 0x29. + ** U+2029 is the same except for the last byte */ + if( sz2<4 + || 0x80!=(u8)zIn[2] + || (0xa8!=(u8)zIn[3] && 0xa9!=(u8)zIn[3]) + ){ + pOut->eErr |= JSTRING_MALFORMED; + sz2 = 2; + break; + } + zIn += 2; + sz2 -= 2; + break; + default: + jsonAppendRawNZ(pOut, zIn, 2); + break; + } + assert( sz2>=2 ); + zIn += 2; + sz2 -= 2; + } + jsonAppendChar(pOut, '"'); + break; + } + case JSONB_TEXTRAW: { + jsonAppendString(pOut, (const char*)&pParse->aBlob[i+n], sz); + break; + } + case JSONB_ARRAY: { + jsonAppendChar(pOut, '['); + j = i+n; + iEnd = j+sz; + while( j0 ) pOut->nUsed--; + jsonAppendChar(pOut, ']'); + break; + } + case JSONB_OBJECT: { + int x = 0; + jsonAppendChar(pOut, '{'); + j = i+n; + iEnd = j+sz; + while( jeErr |= JSTRING_MALFORMED; + if( sz>0 ) pOut->nUsed--; + jsonAppendChar(pOut, '}'); + break; + } + default: { + pOut->eErr |= JSTRING_MALFORMED; break; } } + return i+n+sz; +} + +/* Return true if the input pJson +** +** For performance reasons, this routine does not do a detailed check of the +** input BLOB to ensure that it is well-formed. Hence, false positives are +** possible. False negatives should never occur, however. +*/ +static int jsonFuncArgMightBeBinary(sqlite3_value *pJson){ + u32 sz, n; + const u8 *aBlob; + int nBlob; + JsonParse s; + if( sqlite3_value_type(pJson)!=SQLITE_BLOB ) return 0; + aBlob = sqlite3_value_blob(pJson); + nBlob = sqlite3_value_bytes(pJson); + if( nBlob<1 ) return 0; + if( NEVER(aBlob==0) || (aBlob[0] & 0x0f)>JSONB_OBJECT ) return 0; + memset(&s, 0, sizeof(s)); + s.aBlob = (u8*)aBlob; + s.nBlob = nBlob; + n = jsonbPayloadSize(&s, 0, &sz); + if( n==0 ) return 0; + if( sz+n!=(u32)nBlob ) return 0; + if( (aBlob[0] & 0x0f)<=JSONB_FALSE && sz>0 ) return 0; + return sz+n==(u32)nBlob; } /* -** Compute the parentage of all nodes in a completed parse. +** Given that a JSONB_ARRAY object starts at offset i, return +** the number of entries in that array. */ -static int jsonParseFindParents(JsonParse *pParse){ - u32 *aUp; - assert( pParse->aUp==0 ); - aUp = pParse->aUp = sqlite3_malloc64( sizeof(u32)*pParse->nNode ); - if( aUp==0 ){ - pParse->oom = 1; - return SQLITE_NOMEM; +static u32 jsonbArrayCount(JsonParse *pParse, u32 iRoot){ + u32 n, sz, i, iEnd; + u32 k = 0; + n = jsonbPayloadSize(pParse, iRoot, &sz); + iEnd = iRoot+n+sz; + for(i=iRoot+n; n>0 && idelta. */ -#define JSON_CACHE_ID (-429938) /* First cache entry */ -#define JSON_CACHE_SZ 4 /* Max number of cache entries */ +static void jsonAfterEditSizeAdjust(JsonParse *pParse, u32 iRoot){ + u32 sz = 0; + u32 nBlob; + assert( pParse->delta!=0 ); + assert( pParse->nBlobAlloc >= pParse->nBlob ); + nBlob = pParse->nBlob; + pParse->nBlob = pParse->nBlobAlloc; + (void)jsonbPayloadSize(pParse, iRoot, &sz); + pParse->nBlob = nBlob; + sz += pParse->delta; + pParse->delta += jsonBlobChangePayloadSize(pParse, iRoot, sz); +} /* -** Obtain a complete parse of the JSON found in the pJson argument -** -** Use the sqlite3_get_auxdata() cache to find a preexisting parse -** if it is available. If the cache is not available or if it -** is no longer valid, parse the JSON again and return the new parse. -** Also register the new parse so that it will be available for -** future sqlite3_get_auxdata() calls. -** -** If an error occurs and pErrCtx!=0 then report the error on pErrCtx -** and return NULL. +** Modify the JSONB blob at pParse->aBlob by removing nDel bytes of +** content beginning at iDel, and replacing them with nIns bytes of +** content given by aIns. ** -** The returned pointer (if it is not NULL) is owned by the cache in -** most cases, not the caller. The caller does NOT need to invoke -** jsonParseFree(), in most cases. +** nDel may be zero, in which case no bytes are removed. But iDel is +** still important as new bytes will be insert beginning at iDel. ** -** Except, if an error occurs and pErrCtx==0 then return the JsonParse -** object with JsonParse.nErr non-zero and the caller will own the JsonParse -** object. In that case, it will be the responsibility of the caller to -** invoke jsonParseFree(). To summarize: +** aIns may be zero, in which case space is created to hold nIns bytes +** beginning at iDel, but that space is uninitialized. ** -** pErrCtx!=0 || p->nErr==0 ==> Return value p is owned by the -** cache. Call does not need to -** free it. -** -** pErrCtx==0 && p->nErr!=0 ==> Return value is owned by the caller -** and so the caller must free it. +** Set pParse->oom if an OOM occurs. */ -static JsonParse *jsonParseCached( - sqlite3_context *pCtx, /* Context to use for cache search */ - sqlite3_value *pJson, /* Function param containing JSON text */ - sqlite3_context *pErrCtx, /* Write parse errors here if not NULL */ - int bUnedited /* No prior edits allowed */ +static void jsonBlobEdit( + JsonParse *pParse, /* The JSONB to be modified is in pParse->aBlob */ + u32 iDel, /* First byte to be removed */ + u32 nDel, /* Number of bytes to remove */ + const u8 *aIns, /* Content to insert */ + u32 nIns /* Bytes of content to insert */ ){ - char *zJson = (char*)sqlite3_value_text(pJson); - int nJson = sqlite3_value_bytes(pJson); - JsonParse *p; - JsonParse *pMatch = 0; - int iKey; - int iMinKey = 0; - u32 iMinHold = 0xffffffff; - u32 iMaxHold = 0; - int bJsonRCStr; + i64 d = (i64)nIns - (i64)nDel; + if( d!=0 ){ + if( pParse->nBlob + d > pParse->nBlobAlloc ){ + jsonBlobExpand(pParse, pParse->nBlob+d); + if( pParse->oom ) return; + } + memmove(&pParse->aBlob[iDel+nIns], + &pParse->aBlob[iDel+nDel], + pParse->nBlob - (iDel+nDel)); + pParse->nBlob += d; + pParse->delta += d; + } + if( nIns && aIns ) memcpy(&pParse->aBlob[iDel], aIns, nIns); +} - if( zJson==0 ) return 0; - for(iKey=0; iKeynJson==nJson - && (p->hasMod==0 || bUnedited==0) - && (p->zJson==zJson || memcmp(p->zJson,zJson,nJson)==0) - ){ - p->nErr = 0; - p->useMod = 0; - pMatch = p; - }else - if( pMatch==0 - && p->zAlt!=0 - && bUnedited==0 - && p->nAlt==nJson - && memcmp(p->zAlt, zJson, nJson)==0 - ){ - p->nErr = 0; - p->useMod = 1; - pMatch = p; - }else if( p->iHoldiHold; - iMinKey = iKey; + if( z[i+1]=='\r' ){ + if( i+2iHold>iMaxHold ){ - iMaxHold = p->iHold; + if( 0xe2==(u8)z[i+1] + && i+3nErr = 0; - pMatch->iHold = iMaxHold+1; - assert( pMatch->nJPRef>0 ); /* pMatch is owned by the cache */ - return pMatch; - } + return i; +} - /* The input JSON was not found anywhere in the cache. We will need - ** to parse it ourselves and generate a new JsonParse object. - */ - bJsonRCStr = sqlite3ValueIsOfClass(pJson,sqlite3RCStrUnref); - p = sqlite3_malloc64( sizeof(*p) + (bJsonRCStr ? 0 : nJson+1) ); - if( p==0 ){ - sqlite3_result_error_nomem(pCtx); - return 0; +/* +** Input z[0..n] defines JSON escape sequence including the leading '\\'. +** Decode that escape sequence into a single character. Write that +** character into *piOut. Return the number of bytes in the escape sequence. +** +** If there is a syntax error of some kind (for example too few characters +** after the '\\' to complete the encoding) then *piOut is set to +** JSON_INVALID_CHAR. +*/ +static u32 jsonUnescapeOneChar(const char *z, u32 n, u32 *piOut){ + assert( n>0 ); + assert( z[0]=='\\' ); + if( n<2 ){ + *piOut = JSON_INVALID_CHAR; + return n; } - memset(p, 0, sizeof(*p)); - if( bJsonRCStr ){ - p->zJson = sqlite3RCStrRef(zJson); - p->bJsonIsRCStr = 1; - }else{ - p->zJson = (char*)&p[1]; - memcpy(p->zJson, zJson, nJson+1); + switch( (u8)z[1] ){ + case 'u': { + u32 v, vlo; + if( n<6 ){ + *piOut = JSON_INVALID_CHAR; + return n; + } + v = jsonHexToInt4(&z[2]); + if( (v & 0xfc00)==0xd800 + && n>=12 + && z[6]=='\\' + && z[7]=='u' + && ((vlo = jsonHexToInt4(&z[8]))&0xfc00)==0xdc00 + ){ + *piOut = ((v&0x3ff)<<10) + (vlo&0x3ff) + 0x10000; + return 12; + }else{ + *piOut = v; + return 6; + } + } + case 'b': { *piOut = '\b'; return 2; } + case 'f': { *piOut = '\f'; return 2; } + case 'n': { *piOut = '\n'; return 2; } + case 'r': { *piOut = '\r'; return 2; } + case 't': { *piOut = '\t'; return 2; } + case 'v': { *piOut = '\v'; return 2; } + case '0': { *piOut = 0; return 2; } + case '\'': + case '"': + case '/': + case '\\':{ *piOut = z[1]; return 2; } + case 'x': { + if( n<4 ){ + *piOut = JSON_INVALID_CHAR; + return n; + } + *piOut = (jsonHexToInt(z[2])<<4) | jsonHexToInt(z[3]); + return 4; + } + case 0xe2: + case '\r': + case '\n': { + u32 nSkip = jsonBytesToBypass(z, n); + if( nSkip==0 ){ + *piOut = JSON_INVALID_CHAR; + return n; + }else if( nSkip==n ){ + *piOut = 0; + return n; + }else if( z[nSkip]=='\\' ){ + return nSkip + jsonUnescapeOneChar(&z[nSkip], n-nSkip, piOut); + }else{ + int sz = sqlite3Utf8ReadLimited((u8*)&z[nSkip], n-nSkip, piOut); + return nSkip + sz; + } + } + default: { + *piOut = JSON_INVALID_CHAR; + return 2; + } } - p->nJPRef = 1; - if( jsonParse(p, pErrCtx) ){ - if( pErrCtx==0 ){ - p->nErr = 1; - assert( p->nJPRef==1 ); /* Caller will own the new JsonParse object p */ - return p; +} + + +/* +** Compare two object labels. Return 1 if they are equal and +** 0 if they differ. +** +** In this version, we know that one or the other or both of the +** two comparands contains an escape sequence. +*/ +static SQLITE_NOINLINE int jsonLabelCompareEscaped( + const char *zLeft, /* The left label */ + u32 nLeft, /* Size of the left label in bytes */ + int rawLeft, /* True if zLeft contains no escapes */ + const char *zRight, /* The right label */ + u32 nRight, /* Size of the right label in bytes */ + int rawRight /* True if zRight is escape-free */ +){ + u32 cLeft, cRight; + assert( rawLeft==0 || rawRight==0 ); + while( 1 /*exit-by-return*/ ){ + if( nLeft==0 ){ + cLeft = 0; + }else if( rawLeft || zLeft[0]!='\\' ){ + cLeft = ((u8*)zLeft)[0]; + if( cLeft>=0xc0 ){ + int sz = sqlite3Utf8ReadLimited((u8*)zLeft, nLeft, &cLeft); + zLeft += sz; + nLeft -= sz; + }else{ + zLeft++; + nLeft--; + } + }else{ + u32 n = jsonUnescapeOneChar(zLeft, nLeft, &cLeft); + zLeft += n; + assert( n<=nLeft ); + nLeft -= n; + } + if( nRight==0 ){ + cRight = 0; + }else if( rawRight || zRight[0]!='\\' ){ + cRight = ((u8*)zRight)[0]; + if( cRight>=0xc0 ){ + int sz = sqlite3Utf8ReadLimited((u8*)zRight, nRight, &cRight); + zRight += sz; + nRight -= sz; + }else{ + zRight++; + nRight--; + } + }else{ + u32 n = jsonUnescapeOneChar(zRight, nRight, &cRight); + zRight += n; + assert( n<=nRight ); + nRight -= n; } - jsonParseFree(p); - return 0; + if( cLeft!=cRight ) return 0; + if( cLeft==0 ) return 1; } - p->nJson = nJson; - p->iHold = iMaxHold+1; - /* Transfer ownership of the new JsonParse to the cache */ - sqlite3_set_auxdata(pCtx, JSON_CACHE_ID+iMinKey, p, - (void(*)(void*))jsonParseFree); - return (JsonParse*)sqlite3_get_auxdata(pCtx, JSON_CACHE_ID+iMinKey); } /* -** Compare the OBJECT label at pNode against zKey,nKey. Return true on -** a match. +** Compare two object labels. Return 1 if they are equal and +** 0 if they differ. Return -1 if an OOM occurs. */ -static int jsonLabelCompare(const JsonNode *pNode, const char *zKey, u32 nKey){ - assert( pNode->eU==1 ); - if( pNode->jnFlags & JNODE_RAW ){ - if( pNode->n!=nKey ) return 0; - return strncmp(pNode->u.zJContent, zKey, nKey)==0; +static int jsonLabelCompare( + const char *zLeft, /* The left label */ + u32 nLeft, /* Size of the left label in bytes */ + int rawLeft, /* True if zLeft contains no escapes */ + const char *zRight, /* The right label */ + u32 nRight, /* Size of the right label in bytes */ + int rawRight /* True if zRight is escape-free */ +){ + if( rawLeft && rawRight ){ + /* Simpliest case: Neither label contains escapes. A simple + ** memcmp() is sufficient. */ + if( nLeft!=nRight ) return 0; + return memcmp(zLeft, zRight, nLeft)==0; }else{ - if( pNode->n!=nKey+2 ) return 0; - return strncmp(pNode->u.zJContent+1, zKey, nKey)==0; + return jsonLabelCompareEscaped(zLeft, nLeft, rawLeft, + zRight, nRight, rawRight); } } -static int jsonSameLabel(const JsonNode *p1, const JsonNode *p2){ - if( p1->jnFlags & JNODE_RAW ){ - return jsonLabelCompare(p2, p1->u.zJContent, p1->n); - }else if( p2->jnFlags & JNODE_RAW ){ - return jsonLabelCompare(p1, p2->u.zJContent, p2->n); + +/* +** Error returns from jsonLookupStep() +*/ +#define JSON_LOOKUP_ERROR 0xffffffff +#define JSON_LOOKUP_NOTFOUND 0xfffffffe +#define JSON_LOOKUP_PATHERROR 0xfffffffd +#define JSON_LOOKUP_ISERROR(x) ((x)>=JSON_LOOKUP_PATHERROR) + +/* Forward declaration */ +static u32 jsonLookupStep(JsonParse*,u32,const char*,u32); + + +/* This helper routine for jsonLookupStep() populates pIns with +** binary data that is to be inserted into pParse. +** +** In the common case, pIns just points to pParse->aIns and pParse->nIns. +** But if the zPath of the original edit operation includes path elements +** that go deeper, additional substructure must be created. +** +** For example: +** +** json_insert('{}', '$.a.b.c', 123); +** +** The search stops at '$.a' But additional substructure must be +** created for the ".b.c" part of the patch so that the final result +** is: {"a":{"b":{"c"::123}}}. This routine populates pIns with +** the binary equivalent of {"b":{"c":123}} so that it can be inserted. +** +** The caller is responsible for resetting pIns when it has finished +** using the substructure. +*/ +static u32 jsonCreateEditSubstructure( + JsonParse *pParse, /* The original JSONB that is being edited */ + JsonParse *pIns, /* Populate this with the blob data to insert */ + const char *zTail /* Tail of the path that determins substructure */ +){ + static const u8 emptyObject[] = { JSONB_ARRAY, JSONB_OBJECT }; + int rc; + memset(pIns, 0, sizeof(*pIns)); + pIns->db = pParse->db; + if( zTail[0]==0 ){ + /* No substructure. Just insert what is given in pParse. */ + pIns->aBlob = pParse->aIns; + pIns->nBlob = pParse->nIns; + rc = 0; }else{ - return p1->n==p2->n && strncmp(p1->u.zJContent,p2->u.zJContent,p1->n)==0; + /* Construct the binary substructure */ + pIns->nBlob = 1; + pIns->aBlob = (u8*)&emptyObject[zTail[0]=='.']; + pIns->eEdit = pParse->eEdit; + pIns->nIns = pParse->nIns; + pIns->aIns = pParse->aIns; + rc = jsonLookupStep(pIns, 0, zTail, 0); + pParse->oom |= pIns->oom; } + return rc; /* Error code only */ } -/* forward declaration */ -static JsonNode *jsonLookupAppend(JsonParse*,const char*,int*,const char**); - /* -** Search along zPath to find the node specified. Return a pointer -** to that node, or NULL if zPath is malformed or if there is no such -** node. +** Search along zPath to find the Json element specified. Return an +** index into pParse->aBlob[] for the start of that element's value. ** -** If pApnd!=0, then try to append new nodes to complete zPath if it is -** possible to do so and if no existing node corresponds to zPath. If -** new nodes are appended *pApnd is set to 1. +** If the value found by this routine is the value half of label/value pair +** within an object, then set pPath->iLabel to the start of the corresponding +** label, before returning. +** +** Return one of the JSON_LOOKUP error codes if problems are seen. +** +** This routine will also modify the blob. If pParse->eEdit is one of +** JEDIT_DEL, JEDIT_REPL, JEDIT_INS, or JEDIT_SET, then changes might be +** made to the selected value. If an edit is performed, then the return +** value does not necessarily point to the select element. If an edit +** is performed, the return value is only useful for detecting error +** conditions. */ -static JsonNode *jsonLookupStep( +static u32 jsonLookupStep( JsonParse *pParse, /* The JSON to search */ - u32 iRoot, /* Begin the search at this node */ + u32 iRoot, /* Begin the search at this element of aBlob[] */ const char *zPath, /* The path to search */ - int *pApnd, /* Append nodes to complete path if not NULL */ - const char **pzErr /* Make *pzErr point to any syntax error in zPath */ + u32 iLabel /* Label if iRoot is a value of in an object */ ){ - u32 i, j, nKey; + u32 i, j, k, nKey, sz, n, iEnd, rc; const char *zKey; - JsonNode *pRoot; - if( pParse->oom ) return 0; - pRoot = &pParse->aNode[iRoot]; - if( pRoot->jnFlags & (JNODE_REPLACE|JNODE_REMOVE) && pParse->useMod ){ - while( (pRoot->jnFlags & JNODE_REPLACE)!=0 ){ - u32 idx = (u32)(pRoot - pParse->aNode); - i = pParse->iSubst; - while( 1 /*exit-by-break*/ ){ - assert( inNode ); - assert( pParse->aNode[i].eType==JSON_SUBST ); - assert( pParse->aNode[i].eU==4 ); - assert( pParse->aNode[i].u.iPrevaNode[i].n==idx ){ - pRoot = &pParse->aNode[i+1]; - iRoot = i+1; - break; - } - i = pParse->aNode[i].u.iPrev; + u8 x; + + if( zPath[0]==0 ){ + if( pParse->eEdit && jsonBlobMakeEditable(pParse, pParse->nIns) ){ + n = jsonbPayloadSize(pParse, iRoot, &sz); + sz += n; + if( pParse->eEdit==JEDIT_DEL ){ + if( iLabel>0 ){ + sz += iRoot - iLabel; + iRoot = iLabel; + } + jsonBlobEdit(pParse, iRoot, sz, 0, 0); + }else if( pParse->eEdit==JEDIT_INS ){ + /* Already exists, so json_insert() is a no-op */ + }else{ + /* json_set() or json_replace() */ + jsonBlobEdit(pParse, iRoot, sz, pParse->aIns, pParse->nIns); } } - if( pRoot->jnFlags & JNODE_REMOVE ){ - return 0; - } + pParse->iLabel = iLabel; + return iRoot; } - if( zPath[0]==0 ) return pRoot; if( zPath[0]=='.' ){ - if( pRoot->eType!=JSON_OBJECT ) return 0; + int rawKey = 1; + x = pParse->aBlob[iRoot]; zPath++; if( zPath[0]=='"' ){ zKey = zPath + 1; @@ -204618,315 +205868,802 @@ static JsonNode *jsonLookupStep( if( zPath[i] ){ i++; }else{ - *pzErr = zPath; - return 0; + return JSON_LOOKUP_PATHERROR; } testcase( nKey==0 ); + rawKey = memchr(zKey, '\\', nKey)==0; }else{ zKey = zPath; for(i=0; zPath[i] && zPath[i]!='.' && zPath[i]!='['; i++){} nKey = i; if( nKey==0 ){ - *pzErr = zPath; - return 0; - } - } - j = 1; - for(;;){ - while( j<=pRoot->n ){ - if( jsonLabelCompare(pRoot+j, zKey, nKey) ){ - return jsonLookupStep(pParse, iRoot+j+1, &zPath[i], pApnd, pzErr); - } - j++; - j += jsonNodeSize(&pRoot[j]); + return JSON_LOOKUP_PATHERROR; + } + } + if( (x & 0x0f)!=JSONB_OBJECT ) return JSON_LOOKUP_NOTFOUND; + n = jsonbPayloadSize(pParse, iRoot, &sz); + j = iRoot + n; /* j is the index of a label */ + iEnd = j+sz; + while( jaBlob[j] & 0x0f; + if( xJSONB_TEXTRAW ) return JSON_LOOKUP_ERROR; + n = jsonbPayloadSize(pParse, j, &sz); + if( n==0 ) return JSON_LOOKUP_ERROR; + k = j+n; /* k is the index of the label text */ + if( k+sz>=iEnd ) return JSON_LOOKUP_ERROR; + zLabel = (const char*)&pParse->aBlob[k]; + rawLabel = x==JSONB_TEXT || x==JSONB_TEXTRAW; + if( jsonLabelCompare(zKey, nKey, rawKey, zLabel, sz, rawLabel) ){ + u32 v = k+sz; /* v is the index of the value */ + if( ((pParse->aBlob[v])&0x0f)>JSONB_OBJECT ) return JSON_LOOKUP_ERROR; + n = jsonbPayloadSize(pParse, v, &sz); + if( n==0 || v+n+sz>iEnd ) return JSON_LOOKUP_ERROR; + assert( j>0 ); + rc = jsonLookupStep(pParse, v, &zPath[i], j); + if( pParse->delta ) jsonAfterEditSizeAdjust(pParse, iRoot); + return rc; } - if( (pRoot->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pRoot->eU==2 ); - iRoot = pRoot->u.iAppend; - pRoot = &pParse->aNode[iRoot]; - j = 1; - } - if( pApnd ){ - u32 iStart, iLabel; - JsonNode *pNode; - assert( pParse->useMod ); - iStart = jsonParseAddNode(pParse, JSON_OBJECT, 2, 0); - iLabel = jsonParseAddNode(pParse, JSON_STRING, nKey, zKey); - zPath += i; - pNode = jsonLookupAppend(pParse, zPath, pApnd, pzErr); - if( pParse->oom ) return 0; - if( pNode ){ - pRoot = &pParse->aNode[iRoot]; - assert( pRoot->eU==0 ); - pRoot->u.iAppend = iStart; - pRoot->jnFlags |= JNODE_APPEND; - VVA( pRoot->eU = 2 ); - pParse->aNode[iLabel].jnFlags |= JNODE_RAW; - } - return pNode; + j = k+sz; + if( ((pParse->aBlob[j])&0x0f)>JSONB_OBJECT ) return JSON_LOOKUP_ERROR; + n = jsonbPayloadSize(pParse, j, &sz); + if( n==0 ) return JSON_LOOKUP_ERROR; + j += n+sz; + } + if( j>iEnd ) return JSON_LOOKUP_ERROR; + if( pParse->eEdit>=JEDIT_INS ){ + u32 nIns; /* Total bytes to insert (label+value) */ + JsonParse v; /* BLOB encoding of the value to be inserted */ + JsonParse ix; /* Header of the label to be inserted */ + testcase( pParse->eEdit==JEDIT_INS ); + testcase( pParse->eEdit==JEDIT_SET ); + memset(&ix, 0, sizeof(ix)); + ix.db = pParse->db; + jsonBlobAppendNode(&ix, rawKey?JSONB_TEXTRAW:JSONB_TEXT5, nKey, 0); + pParse->oom |= ix.oom; + rc = jsonCreateEditSubstructure(pParse, &v, &zPath[i]); + if( !JSON_LOOKUP_ISERROR(rc) + && jsonBlobMakeEditable(pParse, ix.nBlob+nKey+v.nBlob) + ){ + assert( !pParse->oom ); + nIns = ix.nBlob + nKey + v.nBlob; + jsonBlobEdit(pParse, j, 0, 0, nIns); + if( !pParse->oom ){ + assert( pParse->aBlob!=0 ); /* Because pParse->oom!=0 */ + assert( ix.aBlob!=0 ); /* Because pPasre->oom!=0 */ + memcpy(&pParse->aBlob[j], ix.aBlob, ix.nBlob); + k = j + ix.nBlob; + memcpy(&pParse->aBlob[k], zKey, nKey); + k += nKey; + memcpy(&pParse->aBlob[k], v.aBlob, v.nBlob); + if( ALWAYS(pParse->delta) ) jsonAfterEditSizeAdjust(pParse, iRoot); + } + } + jsonParseReset(&v); + jsonParseReset(&ix); + return rc; } }else if( zPath[0]=='[' ){ - i = 0; - j = 1; - while( sqlite3Isdigit(zPath[j]) ){ - i = i*10 + zPath[j] - '0'; - j++; + x = pParse->aBlob[iRoot] & 0x0f; + if( x!=JSONB_ARRAY ) return JSON_LOOKUP_NOTFOUND; + n = jsonbPayloadSize(pParse, iRoot, &sz); + k = 0; + i = 1; + while( sqlite3Isdigit(zPath[i]) ){ + k = k*10 + zPath[i] - '0'; + i++; } - if( j<2 || zPath[j]!=']' ){ + if( i<2 || zPath[i]!=']' ){ if( zPath[1]=='#' ){ - JsonNode *pBase = pRoot; - int iBase = iRoot; - if( pRoot->eType!=JSON_ARRAY ) return 0; - for(;;){ - while( j<=pBase->n ){ - if( (pBase[j].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ) i++; - j += jsonNodeSize(&pBase[j]); - } - if( (pBase->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pBase->eU==2 ); - iBase = pBase->u.iAppend; - pBase = &pParse->aNode[iBase]; - j = 1; - } - j = 2; + k = jsonbArrayCount(pParse, iRoot); + i = 2; if( zPath[2]=='-' && sqlite3Isdigit(zPath[3]) ){ - unsigned int x = 0; - j = 3; + unsigned int nn = 0; + i = 3; do{ - x = x*10 + zPath[j] - '0'; - j++; - }while( sqlite3Isdigit(zPath[j]) ); - if( x>i ) return 0; - i -= x; + nn = nn*10 + zPath[i] - '0'; + i++; + }while( sqlite3Isdigit(zPath[i]) ); + if( nn>k ) return JSON_LOOKUP_NOTFOUND; + k -= nn; } - if( zPath[j]!=']' ){ - *pzErr = zPath; - return 0; + if( zPath[i]!=']' ){ + return JSON_LOOKUP_PATHERROR; } }else{ - *pzErr = zPath; - return 0; + return JSON_LOOKUP_PATHERROR; } } - if( pRoot->eType!=JSON_ARRAY ) return 0; - zPath += j + 1; - j = 1; - for(;;){ - while( j<=pRoot->n - && (i>0 || ((pRoot[j].jnFlags & JNODE_REMOVE)!=0 && pParse->useMod)) + j = iRoot+n; + iEnd = j+sz; + while( jdelta ) jsonAfterEditSizeAdjust(pParse, iRoot); + return rc; + } + k--; + n = jsonbPayloadSize(pParse, j, &sz); + if( n==0 ) return JSON_LOOKUP_ERROR; + j += n+sz; + } + if( j>iEnd ) return JSON_LOOKUP_ERROR; + if( k>0 ) return JSON_LOOKUP_NOTFOUND; + if( pParse->eEdit>=JEDIT_INS ){ + JsonParse v; + testcase( pParse->eEdit==JEDIT_INS ); + testcase( pParse->eEdit==JEDIT_SET ); + rc = jsonCreateEditSubstructure(pParse, &v, &zPath[i+1]); + if( !JSON_LOOKUP_ISERROR(rc) + && jsonBlobMakeEditable(pParse, v.nBlob) ){ - if( (pRoot[j].jnFlags & JNODE_REMOVE)==0 || pParse->useMod==0 ) i--; - j += jsonNodeSize(&pRoot[j]); - } - if( i==0 && j<=pRoot->n ) break; - if( (pRoot->jnFlags & JNODE_APPEND)==0 ) break; - if( pParse->useMod==0 ) break; - assert( pRoot->eU==2 ); - iRoot = pRoot->u.iAppend; - pRoot = &pParse->aNode[iRoot]; - j = 1; - } - if( j<=pRoot->n ){ - return jsonLookupStep(pParse, iRoot+j, zPath, pApnd, pzErr); - } - if( i==0 && pApnd ){ - u32 iStart; - JsonNode *pNode; - assert( pParse->useMod ); - iStart = jsonParseAddNode(pParse, JSON_ARRAY, 1, 0); - pNode = jsonLookupAppend(pParse, zPath, pApnd, pzErr); - if( pParse->oom ) return 0; - if( pNode ){ - pRoot = &pParse->aNode[iRoot]; - assert( pRoot->eU==0 ); - pRoot->u.iAppend = iStart; - pRoot->jnFlags |= JNODE_APPEND; - VVA( pRoot->eU = 2 ); + assert( !pParse->oom ); + jsonBlobEdit(pParse, j, 0, v.aBlob, v.nBlob); } - return pNode; + jsonParseReset(&v); + if( pParse->delta ) jsonAfterEditSizeAdjust(pParse, iRoot); + return rc; } }else{ - *pzErr = zPath; + return JSON_LOOKUP_PATHERROR; } - return 0; + return JSON_LOOKUP_NOTFOUND; } /* -** Append content to pParse that will complete zPath. Return a pointer -** to the inserted node, or return NULL if the append fails. +** Convert a JSON BLOB into text and make that text the return value +** of an SQL function. */ -static JsonNode *jsonLookupAppend( - JsonParse *pParse, /* Append content to the JSON parse */ - const char *zPath, /* Description of content to append */ - int *pApnd, /* Set this flag to 1 */ - const char **pzErr /* Make this point to any syntax error */ +static void jsonReturnTextJsonFromBlob( + sqlite3_context *ctx, + const u8 *aBlob, + u32 nBlob ){ - *pApnd = 1; - if( zPath[0]==0 ){ - jsonParseAddNode(pParse, JSON_NULL, 0, 0); - return pParse->oom ? 0 : &pParse->aNode[pParse->nNode-1]; - } - if( zPath[0]=='.' ){ - jsonParseAddNode(pParse, JSON_OBJECT, 0, 0); - }else if( strncmp(zPath,"[0]",3)==0 ){ - jsonParseAddNode(pParse, JSON_ARRAY, 0, 0); - }else{ - return 0; - } - if( pParse->oom ) return 0; - return jsonLookupStep(pParse, pParse->nNode-1, zPath, pApnd, pzErr); + JsonParse x; + JsonString s; + + if( NEVER(aBlob==0) ) return; + memset(&x, 0, sizeof(x)); + x.aBlob = (u8*)aBlob; + x.nBlob = nBlob; + jsonStringInit(&s, ctx); + jsonTranslateBlobToText(&x, 0, &s); + jsonReturnString(&s, 0, 0); } + /* -** Return the text of a syntax error message on a JSON path. Space is -** obtained from sqlite3_malloc(). +** Return the value of the BLOB node at index i. +** +** If the value is a primitive, return it as an SQL value. +** If the value is an array or object, return it as either +** JSON text or the BLOB encoding, depending on the JSON_B flag +** on the userdata. */ -static char *jsonPathSyntaxError(const char *zErr){ - return sqlite3_mprintf("JSON path error near '%q'", zErr); +static void jsonReturnFromBlob( + JsonParse *pParse, /* Complete JSON parse tree */ + u32 i, /* Index of the node */ + sqlite3_context *pCtx, /* Return value for this function */ + int textOnly /* return text JSON. Disregard user-data */ +){ + u32 n, sz; + int rc; + sqlite3 *db = sqlite3_context_db_handle(pCtx); + + n = jsonbPayloadSize(pParse, i, &sz); + if( n==0 ){ + sqlite3_result_error(pCtx, "malformed JSON", -1); + return; + } + switch( pParse->aBlob[i] & 0x0f ){ + case JSONB_NULL: { + if( sz ) goto returnfromblob_malformed; + sqlite3_result_null(pCtx); + break; + } + case JSONB_TRUE: { + if( sz ) goto returnfromblob_malformed; + sqlite3_result_int(pCtx, 1); + break; + } + case JSONB_FALSE: { + if( sz ) goto returnfromblob_malformed; + sqlite3_result_int(pCtx, 0); + break; + } + case JSONB_INT5: + case JSONB_INT: { + sqlite3_int64 iRes = 0; + char *z; + int bNeg = 0; + char x; + if( sz==0 ) goto returnfromblob_malformed; + x = (char)pParse->aBlob[i+n]; + if( x=='-' ){ + if( sz<2 ) goto returnfromblob_malformed; + n++; + sz--; + bNeg = 1; + } + z = sqlite3DbStrNDup(db, (const char*)&pParse->aBlob[i+n], (int)sz); + if( z==0 ) goto returnfromblob_oom; + rc = sqlite3DecOrHexToI64(z, &iRes); + sqlite3DbFree(db, z); + if( rc==0 ){ + sqlite3_result_int64(pCtx, bNeg ? -iRes : iRes); + }else if( rc==3 && bNeg ){ + sqlite3_result_int64(pCtx, SMALLEST_INT64); + }else if( rc==1 ){ + goto returnfromblob_malformed; + }else{ + if( bNeg ){ n--; sz++; } + goto to_double; + } + break; + } + case JSONB_FLOAT5: + case JSONB_FLOAT: { + double r; + char *z; + if( sz==0 ) goto returnfromblob_malformed; + to_double: + z = sqlite3DbStrNDup(db, (const char*)&pParse->aBlob[i+n], (int)sz); + if( z==0 ) goto returnfromblob_oom; + rc = sqlite3AtoF(z, &r, sqlite3Strlen30(z), SQLITE_UTF8); + sqlite3DbFree(db, z); + if( rc<=0 ) goto returnfromblob_malformed; + sqlite3_result_double(pCtx, r); + break; + } + case JSONB_TEXTRAW: + case JSONB_TEXT: { + sqlite3_result_text(pCtx, (char*)&pParse->aBlob[i+n], sz, + SQLITE_TRANSIENT); + break; + } + case JSONB_TEXT5: + case JSONB_TEXTJ: { + /* Translate JSON formatted string into raw text */ + u32 iIn, iOut; + const char *z; + char *zOut; + u32 nOut = sz; + z = (const char*)&pParse->aBlob[i+n]; + zOut = sqlite3DbMallocRaw(db, nOut+1); + if( zOut==0 ) goto returnfromblob_oom; + for(iIn=iOut=0; iIn=2 ); + zOut[iOut++] = (char)(0xc0 | (v>>6)); + zOut[iOut++] = 0x80 | (v&0x3f); + }else if( v<0x10000 ){ + assert( szEscape>=3 ); + zOut[iOut++] = 0xe0 | (v>>12); + zOut[iOut++] = 0x80 | ((v>>6)&0x3f); + zOut[iOut++] = 0x80 | (v&0x3f); + }else if( v==JSON_INVALID_CHAR ){ + /* Silently ignore illegal unicode */ + }else{ + assert( szEscape>=4 ); + zOut[iOut++] = 0xf0 | (v>>18); + zOut[iOut++] = 0x80 | ((v>>12)&0x3f); + zOut[iOut++] = 0x80 | ((v>>6)&0x3f); + zOut[iOut++] = 0x80 | (v&0x3f); + } + iIn += szEscape - 1; + }else{ + zOut[iOut++] = c; + } + } /* end for() */ + assert( iOut<=nOut ); + zOut[iOut] = 0; + sqlite3_result_text(pCtx, zOut, iOut, SQLITE_DYNAMIC); + break; + } + case JSONB_ARRAY: + case JSONB_OBJECT: { + int flags = textOnly ? 0 : SQLITE_PTR_TO_INT(sqlite3_user_data(pCtx)); + if( flags & JSON_BLOB ){ + sqlite3_result_blob(pCtx, &pParse->aBlob[i], sz+n, SQLITE_TRANSIENT); + }else{ + jsonReturnTextJsonFromBlob(pCtx, &pParse->aBlob[i], sz+n); + } + break; + } + default: { + goto returnfromblob_malformed; + } + } + return; + +returnfromblob_oom: + sqlite3_result_error_nomem(pCtx); + return; + +returnfromblob_malformed: + sqlite3_result_error(pCtx, "malformed JSON", -1); + return; } /* -** Do a node lookup using zPath. Return a pointer to the node on success. -** Return NULL if not found or if there is an error. +** pArg is a function argument that might be an SQL value or a JSON +** value. Figure out what it is and encode it as a JSONB blob. +** Return the results in pParse. ** -** On an error, write an error message into pCtx and increment the -** pParse->nErr counter. +** pParse is uninitialized upon entry. This routine will handle the +** initialization of pParse. The result will be contained in +** pParse->aBlob and pParse->nBlob. pParse->aBlob might be dynamically +** allocated (if pParse->nBlobAlloc is greater than zero) in which case +** the caller is responsible for freeing the space allocated to pParse->aBlob +** when it has finished with it. Or pParse->aBlob might be a static string +** or a value obtained from sqlite3_value_blob(pArg). ** -** If pApnd!=NULL then try to append missing nodes and set *pApnd = 1 if -** nodes are appended. +** If the argument is a BLOB that is clearly not a JSONB, then this +** function might set an error message in ctx and return non-zero. +** It might also set an error message and return non-zero on an OOM error. */ -static JsonNode *jsonLookup( - JsonParse *pParse, /* The JSON to search */ - const char *zPath, /* The path to search */ - int *pApnd, /* Append nodes to complete path if not NULL */ - sqlite3_context *pCtx /* Report errors here, if not NULL */ -){ - const char *zErr = 0; - JsonNode *pNode = 0; - char *zMsg; - - if( zPath==0 ) return 0; - if( zPath[0]!='$' ){ - zErr = zPath; - goto lookup_err; +static int jsonFunctionArgToBlob( + sqlite3_context *ctx, + sqlite3_value *pArg, + JsonParse *pParse +){ + int eType = sqlite3_value_type(pArg); + static u8 aNull[] = { 0x00 }; + memset(pParse, 0, sizeof(pParse[0])); + pParse->db = sqlite3_context_db_handle(ctx); + switch( eType ){ + default: { + pParse->aBlob = aNull; + pParse->nBlob = 1; + return 0; + } + case SQLITE_BLOB: { + if( jsonFuncArgMightBeBinary(pArg) ){ + pParse->aBlob = (u8*)sqlite3_value_blob(pArg); + pParse->nBlob = sqlite3_value_bytes(pArg); + }else{ + sqlite3_result_error(ctx, "JSON cannot hold BLOB values", -1); + return 1; + } + break; + } + case SQLITE_TEXT: { + const char *zJson = (const char*)sqlite3_value_text(pArg); + int nJson = sqlite3_value_bytes(pArg); + if( zJson==0 ) return 1; + if( sqlite3_value_subtype(pArg)==JSON_SUBTYPE ){ + pParse->zJson = (char*)zJson; + pParse->nJson = nJson; + if( jsonConvertTextToBlob(pParse, ctx) ){ + sqlite3_result_error(ctx, "malformed JSON", -1); + sqlite3DbFree(pParse->db, pParse->aBlob); + memset(pParse, 0, sizeof(pParse[0])); + return 1; + } + }else{ + jsonBlobAppendNode(pParse, JSONB_TEXTRAW, nJson, zJson); + } + break; + } + case SQLITE_FLOAT: { + double r = sqlite3_value_double(pArg); + if( NEVER(sqlite3IsNaN(r)) ){ + jsonBlobAppendNode(pParse, JSONB_NULL, 0, 0); + }else{ + int n = sqlite3_value_bytes(pArg); + const char *z = (const char*)sqlite3_value_text(pArg); + if( z==0 ) return 1; + if( z[0]=='I' ){ + jsonBlobAppendNode(pParse, JSONB_FLOAT, 5, "9e999"); + }else if( z[0]=='-' && z[1]=='I' ){ + jsonBlobAppendNode(pParse, JSONB_FLOAT, 6, "-9e999"); + }else{ + jsonBlobAppendNode(pParse, JSONB_FLOAT, n, z); + } + } + break; + } + case SQLITE_INTEGER: { + int n = sqlite3_value_bytes(pArg); + const char *z = (const char*)sqlite3_value_text(pArg); + if( z==0 ) return 1; + jsonBlobAppendNode(pParse, JSONB_INT, n, z); + break; + } } - zPath++; - pNode = jsonLookupStep(pParse, 0, zPath, pApnd, &zErr); - if( zErr==0 ) return pNode; + if( pParse->oom ){ + sqlite3_result_error_nomem(ctx); + return 1; + }else{ + return 0; + } +} -lookup_err: - pParse->nErr++; - assert( zErr!=0 && pCtx!=0 ); - zMsg = jsonPathSyntaxError(zErr); +/* +** Generate a bad path error. +** +** If ctx is not NULL then push the error message into ctx and return NULL. +** If ctx is NULL, then return the text of the error message. +*/ +static char *jsonBadPathError( + sqlite3_context *ctx, /* The function call containing the error */ + const char *zPath /* The path with the problem */ +){ + char *zMsg = sqlite3_mprintf("bad JSON path: %Q", zPath); + if( ctx==0 ) return zMsg; if( zMsg ){ - sqlite3_result_error(pCtx, zMsg, -1); + sqlite3_result_error(ctx, zMsg, -1); sqlite3_free(zMsg); }else{ - sqlite3_result_error_nomem(pCtx); + sqlite3_result_error_nomem(ctx); } return 0; } +/* argv[0] is a BLOB that seems likely to be a JSONB. Subsequent +** arguments come in parse where each pair contains a JSON path and +** content to insert or set at that patch. Do the updates +** and return the result. +** +** The specific operation is determined by eEdit, which can be one +** of JEDIT_INS, JEDIT_REPL, or JEDIT_SET. +*/ +static void jsonInsertIntoBlob( + sqlite3_context *ctx, + int argc, + sqlite3_value **argv, + int eEdit /* JEDIT_INS, JEDIT_REPL, or JEDIT_SET */ +){ + int i; + u32 rc = 0; + const char *zPath = 0; + int flgs; + JsonParse *p; + JsonParse ax; + + assert( (argc&1)==1 ); + flgs = argc==1 ? 0 : JSON_EDITABLE; + p = jsonParseFuncArg(ctx, argv[0], flgs); + if( p==0 ) return; + for(i=1; inBlob, ax.aBlob, ax.nBlob); + } + rc = 0; + }else{ + p->eEdit = eEdit; + p->nIns = ax.nBlob; + p->aIns = ax.aBlob; + p->delta = 0; + rc = jsonLookupStep(p, 0, zPath+1, 0); + } + jsonParseReset(&ax); + if( rc==JSON_LOOKUP_NOTFOUND ) continue; + if( JSON_LOOKUP_ISERROR(rc) ) goto jsonInsertIntoBlob_patherror; + } + jsonReturnParse(ctx, p); + jsonParseFree(p); + return; + +jsonInsertIntoBlob_patherror: + jsonParseFree(p); + if( rc==JSON_LOOKUP_ERROR ){ + sqlite3_result_error(ctx, "malformed JSON", -1); + }else{ + jsonBadPathError(ctx, zPath); + } + return; +} /* -** Report the wrong number of arguments for json_insert(), json_replace() -** or json_set(). +** Generate a JsonParse object, containing valid JSONB in aBlob and nBlob, +** from the SQL function argument pArg. Return a pointer to the new +** JsonParse object. +** +** Ownership of the new JsonParse object is passed to the caller. The +** caller should invoke jsonParseFree() on the return value when it +** has finished using it. +** +** If any errors are detected, an appropriate error messages is set +** using sqlite3_result_error() or the equivalent and this routine +** returns NULL. This routine also returns NULL if the pArg argument +** is an SQL NULL value, but no error message is set in that case. This +** is so that SQL functions that are given NULL arguments will return +** a NULL value. */ -static void jsonWrongNumArgs( - sqlite3_context *pCtx, - const char *zFuncName +static JsonParse *jsonParseFuncArg( + sqlite3_context *ctx, + sqlite3_value *pArg, + u32 flgs ){ - char *zMsg = sqlite3_mprintf("json_%s() needs an odd number of arguments", - zFuncName); - sqlite3_result_error(pCtx, zMsg, -1); - sqlite3_free(zMsg); + int eType; /* Datatype of pArg */ + JsonParse *p = 0; /* Value to be returned */ + JsonParse *pFromCache = 0; /* Value taken from cache */ + sqlite3 *db; /* The database connection */ + + assert( ctx!=0 ); + eType = sqlite3_value_type(pArg); + if( eType==SQLITE_NULL ){ + return 0; + } + pFromCache = jsonCacheSearch(ctx, pArg); + if( pFromCache ){ + pFromCache->nJPRef++; + if( (flgs & JSON_EDITABLE)==0 ){ + return pFromCache; + } + } + db = sqlite3_context_db_handle(ctx); +rebuild_from_cache: + p = sqlite3DbMallocZero(db, sizeof(*p)); + if( p==0 ) goto json_pfa_oom; + memset(p, 0, sizeof(*p)); + p->db = db; + p->nJPRef = 1; + if( pFromCache!=0 ){ + u32 nBlob = pFromCache->nBlob; + p->aBlob = sqlite3DbMallocRaw(db, nBlob); + if( p->aBlob==0 ) goto json_pfa_oom; + memcpy(p->aBlob, pFromCache->aBlob, nBlob); + p->nBlobAlloc = p->nBlob = nBlob; + p->hasNonstd = pFromCache->hasNonstd; + jsonParseFree(pFromCache); + return p; + } + if( eType==SQLITE_BLOB ){ + u32 n, sz = 0; + p->aBlob = (u8*)sqlite3_value_blob(pArg); + p->nBlob = (u32)sqlite3_value_bytes(pArg); + if( p->nBlob==0 ){ + goto json_pfa_malformed; + } + if( NEVER(p->aBlob==0) ){ + goto json_pfa_oom; + } + if( (p->aBlob[0] & 0x0f)>JSONB_OBJECT ){ + goto json_pfa_malformed; + } + n = jsonbPayloadSize(p, 0, &sz); + if( n==0 + || sz+n!=p->nBlob + || ((p->aBlob[0] & 0x0f)<=JSONB_FALSE && sz>0) + ){ + goto json_pfa_malformed; + } + if( (flgs & JSON_EDITABLE)!=0 && jsonBlobMakeEditable(p, 0)==0 ){ + goto json_pfa_oom; + } + return p; + } + p->zJson = (char*)sqlite3_value_text(pArg); + p->nJson = sqlite3_value_bytes(pArg); + if( p->nJson==0 ) goto json_pfa_malformed; + if( NEVER(p->zJson==0) ) goto json_pfa_oom; + if( jsonConvertTextToBlob(p, (flgs & JSON_KEEPERROR) ? 0 : ctx) ){ + if( flgs & JSON_KEEPERROR ){ + p->nErr = 1; + return p; + }else{ + jsonParseFree(p); + return 0; + } + }else{ + int isRCStr = sqlite3ValueIsOfClass(pArg, sqlite3RCStrUnref); + int rc; + if( !isRCStr ){ + char *zNew = sqlite3RCStrNew( p->nJson ); + if( zNew==0 ) goto json_pfa_oom; + memcpy(zNew, p->zJson, p->nJson); + p->zJson = zNew; + p->zJson[p->nJson] = 0; + }else{ + sqlite3RCStrRef(p->zJson); + } + p->bJsonIsRCStr = 1; + rc = jsonCacheInsert(ctx, p); + if( rc==SQLITE_NOMEM ) goto json_pfa_oom; + if( flgs & JSON_EDITABLE ){ + pFromCache = p; + p = 0; + goto rebuild_from_cache; + } + } + return p; + +json_pfa_malformed: + if( flgs & JSON_KEEPERROR ){ + p->nErr = 1; + return p; + }else{ + jsonParseFree(p); + sqlite3_result_error(ctx, "malformed JSON", -1); + return 0; + } + +json_pfa_oom: + jsonParseFree(pFromCache); + jsonParseFree(p); + sqlite3_result_error_nomem(ctx); + return 0; } /* -** Mark all NULL entries in the Object passed in as JNODE_REMOVE. +** Make the return value of a JSON function either the raw JSONB blob +** or make it JSON text, depending on whether the JSON_BLOB flag is +** set on the function. */ -static void jsonRemoveAllNulls(JsonNode *pNode){ - int i, n; - assert( pNode->eType==JSON_OBJECT ); - n = pNode->n; - for(i=2; i<=n; i += jsonNodeSize(&pNode[i])+1){ - switch( pNode[i].eType ){ - case JSON_NULL: - pNode[i].jnFlags |= JNODE_REMOVE; - break; - case JSON_OBJECT: - jsonRemoveAllNulls(&pNode[i]); - break; +static void jsonReturnParse( + sqlite3_context *ctx, + JsonParse *p +){ + int flgs; + if( p->oom ){ + sqlite3_result_error_nomem(ctx); + return; + } + flgs = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + if( flgs & JSON_BLOB ){ + if( p->nBlobAlloc>0 && !p->bReadOnly ){ + sqlite3_result_blob(ctx, p->aBlob, p->nBlob, SQLITE_DYNAMIC); + p->nBlobAlloc = 0; + }else{ + sqlite3_result_blob(ctx, p->aBlob, p->nBlob, SQLITE_TRANSIENT); } + }else{ + JsonString s; + jsonStringInit(&s, ctx); + p->delta = 0; + jsonTranslateBlobToText(p, 0, &s); + jsonReturnString(&s, p, ctx); + sqlite3_result_subtype(ctx, JSON_SUBTYPE); } } - /**************************************************************************** ** SQL functions used for testing and debugging ****************************************************************************/ #if SQLITE_DEBUG /* -** Print N node entries. -*/ -static void jsonDebugPrintNodeEntries( - JsonNode *aNode, /* First node entry to print */ - int N /* Number of node entries to print */ -){ - int i; - for(i=0; iaBlob[iStart] & 0x0f; + u32 savedNBlob = pParse->nBlob; + sqlite3_str_appendf(pOut, "%5d:%*s", iStart, nIndent, ""); + if( pParse->nBlobAlloc>pParse->nBlob ){ + pParse->nBlob = pParse->nBlobAlloc; + } + nn = n = jsonbPayloadSize(pParse, iStart, &sz); + if( nn==0 ) nn = 1; + if( sz>0 && xaBlob[iStart+i]); } - printf("node %4u: %-7s n=%-5d", i, zType, aNode[i].n); - if( (aNode[i].jnFlags & ~JNODE_LABEL)!=0 ){ - u8 f = aNode[i].jnFlags; - if( f & JNODE_RAW ) printf(" RAW"); - if( f & JNODE_ESCAPE ) printf(" ESCAPE"); - if( f & JNODE_REMOVE ) printf(" REMOVE"); - if( f & JNODE_REPLACE ) printf(" REPLACE"); - if( f & JNODE_APPEND ) printf(" APPEND"); - if( f & JNODE_JSON5 ) printf(" JSON5"); + if( n==0 ){ + sqlite3_str_appendf(pOut, " ERROR invalid node size\n"); + iStart = n==0 ? iStart+1 : iEnd; + continue; } - switch( aNode[i].eU ){ - case 1: printf(" zJContent=[%.*s]\n", - aNode[i].n, aNode[i].u.zJContent); break; - case 2: printf(" iAppend=%u\n", aNode[i].u.iAppend); break; - case 3: printf(" iKey=%u\n", aNode[i].u.iKey); break; - case 4: printf(" iPrev=%u\n", aNode[i].u.iPrev); break; - default: printf("\n"); + pParse->nBlob = savedNBlob; + if( iStart+n+sz>iEnd ){ + iEnd = iStart+n+sz; + if( iEnd>pParse->nBlob ){ + if( pParse->nBlobAlloc>0 && iEnd>pParse->nBlobAlloc ){ + iEnd = pParse->nBlobAlloc; + }else{ + iEnd = pParse->nBlob; + } + } } + sqlite3_str_appendall(pOut," <-- "); + switch( x ){ + case JSONB_NULL: sqlite3_str_appendall(pOut,"null"); break; + case JSONB_TRUE: sqlite3_str_appendall(pOut,"true"); break; + case JSONB_FALSE: sqlite3_str_appendall(pOut,"false"); break; + case JSONB_INT: sqlite3_str_appendall(pOut,"int"); break; + case JSONB_INT5: sqlite3_str_appendall(pOut,"int5"); break; + case JSONB_FLOAT: sqlite3_str_appendall(pOut,"float"); break; + case JSONB_FLOAT5: sqlite3_str_appendall(pOut,"float5"); break; + case JSONB_TEXT: sqlite3_str_appendall(pOut,"text"); break; + case JSONB_TEXTJ: sqlite3_str_appendall(pOut,"textj"); break; + case JSONB_TEXT5: sqlite3_str_appendall(pOut,"text5"); break; + case JSONB_TEXTRAW: sqlite3_str_appendall(pOut,"textraw"); break; + case JSONB_ARRAY: { + sqlite3_str_appendf(pOut,"array, %u bytes\n", sz); + jsonDebugPrintBlob(pParse, iStart+n, iStart+n+sz, nIndent+2, pOut); + showContent = 0; + break; + } + case JSONB_OBJECT: { + sqlite3_str_appendf(pOut, "object, %u bytes\n", sz); + jsonDebugPrintBlob(pParse, iStart+n, iStart+n+sz, nIndent+2, pOut); + showContent = 0; + break; + } + default: { + sqlite3_str_appendall(pOut, "ERROR: unknown node type\n"); + showContent = 0; + break; + } + } + if( showContent ){ + if( sz==0 && x<=JSONB_FALSE ){ + sqlite3_str_append(pOut, "\n", 1); + }else{ + u32 i; + sqlite3_str_appendall(pOut, ": \""); + for(i=iStart+n; iaBlob[i]; + if( c<0x20 || c>=0x7f ) c = '.'; + sqlite3_str_append(pOut, (char*)&c, 1); + } + sqlite3_str_append(pOut, "\"\n", 2); + } + } + iStart += n + sz; } } -#endif /* SQLITE_DEBUG */ - - -#if 0 /* 1 for debugging. 0 normally. Requires -DSQLITE_DEBUG too */ -static void jsonDebugPrintParse(JsonParse *p){ - jsonDebugPrintNodeEntries(p->aNode, p->nNode); -} -static void jsonDebugPrintNode(JsonNode *pNode){ - jsonDebugPrintNodeEntries(pNode, jsonNodeSize(pNode)); +static void jsonShowParse(JsonParse *pParse){ + sqlite3_str out; + char zBuf[1000]; + if( pParse==0 ){ + printf("NULL pointer\n"); + return; + }else{ + printf("nBlobAlloc = %u\n", pParse->nBlobAlloc); + printf("nBlob = %u\n", pParse->nBlob); + printf("delta = %d\n", pParse->delta); + if( pParse->nBlob==0 ) return; + printf("content (bytes 0..%u):\n", pParse->nBlob-1); + } + sqlite3StrAccumInit(&out, 0, zBuf, sizeof(zBuf), 1000000); + jsonDebugPrintBlob(pParse, 0, pParse->nBlob, 0, &out); + printf("%s", sqlite3_str_value(&out)); + sqlite3_str_reset(&out); } -#else - /* The usual case */ -# define jsonDebugPrintNode(X) -# define jsonDebugPrintParse(X) -#endif +#endif /* SQLITE_DEBUG */ #ifdef SQLITE_DEBUG /* ** SQL function: json_parse(JSON) ** -** Parse JSON using jsonParseCached(). Then print a dump of that -** parse on standard output. Return the mimified JSON result, just -** like the json() function. +** Parse JSON using jsonParseFuncArg(). Return text that is a +** human-readable dump of the binary JSONB for the input parameter. */ static void jsonParseFunc( sqlite3_context *ctx, @@ -204934,38 +206671,19 @@ static void jsonParseFunc( sqlite3_value **argv ){ JsonParse *p; /* The parse */ + sqlite3_str out; - assert( argc==1 ); - p = jsonParseCached(ctx, argv[0], ctx, 0); + assert( argc>=1 ); + sqlite3StrAccumInit(&out, 0, 0, 0, 1000000); + p = jsonParseFuncArg(ctx, argv[0], 0); if( p==0 ) return; - printf("nNode = %u\n", p->nNode); - printf("nAlloc = %u\n", p->nAlloc); - printf("nJson = %d\n", p->nJson); - printf("nAlt = %d\n", p->nAlt); - printf("nErr = %u\n", p->nErr); - printf("oom = %u\n", p->oom); - printf("hasNonstd = %u\n", p->hasNonstd); - printf("useMod = %u\n", p->useMod); - printf("hasMod = %u\n", p->hasMod); - printf("nJPRef = %u\n", p->nJPRef); - printf("iSubst = %u\n", p->iSubst); - printf("iHold = %u\n", p->iHold); - jsonDebugPrintNodeEntries(p->aNode, p->nNode); - jsonReturnJson(p, p->aNode, ctx, 1); -} - -/* -** The json_test1(JSON) function return true (1) if the input is JSON -** text generated by another json function. It returns (0) if the input -** is not known to be JSON. -*/ -static void jsonTest1Func( - sqlite3_context *ctx, - int argc, - sqlite3_value **argv -){ - UNUSED_PARAMETER(argc); - sqlite3_result_int(ctx, sqlite3_value_subtype(argv[0])==JSON_SUBTYPE); + if( argc==1 ){ + jsonDebugPrintBlob(p, 0, p->nBlob, 0, &out); + sqlite3_result_text64(ctx, out.zText, out.nChar, SQLITE_DYNAMIC, SQLITE_UTF8); + }else{ + jsonShowParse(p); + } + jsonParseFree(p); } #endif /* SQLITE_DEBUG */ @@ -204974,7 +206692,7 @@ static void jsonTest1Func( ****************************************************************************/ /* -** Implementation of the json_QUOTE(VALUE) function. Return a JSON value +** Implementation of the json_quote(VALUE) function. Return a JSON value ** corresponding to the SQL value input. Mostly this means putting ** double-quotes around strings and returning the unquoted string "null" ** when given a NULL input. @@ -204987,9 +206705,9 @@ static void jsonQuoteFunc( JsonString jx; UNUSED_PARAMETER(argc); - jsonInit(&jx, ctx); - jsonAppendValue(&jx, argv[0]); - jsonResult(&jx); + jsonStringInit(&jx, ctx); + jsonAppendSqlValue(&jx, argv[0]); + jsonReturnString(&jx, 0, 0); sqlite3_result_subtype(ctx, JSON_SUBTYPE); } @@ -205006,18 +206724,17 @@ static void jsonArrayFunc( int i; JsonString jx; - jsonInit(&jx, ctx); + jsonStringInit(&jx, ctx); jsonAppendChar(&jx, '['); for(i=0; inNode ); if( argc==2 ){ const char *zPath = (const char*)sqlite3_value_text(argv[1]); - pNode = jsonLookup(p, zPath, 0, ctx); - }else{ - pNode = p->aNode; - } - if( pNode==0 ){ - return; - } - if( pNode->eType==JSON_ARRAY ){ - while( 1 /*exit-by-break*/ ){ - i = 1; - while( i<=pNode->n ){ - if( (pNode[i].jnFlags & JNODE_REMOVE)==0 ) n++; - i += jsonNodeSize(&pNode[i]); + if( zPath==0 ){ + jsonParseFree(p); + return; + } + i = jsonLookupStep(p, 0, zPath[0]=='$' ? zPath+1 : "@", 0); + if( JSON_LOOKUP_ISERROR(i) ){ + if( i==JSON_LOOKUP_NOTFOUND ){ + /* no-op */ + }else if( i==JSON_LOOKUP_PATHERROR ){ + jsonBadPathError(ctx, zPath); + }else{ + sqlite3_result_error(ctx, "malformed JSON", -1); } - if( (pNode->jnFlags & JNODE_APPEND)==0 ) break; - if( p->useMod==0 ) break; - assert( pNode->eU==2 ); - pNode = &p->aNode[pNode->u.iAppend]; + eErr = 1; + i = 0; } + }else{ + i = 0; + } + if( (p->aBlob[i] & 0x0f)==JSONB_ARRAY ){ + cnt = jsonbArrayCount(p, i); } - sqlite3_result_int64(ctx, n); + if( !eErr ) sqlite3_result_int64(ctx, cnt); + jsonParseFree(p); } -/* -** Bit values for the flags passed into jsonExtractFunc() or -** jsonSetFunc() via the user-data value. -*/ -#define JSON_JSON 0x01 /* Result is always JSON */ -#define JSON_SQL 0x02 /* Result is always SQL */ -#define JSON_ABPATH 0x03 /* Allow abbreviated JSON path specs */ -#define JSON_ISSET 0x04 /* json_set(), not json_insert() */ +/* True if the string is all digits */ +static int jsonAllDigits(const char *z, int n){ + int i; + for(i=0; i2 ){ + jsonAppendChar(&jx, '['); + } + for(i=1; i and ->> operators accept abbreviated PATH arguments. This - ** is mostly for compatibility with PostgreSQL, but also for - ** convenience. - ** - ** NUMBER ==> $[NUMBER] // PG compatible - ** LABEL ==> $.LABEL // PG compatible - ** [NUMBER] ==> $[NUMBER] // Not PG. Purely for convenience - */ - jsonInit(&jx, ctx); - if( sqlite3Isdigit(zPath[0]) ){ - jsonAppendRawNZ(&jx, "$[", 2); - jsonAppendRaw(&jx, zPath, (int)strlen(zPath)); - jsonAppendRawNZ(&jx, "]", 2); - }else{ - jsonAppendRawNZ(&jx, "$.", 1 + (zPath[0]!='[')); - jsonAppendRaw(&jx, zPath, (int)strlen(zPath)); - jsonAppendChar(&jx, 0); - } - pNode = jx.bErr ? 0 : jsonLookup(p, jx.zBuf, 0, ctx); - jsonReset(&jx); + const char *zPath = (const char*)sqlite3_value_text(argv[i]); + int nPath; + u32 j; + if( zPath==0 ) goto json_extract_error; + nPath = sqlite3Strlen30(zPath); + if( zPath[0]=='$' ){ + j = jsonLookupStep(p, 0, zPath+1, 0); + }else if( (flags & JSON_ABPATH) ){ + /* The -> and ->> operators accept abbreviated PATH arguments. This + ** is mostly for compatibility with PostgreSQL, but also for + ** convenience. + ** + ** NUMBER ==> $[NUMBER] // PG compatible + ** LABEL ==> $.LABEL // PG compatible + ** [NUMBER] ==> $[NUMBER] // Not PG. Purely for convenience + */ + jsonStringInit(&jx, ctx); + if( jsonAllDigits(zPath, nPath) ){ + jsonAppendRawNZ(&jx, "[", 1); + jsonAppendRaw(&jx, zPath, nPath); + jsonAppendRawNZ(&jx, "]", 2); + }else if( jsonAllAlphanum(zPath, nPath) ){ + jsonAppendRawNZ(&jx, ".", 1); + jsonAppendRaw(&jx, zPath, nPath); + }else if( zPath[0]=='[' && nPath>=3 && zPath[nPath-1]==']' ){ + jsonAppendRaw(&jx, zPath, nPath); }else{ - pNode = jsonLookup(p, zPath, 0, ctx); + jsonAppendRawNZ(&jx, ".\"", 2); + jsonAppendRaw(&jx, zPath, nPath); + jsonAppendRawNZ(&jx, "\"", 1); } - if( pNode ){ + jsonStringTerminate(&jx); + j = jsonLookupStep(p, 0, jx.zBuf, 0); + jsonStringReset(&jx); + }else{ + jsonBadPathError(ctx, zPath); + goto json_extract_error; + } + if( jnBlob ){ + if( argc==2 ){ if( flags & JSON_JSON ){ - jsonReturnJson(p, pNode, ctx, 0); + jsonStringInit(&jx, ctx); + jsonTranslateBlobToText(p, j, &jx); + jsonReturnString(&jx, 0, 0); + jsonStringReset(&jx); + assert( (flags & JSON_BLOB)==0 ); + sqlite3_result_subtype(ctx, JSON_SUBTYPE); }else{ - jsonReturn(p, pNode, ctx); - sqlite3_result_subtype(ctx, 0); + jsonReturnFromBlob(p, j, ctx, 0); + if( (flags & (JSON_SQL|JSON_BLOB))==0 + && (p->aBlob[j]&0x0f)>=JSONB_ARRAY + ){ + sqlite3_result_subtype(ctx, JSON_SUBTYPE); + } } + }else{ + jsonAppendSeparator(&jx); + jsonTranslateBlobToText(p, j, &jx); } - }else{ - pNode = jsonLookup(p, zPath, 0, ctx); - if( p->nErr==0 && pNode ) jsonReturn(p, pNode, ctx); - } - }else{ - /* Two or more PATH arguments results in a JSON array with each - ** element of the array being the value selected by one of the PATHs */ - int i; - jsonInit(&jx, ctx); - jsonAppendChar(&jx, '['); - for(i=1; inErr ) break; - jsonAppendSeparator(&jx); - if( pNode ){ - jsonRenderNode(p, pNode, &jx); + }else if( j==JSON_LOOKUP_NOTFOUND ){ + if( argc==2 ){ + goto json_extract_error; /* Return NULL if not found */ }else{ + jsonAppendSeparator(&jx); jsonAppendRawNZ(&jx, "null", 4); } + }else if( j==JSON_LOOKUP_ERROR ){ + sqlite3_result_error(ctx, "malformed JSON", -1); + goto json_extract_error; + }else{ + jsonBadPathError(ctx, zPath); + goto json_extract_error; } - if( i==argc ){ - jsonAppendChar(&jx, ']'); - jsonResult(&jx); + } + if( argc>2 ){ + jsonAppendChar(&jx, ']'); + jsonReturnString(&jx, 0, 0); + if( (flags & JSON_BLOB)==0 ){ sqlite3_result_subtype(ctx, JSON_SUBTYPE); } - jsonReset(&jx); } +json_extract_error: + jsonStringReset(&jx); + jsonParseFree(p); + return; } -/* This is the RFC 7396 MergePatch algorithm. -*/ -static JsonNode *jsonMergePatch( - JsonParse *pParse, /* The JSON parser that contains the TARGET */ - u32 iTarget, /* Node of the TARGET in pParse */ - JsonNode *pPatch /* The PATCH */ -){ - u32 i, j; - u32 iRoot; - JsonNode *pTarget; - if( pPatch->eType!=JSON_OBJECT ){ - return pPatch; - } - assert( iTargetnNode ); - pTarget = &pParse->aNode[iTarget]; - assert( (pPatch->jnFlags & JNODE_APPEND)==0 ); - if( pTarget->eType!=JSON_OBJECT ){ - jsonRemoveAllNulls(pPatch); - return pPatch; - } - iRoot = iTarget; - for(i=1; in; i += jsonNodeSize(&pPatch[i+1])+1){ - u32 nKey; - const char *zKey; - assert( pPatch[i].eType==JSON_STRING ); - assert( pPatch[i].jnFlags & JNODE_LABEL ); - assert( pPatch[i].eU==1 ); - nKey = pPatch[i].n; - zKey = pPatch[i].u.zJContent; - for(j=1; jn; j += jsonNodeSize(&pTarget[j+1])+1 ){ - assert( pTarget[j].eType==JSON_STRING ); - assert( pTarget[j].jnFlags & JNODE_LABEL ); - if( jsonSameLabel(&pPatch[i], &pTarget[j]) ){ - if( pTarget[j+1].jnFlags & (JNODE_REMOVE|JNODE_REPLACE) ) break; - if( pPatch[i+1].eType==JSON_NULL ){ - pTarget[j+1].jnFlags |= JNODE_REMOVE; - }else{ - JsonNode *pNew = jsonMergePatch(pParse, iTarget+j+1, &pPatch[i+1]); - if( pNew==0 ) return 0; - if( pNew!=&pParse->aNode[iTarget+j+1] ){ - jsonParseAddSubstNode(pParse, iTarget+j+1); - jsonParseAddNodeArray(pParse, pNew, jsonNodeSize(pNew)); - } - pTarget = &pParse->aNode[iTarget]; - } - break; +/* +** Return codes for jsonMergePatch() +*/ +#define JSON_MERGE_OK 0 /* Success */ +#define JSON_MERGE_BADTARGET 1 /* Malformed TARGET blob */ +#define JSON_MERGE_BADPATCH 2 /* Malformed PATCH blob */ +#define JSON_MERGE_OOM 3 /* Out-of-memory condition */ + +/* +** RFC-7396 MergePatch for two JSONB blobs. +** +** pTarget is the target. pPatch is the patch. The target is updated +** in place. The patch is read-only. +** +** The original RFC-7396 algorithm is this: +** +** define MergePatch(Target, Patch): +** if Patch is an Object: +** if Target is not an Object: +** Target = {} # Ignore the contents and set it to an empty Object +** for each Name/Value pair in Patch: +** if Value is null: +** if Name exists in Target: +** remove the Name/Value pair from Target +** else: +** Target[Name] = MergePatch(Target[Name], Value) +** return Target +** else: +** return Patch +** +** Here is an equivalent algorithm restructured to show the actual +** implementation: +** +** 01 define MergePatch(Target, Patch): +** 02 if Patch is not an Object: +** 03 return Patch +** 04 else: // if Patch is an Object +** 05 if Target is not an Object: +** 06 Target = {} +** 07 for each Name/Value pair in Patch: +** 08 if Name exists in Target: +** 09 if Value is null: +** 10 remove the Name/Value pair from Target +** 11 else +** 12 Target[name] = MergePatch(Target[Name], Value) +** 13 else if Value is not NULL: +** 14 if Value is not an Object: +** 15 Target[name] = Value +** 16 else: +** 17 Target[name] = MergePatch('{}',value) +** 18 return Target +** | +** ^---- Line numbers referenced in comments in the implementation +*/ +static int jsonMergePatch( + JsonParse *pTarget, /* The JSON parser that contains the TARGET */ + u32 iTarget, /* Index of TARGET in pTarget->aBlob[] */ + const JsonParse *pPatch, /* The PATCH */ + u32 iPatch /* Index of PATCH in pPatch->aBlob[] */ +){ + u8 x; /* Type of a single node */ + u32 n, sz=0; /* Return values from jsonbPayloadSize() */ + u32 iTCursor; /* Cursor position while scanning the target object */ + u32 iTStart; /* First label in the target object */ + u32 iTEndBE; /* Original first byte past end of target, before edit */ + u32 iTEnd; /* Current first byte past end of target */ + u8 eTLabel; /* Node type of the target label */ + u32 iTLabel = 0; /* Index of the label */ + u32 nTLabel = 0; /* Header size in bytes for the target label */ + u32 szTLabel = 0; /* Size of the target label payload */ + u32 iTValue = 0; /* Index of the target value */ + u32 nTValue = 0; /* Header size of the target value */ + u32 szTValue = 0; /* Payload size for the target value */ + + u32 iPCursor; /* Cursor position while scanning the patch */ + u32 iPEnd; /* First byte past the end of the patch */ + u8 ePLabel; /* Node type of the patch label */ + u32 iPLabel; /* Start of patch label */ + u32 nPLabel; /* Size of header on the patch label */ + u32 szPLabel; /* Payload size of the patch label */ + u32 iPValue; /* Start of patch value */ + u32 nPValue; /* Header size for the patch value */ + u32 szPValue; /* Payload size of the patch value */ + + assert( iTarget>=0 && iTargetnBlob ); + assert( iPatch>=0 && iPatchnBlob ); + x = pPatch->aBlob[iPatch] & 0x0f; + if( x!=JSONB_OBJECT ){ /* Algorithm line 02 */ + u32 szPatch; /* Total size of the patch, header+payload */ + u32 szTarget; /* Total size of the target, header+payload */ + n = jsonbPayloadSize(pPatch, iPatch, &sz); + szPatch = n+sz; + sz = 0; + n = jsonbPayloadSize(pTarget, iTarget, &sz); + szTarget = n+sz; + jsonBlobEdit(pTarget, iTarget, szTarget, pPatch->aBlob+iPatch, szPatch); + return pTarget->oom ? JSON_MERGE_OOM : JSON_MERGE_OK; /* Line 03 */ + } + x = pTarget->aBlob[iTarget] & 0x0f; + if( x!=JSONB_OBJECT ){ /* Algorithm line 05 */ + n = jsonbPayloadSize(pTarget, iTarget, &sz); + jsonBlobEdit(pTarget, iTarget+n, sz, 0, 0); + x = pTarget->aBlob[iTarget]; + pTarget->aBlob[iTarget] = (x & 0xf0) | JSONB_OBJECT; + } + n = jsonbPayloadSize(pPatch, iPatch, &sz); + if( NEVER(n==0) ) return JSON_MERGE_BADPATCH; + iPCursor = iPatch+n; + iPEnd = iPCursor+sz; + n = jsonbPayloadSize(pTarget, iTarget, &sz); + if( NEVER(n==0) ) return JSON_MERGE_BADTARGET; + iTStart = iTarget+n; + iTEndBE = iTStart+sz; + + while( iPCursoraBlob[iPCursor] & 0x0f; + if( ePLabelJSONB_TEXTRAW ){ + return JSON_MERGE_BADPATCH; + } + nPLabel = jsonbPayloadSize(pPatch, iPCursor, &szPLabel); + if( nPLabel==0 ) return JSON_MERGE_BADPATCH; + iPValue = iPCursor + nPLabel + szPLabel; + if( iPValue>=iPEnd ) return JSON_MERGE_BADPATCH; + nPValue = jsonbPayloadSize(pPatch, iPValue, &szPValue); + if( nPValue==0 ) return JSON_MERGE_BADPATCH; + iPCursor = iPValue + nPValue + szPValue; + if( iPCursor>iPEnd ) return JSON_MERGE_BADPATCH; + + iTCursor = iTStart; + iTEnd = iTEndBE + pTarget->delta; + while( iTCursoraBlob[iTCursor] & 0x0f; + if( eTLabelJSONB_TEXTRAW ){ + return JSON_MERGE_BADTARGET; + } + nTLabel = jsonbPayloadSize(pTarget, iTCursor, &szTLabel); + if( nTLabel==0 ) return JSON_MERGE_BADTARGET; + iTValue = iTLabel + nTLabel + szTLabel; + if( iTValue>=iTEnd ) return JSON_MERGE_BADTARGET; + nTValue = jsonbPayloadSize(pTarget, iTValue, &szTValue); + if( nTValue==0 ) return JSON_MERGE_BADTARGET; + if( iTValue + nTValue + szTValue > iTEnd ) return JSON_MERGE_BADTARGET; + isEqual = jsonLabelCompare( + (const char*)&pPatch->aBlob[iPLabel+nPLabel], + szPLabel, + (ePLabel==JSONB_TEXT || ePLabel==JSONB_TEXTRAW), + (const char*)&pTarget->aBlob[iTLabel+nTLabel], + szTLabel, + (eTLabel==JSONB_TEXT || eTLabel==JSONB_TEXTRAW)); + if( isEqual ) break; + iTCursor = iTValue + nTValue + szTValue; + } + x = pPatch->aBlob[iPValue] & 0x0f; + if( iTCursoroom) ) return JSON_MERGE_OOM; + }else{ + /* Algorithm line 12 */ + int rc, savedDelta = pTarget->delta; + pTarget->delta = 0; + rc = jsonMergePatch(pTarget, iTValue, pPatch, iPValue); + if( rc ) return rc; + pTarget->delta += savedDelta; + } + }else if( x>0 ){ /* Algorithm line 13 */ + /* No match and patch value is not NULL */ + u32 szNew = szPLabel+nPLabel; + if( (pPatch->aBlob[iPValue] & 0x0f)!=JSONB_OBJECT ){ /* Line 14 */ + jsonBlobEdit(pTarget, iTEnd, 0, 0, szPValue+nPValue+szNew); + if( pTarget->oom ) return JSON_MERGE_OOM; + memcpy(&pTarget->aBlob[iTEnd], &pPatch->aBlob[iPLabel], szNew); + memcpy(&pTarget->aBlob[iTEnd+szNew], + &pPatch->aBlob[iPValue], szPValue+nPValue); + }else{ + int rc, savedDelta; + jsonBlobEdit(pTarget, iTEnd, 0, 0, szNew+1); + if( pTarget->oom ) return JSON_MERGE_OOM; + memcpy(&pTarget->aBlob[iTEnd], &pPatch->aBlob[iPLabel], szNew); + pTarget->aBlob[iTEnd+szNew] = 0x00; + savedDelta = pTarget->delta; + pTarget->delta = 0; + rc = jsonMergePatch(pTarget, iTEnd+szNew,pPatch,iPValue); + if( rc ) return rc; + pTarget->delta += savedDelta; } } - if( j>=pTarget->n && pPatch[i+1].eType!=JSON_NULL ){ - int iStart; - JsonNode *pApnd; - u32 nApnd; - iStart = jsonParseAddNode(pParse, JSON_OBJECT, 0, 0); - jsonParseAddNode(pParse, JSON_STRING, nKey, zKey); - pApnd = &pPatch[i+1]; - if( pApnd->eType==JSON_OBJECT ) jsonRemoveAllNulls(pApnd); - nApnd = jsonNodeSize(pApnd); - jsonParseAddNodeArray(pParse, pApnd, jsonNodeSize(pApnd)); - if( pParse->oom ) return 0; - pParse->aNode[iStart].n = 1+nApnd; - pParse->aNode[iRoot].jnFlags |= JNODE_APPEND; - pParse->aNode[iRoot].u.iAppend = iStart; - VVA( pParse->aNode[iRoot].eU = 2 ); - iRoot = iStart; - pTarget = &pParse->aNode[iTarget]; - } } - return pTarget; + if( pTarget->delta ) jsonAfterEditSizeAdjust(pTarget, iTarget); + return pTarget->oom ? JSON_MERGE_OOM : JSON_MERGE_OK; } + /* ** Implementation of the json_mergepatch(JSON1,JSON2) function. Return a JSON ** object that is the result of running the RFC 7396 MergePatch() algorithm @@ -205253,28 +207132,27 @@ static void jsonPatchFunc( int argc, sqlite3_value **argv ){ - JsonParse *pX; /* The JSON that is being patched */ - JsonParse *pY; /* The patch */ - JsonNode *pResult; /* The result of the merge */ + JsonParse *pTarget; /* The TARGET */ + JsonParse *pPatch; /* The PATCH */ + int rc; /* Result code */ UNUSED_PARAMETER(argc); - pX = jsonParseCached(ctx, argv[0], ctx, 1); - if( pX==0 ) return; - assert( pX->hasMod==0 ); - pX->hasMod = 1; - pY = jsonParseCached(ctx, argv[1], ctx, 1); - if( pY==0 ) return; - pX->useMod = 1; - pY->useMod = 1; - pResult = jsonMergePatch(pX, 0, pY->aNode); - assert( pResult!=0 || pX->oom ); - if( pResult && pX->oom==0 ){ - jsonDebugPrintParse(pX); - jsonDebugPrintNode(pResult); - jsonReturnJson(pX, pResult, ctx, 0); - }else{ - sqlite3_result_error_nomem(ctx); + assert( argc==2 ); + pTarget = jsonParseFuncArg(ctx, argv[0], JSON_EDITABLE); + if( pTarget==0 ) return; + pPatch = jsonParseFuncArg(ctx, argv[1], 0); + if( pPatch ){ + rc = jsonMergePatch(pTarget, 0, pPatch, 0); + if( rc==JSON_MERGE_OK ){ + jsonReturnParse(ctx, pTarget); + }else if( rc==JSON_MERGE_OOM ){ + sqlite3_result_error_nomem(ctx); + }else{ + sqlite3_result_error(ctx, "malformed JSON", -1); + } + jsonParseFree(pPatch); } + jsonParseFree(pTarget); } @@ -205298,23 +207176,23 @@ static void jsonObjectFunc( "of arguments", -1); return; } - jsonInit(&jx, ctx); + jsonStringInit(&jx, ctx); jsonAppendChar(&jx, '{'); for(i=0; i1); - if( pParse==0 ) return; - for(i=1; i<(u32)argc; i++){ + p = jsonParseFuncArg(ctx, argv[0], argc>1 ? JSON_EDITABLE : 0); + if( p==0 ) return; + for(i=1; inErr ) goto remove_done; - if( pNode ){ - pNode->jnFlags |= JNODE_REMOVE; - pParse->hasMod = 1; - pParse->useMod = 1; - } - } - if( (pParse->aNode[0].jnFlags & JNODE_REMOVE)==0 ){ - jsonReturnJson(pParse, pParse->aNode, ctx, 1); - } -remove_done: - jsonDebugPrintParse(p); -} - -/* -** Substitute the value at iNode with the pValue parameter. -*/ -static void jsonReplaceNode( - sqlite3_context *pCtx, - JsonParse *p, - int iNode, - sqlite3_value *pValue -){ - int idx = jsonParseAddSubstNode(p, iNode); - if( idx<=0 ){ - assert( p->oom ); - return; - } - switch( sqlite3_value_type(pValue) ){ - case SQLITE_NULL: { - jsonParseAddNode(p, JSON_NULL, 0, 0); - break; - } - case SQLITE_FLOAT: { - char *z = sqlite3_mprintf("%!0.15g", sqlite3_value_double(pValue)); - int n; - if( z==0 ){ - p->oom = 1; - break; - } - n = sqlite3Strlen30(z); - jsonParseAddNode(p, JSON_REAL, n, z); - jsonParseAddCleanup(p, sqlite3_free, z); - break; - } - case SQLITE_INTEGER: { - char *z = sqlite3_mprintf("%lld", sqlite3_value_int64(pValue)); - int n; - if( z==0 ){ - p->oom = 1; - break; - } - n = sqlite3Strlen30(z); - jsonParseAddNode(p, JSON_INT, n, z); - jsonParseAddCleanup(p, sqlite3_free, z); - - break; - } - case SQLITE_TEXT: { - const char *z = (const char*)sqlite3_value_text(pValue); - u32 n = (u32)sqlite3_value_bytes(pValue); - if( z==0 ){ - p->oom = 1; - break; - } - if( sqlite3_value_subtype(pValue)!=JSON_SUBTYPE ){ - char *zCopy = sqlite3_malloc64( n+1 ); - int k; - if( zCopy ){ - memcpy(zCopy, z, n); - zCopy[n] = 0; - jsonParseAddCleanup(p, sqlite3_free, zCopy); - }else{ - p->oom = 1; - sqlite3_result_error_nomem(pCtx); - } - k = jsonParseAddNode(p, JSON_STRING, n, zCopy); - assert( k>0 || p->oom ); - if( p->oom==0 ) p->aNode[k].jnFlags |= JNODE_RAW; + if( zPath==0 ){ + goto json_remove_done; + } + if( zPath[0]!='$' ){ + goto json_remove_patherror; + } + if( zPath[1]==0 ){ + /* json_remove(j,'$') returns NULL */ + goto json_remove_done; + } + p->eEdit = JEDIT_DEL; + p->delta = 0; + rc = jsonLookupStep(p, 0, zPath+1, 0); + if( JSON_LOOKUP_ISERROR(rc) ){ + if( rc==JSON_LOOKUP_NOTFOUND ){ + continue; /* No-op */ + }else if( rc==JSON_LOOKUP_PATHERROR ){ + jsonBadPathError(ctx, zPath); }else{ - JsonParse *pPatch = jsonParseCached(pCtx, pValue, pCtx, 1); - if( pPatch==0 ){ - p->oom = 1; - break; - } - jsonParseAddNodeArray(p, pPatch->aNode, pPatch->nNode); - /* The nodes copied out of pPatch and into p likely contain - ** u.zJContent pointers into pPatch->zJson. So preserve the - ** content of pPatch until p is destroyed. */ - assert( pPatch->nJPRef>=1 ); - pPatch->nJPRef++; - jsonParseAddCleanup(p, (void(*)(void*))jsonParseFree, pPatch); + sqlite3_result_error(ctx, "malformed JSON", -1); } - break; - } - default: { - jsonParseAddNode(p, JSON_NULL, 0, 0); - sqlite3_result_error(pCtx, "JSON cannot hold BLOB values", -1); - p->nErr++; - break; + goto json_remove_done; } } + jsonReturnParse(ctx, p); + jsonParseFree(p); + return; + +json_remove_patherror: + jsonBadPathError(ctx, zPath); + +json_remove_done: + jsonParseFree(p); + return; } /* @@ -205457,32 +207265,12 @@ static void jsonReplaceFunc( int argc, sqlite3_value **argv ){ - JsonParse *pParse; /* The parse */ - JsonNode *pNode; - const char *zPath; - u32 i; - if( argc<1 ) return; if( (argc&1)==0 ) { jsonWrongNumArgs(ctx, "replace"); return; } - pParse = jsonParseCached(ctx, argv[0], ctx, argc>1); - if( pParse==0 ) return; - pParse->nJPRef++; - for(i=1; i<(u32)argc; i+=2){ - zPath = (const char*)sqlite3_value_text(argv[i]); - pParse->useMod = 1; - pNode = jsonLookup(pParse, zPath, 0, ctx); - if( pParse->nErr ) goto replace_err; - if( pNode ){ - jsonReplaceNode(ctx, pParse, (u32)(pNode - pParse->aNode), argv[i+1]); - } - } - jsonReturnJson(pParse, pParse->aNode, ctx, 1); -replace_err: - jsonDebugPrintParse(pParse); - jsonParseFree(pParse); + jsonInsertIntoBlob(ctx, argc, argv, JEDIT_REPL); } @@ -205503,39 +207291,16 @@ static void jsonSetFunc( int argc, sqlite3_value **argv ){ - JsonParse *pParse; /* The parse */ - JsonNode *pNode; - const char *zPath; - u32 i; - int bApnd; - int bIsSet = sqlite3_user_data(ctx)!=0; + + int flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + int bIsSet = (flags&JSON_ISSET)!=0; if( argc<1 ) return; if( (argc&1)==0 ) { jsonWrongNumArgs(ctx, bIsSet ? "set" : "insert"); return; } - pParse = jsonParseCached(ctx, argv[0], ctx, argc>1); - if( pParse==0 ) return; - pParse->nJPRef++; - for(i=1; i<(u32)argc; i+=2){ - zPath = (const char*)sqlite3_value_text(argv[i]); - bApnd = 0; - pParse->useMod = 1; - pNode = jsonLookup(pParse, zPath, &bApnd, ctx); - if( pParse->oom ){ - sqlite3_result_error_nomem(ctx); - goto jsonSetDone; - }else if( pParse->nErr ){ - goto jsonSetDone; - }else if( pNode && (bApnd || bIsSet) ){ - jsonReplaceNode(ctx, pParse, (u32)(pNode - pParse->aNode), argv[i+1]); - } - } - jsonDebugPrintParse(pParse); - jsonReturnJson(pParse, pParse->aNode, ctx, 1); -jsonSetDone: - jsonParseFree(pParse); + jsonInsertIntoBlob(ctx, argc, argv, bIsSet ? JEDIT_SET : JEDIT_INS); } /* @@ -205551,27 +207316,93 @@ static void jsonTypeFunc( sqlite3_value **argv ){ JsonParse *p; /* The parse */ - const char *zPath; - JsonNode *pNode; + const char *zPath = 0; + u32 i; - p = jsonParseCached(ctx, argv[0], ctx, 0); + p = jsonParseFuncArg(ctx, argv[0], 0); if( p==0 ) return; if( argc==2 ){ zPath = (const char*)sqlite3_value_text(argv[1]); - pNode = jsonLookup(p, zPath, 0, ctx); + if( zPath==0 ) goto json_type_done; + if( zPath[0]!='$' ){ + jsonBadPathError(ctx, zPath); + goto json_type_done; + } + i = jsonLookupStep(p, 0, zPath+1, 0); + if( JSON_LOOKUP_ISERROR(i) ){ + if( i==JSON_LOOKUP_NOTFOUND ){ + /* no-op */ + }else if( i==JSON_LOOKUP_PATHERROR ){ + jsonBadPathError(ctx, zPath); + }else{ + sqlite3_result_error(ctx, "malformed JSON", -1); + } + goto json_type_done; + } }else{ - pNode = p->aNode; - } - if( pNode ){ - sqlite3_result_text(ctx, jsonType[pNode->eType], -1, SQLITE_STATIC); + i = 0; } + sqlite3_result_text(ctx, jsonbType[p->aBlob[i]&0x0f], -1, SQLITE_STATIC); +json_type_done: + jsonParseFree(p); } /* ** json_valid(JSON) -** -** Return 1 if JSON is a well-formed canonical JSON string according -** to RFC-7159. Return 0 otherwise. +** json_valid(JSON, FLAGS) +** +** Check the JSON argument to see if it is well-formed. The FLAGS argument +** encodes the various constraints on what is meant by "well-formed": +** +** 0x01 Canonical RFC-8259 JSON text +** 0x02 JSON text with optional JSON-5 extensions +** 0x04 Superficially appears to be JSONB +** 0x08 Strictly well-formed JSONB +** +** If the FLAGS argument is omitted, it defaults to 1. Useful values for +** FLAGS include: +** +** 1 Strict canonical JSON text +** 2 JSON text perhaps with JSON-5 extensions +** 4 Superficially appears to be JSONB +** 5 Canonical JSON text or superficial JSONB +** 6 JSON-5 text or superficial JSONB +** 8 Strict JSONB +** 9 Canonical JSON text or strict JSONB +** 10 JSON-5 text or strict JSONB +** +** Other flag combinations are redundant. For example, every canonical +** JSON text is also well-formed JSON-5 text, so FLAG values 2 and 3 +** are the same. Similarly, any input that passes a strict JSONB validation +** will also pass the superficial validation so 12 through 15 are the same +** as 8 through 11 respectively. +** +** This routine runs in linear time to validate text and when doing strict +** JSONB validation. Superficial JSONB validation is constant time, +** assuming the BLOB is already in memory. The performance advantage +** of superficial JSONB validation is why that option is provided. +** Application developers can choose to do fast superficial validation or +** slower strict validation, according to their specific needs. +** +** Only the lower four bits of the FLAGS argument are currently used. +** Higher bits are reserved for future expansion. To facilitate +** compatibility, the current implementation raises an error if any bit +** in FLAGS is set other than the lower four bits. +** +** The original circa 2015 implementation of the JSON routines in +** SQLite only supported canonical RFC-8259 JSON text and the json_valid() +** function only accepted one argument. That is why the default value +** for the FLAGS argument is 1, since FLAGS=1 causes this routine to only +** recognize canonical RFC-8259 JSON text as valid. The extra FLAGS +** argument was added when the JSON routines were extended to support +** JSON5-like extensions and binary JSONB stored in BLOBs. +** +** Return Values: +** +** * Raise an error if FLAGS is outside the range of 1 to 15. +** * Return NULL if the input is NULL +** * Return 1 if the input is well-formed. +** * Return 0 if the input is not well-formed. */ static void jsonValidFunc( sqlite3_context *ctx, @@ -205579,79 +207410,125 @@ static void jsonValidFunc( sqlite3_value **argv ){ JsonParse *p; /* The parse */ - UNUSED_PARAMETER(argc); - if( sqlite3_value_type(argv[0])==SQLITE_NULL ){ + u8 flags = 1; + u8 res = 0; + if( argc==2 ){ + i64 f = sqlite3_value_int64(argv[1]); + if( f<1 || f>15 ){ + sqlite3_result_error(ctx, "FLAGS parameter to json_valid() must be" + " between 1 and 15", -1); + return; + } + flags = f & 0x0f; + } + switch( sqlite3_value_type(argv[0]) ){ + case SQLITE_NULL: { #ifdef SQLITE_LEGACY_JSON_VALID - /* Incorrect legacy behavior was to return FALSE for a NULL input */ - sqlite3_result_int(ctx, 0); + /* Incorrect legacy behavior was to return FALSE for a NULL input */ + sqlite3_result_int(ctx, 0); #endif - return; - } - p = jsonParseCached(ctx, argv[0], 0, 0); - if( p==0 || p->oom ){ - sqlite3_result_error_nomem(ctx); - sqlite3_free(p); - }else{ - sqlite3_result_int(ctx, p->nErr==0 && (p->hasNonstd==0 || p->useMod)); - if( p->nErr ) jsonParseFree(p); + return; + } + case SQLITE_BLOB: { + if( (flags & 0x0c)!=0 && jsonFuncArgMightBeBinary(argv[0]) ){ + if( flags & 0x04 ){ + /* Superficial checking only - accomplished by the + ** jsonFuncArgMightBeBinary() call above. */ + res = 1; + }else{ + /* Strict checking. Check by translating BLOB->TEXT->BLOB. If + ** no errors occur, call that a "strict check". */ + JsonParse px; + u32 iErr; + memset(&px, 0, sizeof(px)); + px.aBlob = (u8*)sqlite3_value_blob(argv[0]); + px.nBlob = sqlite3_value_bytes(argv[0]); + iErr = jsonbValidityCheck(&px, 0, px.nBlob, 1); + res = iErr==0; + } + } + break; + } + default: { + JsonParse px; + if( (flags & 0x3)==0 ) break; + memset(&px, 0, sizeof(px)); + + p = jsonParseFuncArg(ctx, argv[0], JSON_KEEPERROR); + if( p ){ + if( p->oom ){ + sqlite3_result_error_nomem(ctx); + }else if( p->nErr ){ + /* no-op */ + }else if( (flags & 0x02)!=0 || p->hasNonstd==0 ){ + res = 1; + } + jsonParseFree(p); + }else{ + sqlite3_result_error_nomem(ctx); + } + break; + } } + sqlite3_result_int(ctx, res); } /* ** json_error_position(JSON) ** -** If the argument is not an interpretable JSON string, then return the 1-based -** character position at which the parser first recognized that the input -** was in error. The left-most character is 1. If the string is valid -** JSON, then return 0. -** -** Note that json_valid() is only true for strictly conforming canonical JSON. -** But this routine returns zero if the input contains extension. Thus: +** If the argument is NULL, return NULL ** -** (1) If the input X is strictly conforming canonical JSON: +** If the argument is BLOB, do a full validity check and return non-zero +** if the check fails. The return value is the approximate 1-based offset +** to the byte of the element that contains the first error. ** -** json_valid(X) returns true -** json_error_position(X) returns 0 -** -** (2) If the input X is JSON but it includes extension (such as JSON5) that -** are not part of RFC-8259: -** -** json_valid(X) returns false -** json_error_position(X) return 0 -** -** (3) If the input X cannot be interpreted as JSON even taking extensions -** into account: -** -** json_valid(X) return false -** json_error_position(X) returns 1 or more +** Otherwise interpret the argument is TEXT (even if it is numeric) and +** return the 1-based character position for where the parser first recognized +** that the input was not valid JSON, or return 0 if the input text looks +** ok. JSON-5 extensions are accepted. */ static void jsonErrorFunc( sqlite3_context *ctx, int argc, sqlite3_value **argv ){ - JsonParse *p; /* The parse */ + i64 iErrPos = 0; /* Error position to be returned */ + JsonParse s; + + assert( argc==1 ); UNUSED_PARAMETER(argc); - if( sqlite3_value_type(argv[0])==SQLITE_NULL ) return; - p = jsonParseCached(ctx, argv[0], 0, 0); - if( p==0 || p->oom ){ + memset(&s, 0, sizeof(s)); + s.db = sqlite3_context_db_handle(ctx); + if( jsonFuncArgMightBeBinary(argv[0]) ){ + s.aBlob = (u8*)sqlite3_value_blob(argv[0]); + s.nBlob = sqlite3_value_bytes(argv[0]); + iErrPos = (i64)jsonbValidityCheck(&s, 0, s.nBlob, 1); + }else{ + s.zJson = (char*)sqlite3_value_text(argv[0]); + if( s.zJson==0 ) return; /* NULL input or OOM */ + s.nJson = sqlite3_value_bytes(argv[0]); + if( jsonConvertTextToBlob(&s,0) ){ + if( s.oom ){ + iErrPos = -1; + }else{ + /* Convert byte-offset s.iErr into a character offset */ + u32 k; + assert( s.zJson!=0 ); /* Because s.oom is false */ + for(k=0; knErr==0 ){ - sqlite3_result_int(ctx, 0); }else{ - int n = 1; - u32 i; - const char *z = (const char*)sqlite3_value_text(argv[0]); - for(i=0; iiErr && ALWAYS(z[i]); i++){ - if( (z[i]&0xc0)!=0x80 ) n++; - } - sqlite3_result_int(ctx, n); - jsonParseFree(p); + sqlite3_result_int64(ctx, iErrPos); } } - /**************************************************************************** ** Aggregate SQL function implementations ****************************************************************************/ @@ -205670,24 +207547,34 @@ static void jsonArrayStep( pStr = (JsonString*)sqlite3_aggregate_context(ctx, sizeof(*pStr)); if( pStr ){ if( pStr->zBuf==0 ){ - jsonInit(pStr, ctx); + jsonStringInit(pStr, ctx); jsonAppendChar(pStr, '['); }else if( pStr->nUsed>1 ){ jsonAppendChar(pStr, ','); } pStr->pCtx = ctx; - jsonAppendValue(pStr, argv[0]); + jsonAppendSqlValue(pStr, argv[0]); } } static void jsonArrayCompute(sqlite3_context *ctx, int isFinal){ JsonString *pStr; pStr = (JsonString*)sqlite3_aggregate_context(ctx, 0); if( pStr ){ + int flags; pStr->pCtx = ctx; jsonAppendChar(pStr, ']'); - if( pStr->bErr ){ - if( pStr->bErr==1 ) sqlite3_result_error_nomem(ctx); - assert( pStr->bStatic ); + flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + if( pStr->eErr ){ + jsonReturnString(pStr, 0, 0); + return; + }else if( flags & JSON_BLOB ){ + jsonReturnStringAsBlob(pStr); + if( isFinal ){ + if( !pStr->bStatic ) sqlite3RCStrUnref(pStr->zBuf); + }else{ + pStr->nUsed--; + } + return; }else if( isFinal ){ sqlite3_result_text(ctx, pStr->zBuf, (int)pStr->nUsed, pStr->bStatic ? SQLITE_TRANSIENT : @@ -205776,27 +207663,38 @@ static void jsonObjectStep( pStr = (JsonString*)sqlite3_aggregate_context(ctx, sizeof(*pStr)); if( pStr ){ if( pStr->zBuf==0 ){ - jsonInit(pStr, ctx); + jsonStringInit(pStr, ctx); jsonAppendChar(pStr, '{'); }else if( pStr->nUsed>1 ){ jsonAppendChar(pStr, ','); } pStr->pCtx = ctx; z = (const char*)sqlite3_value_text(argv[0]); - n = (u32)sqlite3_value_bytes(argv[0]); + n = sqlite3Strlen30(z); jsonAppendString(pStr, z, n); jsonAppendChar(pStr, ':'); - jsonAppendValue(pStr, argv[1]); + jsonAppendSqlValue(pStr, argv[1]); } } static void jsonObjectCompute(sqlite3_context *ctx, int isFinal){ JsonString *pStr; pStr = (JsonString*)sqlite3_aggregate_context(ctx, 0); if( pStr ){ + int flags; jsonAppendChar(pStr, '}'); - if( pStr->bErr ){ - if( pStr->bErr==1 ) sqlite3_result_error_nomem(ctx); - assert( pStr->bStatic ); + pStr->pCtx = ctx; + flags = SQLITE_PTR_TO_INT(sqlite3_user_data(ctx)); + if( pStr->eErr ){ + jsonReturnString(pStr, 0, 0); + return; + }else if( flags & JSON_BLOB ){ + jsonReturnStringAsBlob(pStr); + if( isFinal ){ + if( !pStr->bStatic ) sqlite3RCStrUnref(pStr->zBuf); + }else{ + pStr->nUsed--; + } + return; }else if( isFinal ){ sqlite3_result_text(ctx, pStr->zBuf, (int)pStr->nUsed, pStr->bStatic ? SQLITE_TRANSIENT : @@ -205824,19 +207722,37 @@ static void jsonObjectFinal(sqlite3_context *ctx){ /**************************************************************************** ** The json_each virtual table ****************************************************************************/ +typedef struct JsonParent JsonParent; +struct JsonParent { + u32 iHead; /* Start of object or array */ + u32 iValue; /* Start of the value */ + u32 iEnd; /* First byte past the end */ + u32 nPath; /* Length of path */ + i64 iKey; /* Key for JSONB_ARRAY */ +}; + typedef struct JsonEachCursor JsonEachCursor; struct JsonEachCursor { sqlite3_vtab_cursor base; /* Base class - must be first */ u32 iRowid; /* The rowid */ - u32 iBegin; /* The first node of the scan */ - u32 i; /* Index in sParse.aNode[] of current row */ + u32 i; /* Index in sParse.aBlob[] of current row */ u32 iEnd; /* EOF when i equals or exceeds this value */ - u8 eType; /* Type of top-level element */ + u32 nRoot; /* Size of the root path in bytes */ + u8 eType; /* Type of the container for element i */ u8 bRecursive; /* True for json_tree(). False for json_each() */ - char *zJson; /* Input JSON */ - char *zRoot; /* Path by which to filter zJson */ + u32 nParent; /* Current nesting depth */ + u32 nParentAlloc; /* Space allocated for aParent[] */ + JsonParent *aParent; /* Parent elements of i */ + sqlite3 *db; /* Database connection */ + JsonString path; /* Current path */ JsonParse sParse; /* Parse of the input JSON */ }; +typedef struct JsonEachConnection JsonEachConnection; +struct JsonEachConnection { + sqlite3_vtab base; /* Base class - must be first */ + sqlite3 *db; /* Database connection */ +}; + /* Constructor for the json_each virtual table */ static int jsonEachConnect( @@ -205846,7 +207762,7 @@ static int jsonEachConnect( sqlite3_vtab **ppVtab, char **pzErr ){ - sqlite3_vtab *pNew; + JsonEachConnection *pNew; int rc; /* Column numbers */ @@ -205872,28 +207788,32 @@ static int jsonEachConnect( "CREATE TABLE x(key,value,type,atom,id,parent,fullkey,path," "json HIDDEN,root HIDDEN)"); if( rc==SQLITE_OK ){ - pNew = *ppVtab = sqlite3_malloc( sizeof(*pNew) ); + pNew = (JsonEachConnection*)sqlite3DbMallocZero(db, sizeof(*pNew)); + *ppVtab = (sqlite3_vtab*)pNew; if( pNew==0 ) return SQLITE_NOMEM; - memset(pNew, 0, sizeof(*pNew)); sqlite3_vtab_config(db, SQLITE_VTAB_INNOCUOUS); + pNew->db = db; } return rc; } /* destructor for json_each virtual table */ static int jsonEachDisconnect(sqlite3_vtab *pVtab){ - sqlite3_free(pVtab); + JsonEachConnection *p = (JsonEachConnection*)pVtab; + sqlite3DbFree(p->db, pVtab); return SQLITE_OK; } /* constructor for a JsonEachCursor object for json_each(). */ static int jsonEachOpenEach(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ + JsonEachConnection *pVtab = (JsonEachConnection*)p; JsonEachCursor *pCur; UNUSED_PARAMETER(p); - pCur = sqlite3_malloc( sizeof(*pCur) ); + pCur = sqlite3DbMallocZero(pVtab->db, sizeof(*pCur)); if( pCur==0 ) return SQLITE_NOMEM; - memset(pCur, 0, sizeof(*pCur)); + pCur->db = pVtab->db; + jsonStringZero(&pCur->path); *ppCursor = &pCur->base; return SQLITE_OK; } @@ -205911,21 +207831,24 @@ static int jsonEachOpenTree(sqlite3_vtab *p, sqlite3_vtab_cursor **ppCursor){ /* Reset a JsonEachCursor back to its original state. Free any memory ** held. */ static void jsonEachCursorReset(JsonEachCursor *p){ - sqlite3_free(p->zRoot); jsonParseReset(&p->sParse); + jsonStringReset(&p->path); + sqlite3DbFree(p->db, p->aParent); p->iRowid = 0; p->i = 0; + p->aParent = 0; + p->nParent = 0; + p->nParentAlloc = 0; p->iEnd = 0; p->eType = 0; - p->zJson = 0; - p->zRoot = 0; } /* Destructor for a jsonEachCursor object */ static int jsonEachClose(sqlite3_vtab_cursor *cur){ JsonEachCursor *p = (JsonEachCursor*)cur; jsonEachCursorReset(p); - sqlite3_free(cur); + + sqlite3DbFree(p->db, cur); return SQLITE_OK; } @@ -205936,200 +207859,230 @@ static int jsonEachEof(sqlite3_vtab_cursor *cur){ return p->i >= p->iEnd; } -/* Advance the cursor to the next element for json_tree() */ -static int jsonEachNext(sqlite3_vtab_cursor *cur){ - JsonEachCursor *p = (JsonEachCursor*)cur; - if( p->bRecursive ){ - if( p->sParse.aNode[p->i].jnFlags & JNODE_LABEL ) p->i++; - p->i++; - p->iRowid++; - if( p->iiEnd ){ - u32 iUp = p->sParse.aUp[p->i]; - JsonNode *pUp = &p->sParse.aNode[iUp]; - p->eType = pUp->eType; - if( pUp->eType==JSON_ARRAY ){ - assert( pUp->eU==0 || pUp->eU==3 ); - testcase( pUp->eU==3 ); - VVA( pUp->eU = 3 ); - if( iUp==p->i-1 ){ - pUp->u.iKey = 0; - }else{ - pUp->u.iKey++; +/* +** If the cursor is currently pointing at the label of a object entry, +** then return the index of the value. For all other cases, return the +** current pointer position, which is the value. +*/ +static int jsonSkipLabel(JsonEachCursor *p){ + if( p->eType==JSONB_OBJECT ){ + u32 sz = 0; + u32 n = jsonbPayloadSize(&p->sParse, p->i, &sz); + return p->i + n + sz; + }else{ + return p->i; + } +} + +/* +** Append the path name for the current element. +*/ +static void jsonAppendPathName(JsonEachCursor *p){ + assert( p->nParent>0 ); + assert( p->eType==JSONB_ARRAY || p->eType==JSONB_OBJECT ); + if( p->eType==JSONB_ARRAY ){ + jsonPrintf(30, &p->path, "[%lld]", p->aParent[p->nParent-1].iKey); + }else{ + u32 n, sz = 0, k, i; + const char *z; + int needQuote = 0; + n = jsonbPayloadSize(&p->sParse, p->i, &sz); + k = p->i + n; + z = (const char*)&p->sParse.aBlob[k]; + if( sz==0 || !sqlite3Isalpha(z[0]) ){ + needQuote = 1; + }else{ + for(i=0; ieType ){ - case JSON_ARRAY: { - p->i += jsonNodeSize(&p->sParse.aNode[p->i]); - p->iRowid++; - break; - } - case JSON_OBJECT: { - p->i += 1 + jsonNodeSize(&p->sParse.aNode[p->i+1]); - p->iRowid++; - break; - } - default: { - p->i = p->iEnd; - break; - } + if( needQuote ){ + jsonPrintf(sz+4,&p->path,".\"%.*s\"", sz, z); + }else{ + jsonPrintf(sz+2,&p->path,".%.*s", sz, z); } } - return SQLITE_OK; } -/* Append an object label to the JSON Path being constructed -** in pStr. -*/ -static void jsonAppendObjectPathElement( - JsonString *pStr, - JsonNode *pNode -){ - int jj, nn; - const char *z; - assert( pNode->eType==JSON_STRING ); - assert( pNode->jnFlags & JNODE_LABEL ); - assert( pNode->eU==1 ); - z = pNode->u.zJContent; - nn = pNode->n; - if( (pNode->jnFlags & JNODE_RAW)==0 ){ - assert( nn>=2 ); - assert( z[0]=='"' || z[0]=='\'' ); - assert( z[nn-1]=='"' || z[0]=='\'' ); - if( nn>2 && sqlite3Isalpha(z[1]) ){ - for(jj=2; jjbRecursive ){ + u8 x; + u8 levelChange = 0; + u32 n, sz = 0; + u32 i = jsonSkipLabel(p); + x = p->sParse.aBlob[i] & 0x0f; + n = jsonbPayloadSize(&p->sParse, i, &sz); + if( x==JSONB_OBJECT || x==JSONB_ARRAY ){ + JsonParent *pParent; + if( p->nParent>=p->nParentAlloc ){ + JsonParent *pNew; + u64 nNew; + nNew = p->nParentAlloc*2 + 3; + pNew = sqlite3DbRealloc(p->db, p->aParent, sizeof(JsonParent)*nNew); + if( pNew==0 ) return SQLITE_NOMEM; + p->nParentAlloc = (u32)nNew; + p->aParent = pNew; + } + levelChange = 1; + pParent = &p->aParent[p->nParent]; + pParent->iHead = p->i; + pParent->iValue = i; + pParent->iEnd = i + n + sz; + pParent->iKey = -1; + pParent->nPath = (u32)p->path.nUsed; + if( p->eType && p->nParent ){ + jsonAppendPathName(p); + if( p->path.eErr ) rc = SQLITE_NOMEM; + } + p->nParent++; + p->i = i + n; + }else{ + p->i = i + n + sz; + } + while( p->nParent>0 && p->i >= p->aParent[p->nParent-1].iEnd ){ + p->nParent--; + p->path.nUsed = p->aParent[p->nParent].nPath; + levelChange = 1; + } + if( levelChange ){ + if( p->nParent>0 ){ + JsonParent *pParent = &p->aParent[p->nParent-1]; + u32 iVal = pParent->iValue; + p->eType = p->sParse.aBlob[iVal] & 0x0f; + }else{ + p->eType = 0; } } + }else{ + u32 n, sz = 0; + u32 i = jsonSkipLabel(p); + n = jsonbPayloadSize(&p->sParse, i, &sz); + p->i = i + n + sz; + } + if( p->eType==JSONB_ARRAY && p->nParent ){ + p->aParent[p->nParent-1].iKey++; } - jsonPrintf(nn+2, pStr, ".%.*s", nn, z); + p->iRowid++; + return rc; } -/* Append the name of the path for element i to pStr +/* Length of the path for rowid==0 in bRecursive mode. */ -static void jsonEachComputePath( - JsonEachCursor *p, /* The cursor */ - JsonString *pStr, /* Write the path here */ - u32 i /* Path to this element */ -){ - JsonNode *pNode, *pUp; - u32 iUp; - if( i==0 ){ - jsonAppendChar(pStr, '$'); - return; - } - iUp = p->sParse.aUp[i]; - jsonEachComputePath(p, pStr, iUp); - pNode = &p->sParse.aNode[i]; - pUp = &p->sParse.aNode[iUp]; - if( pUp->eType==JSON_ARRAY ){ - assert( pUp->eU==3 || (pUp->eU==0 && pUp->u.iKey==0) ); - testcase( pUp->eU==0 ); - jsonPrintf(30, pStr, "[%d]", pUp->u.iKey); - }else{ - assert( pUp->eType==JSON_OBJECT ); - if( (pNode->jnFlags & JNODE_LABEL)==0 ) pNode--; - jsonAppendObjectPathElement(pStr, pNode); +static int jsonEachPathLength(JsonEachCursor *p){ + u32 n = p->path.nUsed; + char *z = p->path.zBuf; + if( p->iRowid==0 && p->bRecursive && n>=2 ){ + while( n>1 ){ + n--; + if( z[n]=='[' || z[n]=='.' ){ + u32 x, sz = 0; + char cSaved = z[n]; + z[n] = 0; + assert( p->sParse.eEdit==0 ); + x = jsonLookupStep(&p->sParse, 0, z+1, 0); + z[n] = cSaved; + if( JSON_LOOKUP_ISERROR(x) ) continue; + if( x + jsonbPayloadSize(&p->sParse, x, &sz) == p->i ) break; + } + } } + return n; } /* Return the value of a column */ static int jsonEachColumn( sqlite3_vtab_cursor *cur, /* The cursor */ sqlite3_context *ctx, /* First argument to sqlite3_result_...() */ - int i /* Which column to return */ + int iColumn /* Which column to return */ ){ JsonEachCursor *p = (JsonEachCursor*)cur; - JsonNode *pThis = &p->sParse.aNode[p->i]; - switch( i ){ + switch( iColumn ){ case JEACH_KEY: { - if( p->i==0 ) break; - if( p->eType==JSON_OBJECT ){ - jsonReturn(&p->sParse, pThis, ctx); - }else if( p->eType==JSON_ARRAY ){ - u32 iKey; - if( p->bRecursive ){ - if( p->iRowid==0 ) break; - assert( p->sParse.aNode[p->sParse.aUp[p->i]].eU==3 ); - iKey = p->sParse.aNode[p->sParse.aUp[p->i]].u.iKey; + if( p->nParent==0 ){ + u32 n, j; + if( p->nRoot==1 ) break; + j = jsonEachPathLength(p); + n = p->nRoot - j; + if( n==0 ){ + break; + }else if( p->path.zBuf[j]=='[' ){ + i64 x; + sqlite3Atoi64(&p->path.zBuf[j+1], &x, n-1, SQLITE_UTF8); + sqlite3_result_int64(ctx, x); + }else if( p->path.zBuf[j+1]=='"' ){ + sqlite3_result_text(ctx, &p->path.zBuf[j+2], n-3, SQLITE_TRANSIENT); }else{ - iKey = p->iRowid; + sqlite3_result_text(ctx, &p->path.zBuf[j+1], n-1, SQLITE_TRANSIENT); } - sqlite3_result_int64(ctx, (sqlite3_int64)iKey); + break; + } + if( p->eType==JSONB_OBJECT ){ + jsonReturnFromBlob(&p->sParse, p->i, ctx, 1); + }else{ + assert( p->eType==JSONB_ARRAY ); + sqlite3_result_int64(ctx, p->aParent[p->nParent-1].iKey); } break; } case JEACH_VALUE: { - if( pThis->jnFlags & JNODE_LABEL ) pThis++; - jsonReturn(&p->sParse, pThis, ctx); + u32 i = jsonSkipLabel(p); + jsonReturnFromBlob(&p->sParse, i, ctx, 1); break; } case JEACH_TYPE: { - if( pThis->jnFlags & JNODE_LABEL ) pThis++; - sqlite3_result_text(ctx, jsonType[pThis->eType], -1, SQLITE_STATIC); + u32 i = jsonSkipLabel(p); + u8 eType = p->sParse.aBlob[i] & 0x0f; + sqlite3_result_text(ctx, jsonbType[eType], -1, SQLITE_STATIC); break; } case JEACH_ATOM: { - if( pThis->jnFlags & JNODE_LABEL ) pThis++; - if( pThis->eType>=JSON_ARRAY ) break; - jsonReturn(&p->sParse, pThis, ctx); + u32 i = jsonSkipLabel(p); + if( (p->sParse.aBlob[i] & 0x0f)sParse, i, ctx, 1); + } break; } case JEACH_ID: { - sqlite3_result_int64(ctx, - (sqlite3_int64)p->i + ((pThis->jnFlags & JNODE_LABEL)!=0)); + sqlite3_result_int64(ctx, (sqlite3_int64)p->i); break; } case JEACH_PARENT: { - if( p->i>p->iBegin && p->bRecursive ){ - sqlite3_result_int64(ctx, (sqlite3_int64)p->sParse.aUp[p->i]); + if( p->nParent>0 && p->bRecursive ){ + sqlite3_result_int64(ctx, p->aParent[p->nParent-1].iHead); } break; } case JEACH_FULLKEY: { - JsonString x; - jsonInit(&x, ctx); - if( p->bRecursive ){ - jsonEachComputePath(p, &x, p->i); - }else{ - if( p->zRoot ){ - jsonAppendRaw(&x, p->zRoot, (int)strlen(p->zRoot)); - }else{ - jsonAppendChar(&x, '$'); - } - if( p->eType==JSON_ARRAY ){ - jsonPrintf(30, &x, "[%d]", p->iRowid); - }else if( p->eType==JSON_OBJECT ){ - jsonAppendObjectPathElement(&x, pThis); - } - } - jsonResult(&x); + u64 nBase = p->path.nUsed; + if( p->nParent ) jsonAppendPathName(p); + sqlite3_result_text64(ctx, p->path.zBuf, p->path.nUsed, + SQLITE_TRANSIENT, SQLITE_UTF8); + p->path.nUsed = nBase; break; } case JEACH_PATH: { - if( p->bRecursive ){ - JsonString x; - jsonInit(&x, ctx); - jsonEachComputePath(p, &x, p->sParse.aUp[p->i]); - jsonResult(&x); - break; - } - /* For json_each() path and root are the same so fall through - ** into the root case */ - /* no break */ deliberate_fall_through + u32 n = jsonEachPathLength(p); + sqlite3_result_text64(ctx, p->path.zBuf, n, + SQLITE_TRANSIENT, SQLITE_UTF8); + break; } default: { - const char *zRoot = p->zRoot; - if( zRoot==0 ) zRoot = "$"; - sqlite3_result_text(ctx, zRoot, -1, SQLITE_STATIC); + sqlite3_result_text(ctx, p->path.zBuf, p->nRoot, SQLITE_STATIC); break; } case JEACH_JSON: { - assert( i==JEACH_JSON ); - sqlite3_result_text(ctx, p->sParse.zJson, -1, SQLITE_STATIC); + if( p->sParse.zJson==0 ){ + sqlite3_result_blob(ctx, p->sParse.aBlob, p->sParse.nBlob, + SQLITE_STATIC); + }else{ + sqlite3_result_text(ctx, p->sParse.zJson, -1, SQLITE_STATIC); + } break; } } @@ -206220,86 +208173,101 @@ static int jsonEachFilter( int argc, sqlite3_value **argv ){ JsonEachCursor *p = (JsonEachCursor*)cur; - const char *z; const char *zRoot = 0; - sqlite3_int64 n; + u32 i, n, sz; UNUSED_PARAMETER(idxStr); UNUSED_PARAMETER(argc); jsonEachCursorReset(p); if( idxNum==0 ) return SQLITE_OK; - z = (const char*)sqlite3_value_text(argv[0]); - if( z==0 ) return SQLITE_OK; memset(&p->sParse, 0, sizeof(p->sParse)); p->sParse.nJPRef = 1; - if( sqlite3ValueIsOfClass(argv[0], sqlite3RCStrUnref) ){ - p->sParse.zJson = sqlite3RCStrRef((char*)z); - }else{ - n = sqlite3_value_bytes(argv[0]); - p->sParse.zJson = sqlite3RCStrNew( n+1 ); - if( p->sParse.zJson==0 ) return SQLITE_NOMEM; - memcpy(p->sParse.zJson, z, (size_t)n+1); - } - p->sParse.bJsonIsRCStr = 1; - p->zJson = p->sParse.zJson; - if( jsonParse(&p->sParse, 0) ){ - int rc = SQLITE_NOMEM; - if( p->sParse.oom==0 ){ - sqlite3_free(cur->pVtab->zErrMsg); - cur->pVtab->zErrMsg = sqlite3_mprintf("malformed JSON"); - if( cur->pVtab->zErrMsg ) rc = SQLITE_ERROR; + p->sParse.db = p->db; + if( sqlite3_value_type(argv[0])==SQLITE_BLOB ){ + if( jsonFuncArgMightBeBinary(argv[0]) ){ + p->sParse.nBlob = sqlite3_value_bytes(argv[0]); + p->sParse.aBlob = (u8*)sqlite3_value_blob(argv[0]); + }else{ + goto json_each_malformed_input; } - jsonEachCursorReset(p); - return rc; - }else if( p->bRecursive && jsonParseFindParents(&p->sParse) ){ - jsonEachCursorReset(p); - return SQLITE_NOMEM; }else{ - JsonNode *pNode = 0; - if( idxNum==3 ){ - const char *zErr = 0; - zRoot = (const char*)sqlite3_value_text(argv[1]); - if( zRoot==0 ) return SQLITE_OK; - n = sqlite3_value_bytes(argv[1]); - p->zRoot = sqlite3_malloc64( n+1 ); - if( p->zRoot==0 ) return SQLITE_NOMEM; - memcpy(p->zRoot, zRoot, (size_t)n+1); - if( zRoot[0]!='$' ){ - zErr = zRoot; - }else{ - pNode = jsonLookupStep(&p->sParse, 0, p->zRoot+1, 0, &zErr); + p->sParse.zJson = (char*)sqlite3_value_text(argv[0]); + p->sParse.nJson = sqlite3_value_bytes(argv[0]); + if( p->sParse.zJson==0 ){ + p->i = p->iEnd = 0; + return SQLITE_OK; + } + if( jsonConvertTextToBlob(&p->sParse, 0) ){ + if( p->sParse.oom ){ + return SQLITE_NOMEM; } - if( zErr ){ + goto json_each_malformed_input; + } + } + if( idxNum==3 ){ + zRoot = (const char*)sqlite3_value_text(argv[1]); + if( zRoot==0 ) return SQLITE_OK; + if( zRoot[0]!='$' ){ + sqlite3_free(cur->pVtab->zErrMsg); + cur->pVtab->zErrMsg = jsonBadPathError(0, zRoot); + jsonEachCursorReset(p); + return cur->pVtab->zErrMsg ? SQLITE_ERROR : SQLITE_NOMEM; + } + p->nRoot = sqlite3Strlen30(zRoot); + if( zRoot[1]==0 ){ + i = p->i = 0; + p->eType = 0; + }else{ + i = jsonLookupStep(&p->sParse, 0, zRoot+1, 0); + if( JSON_LOOKUP_ISERROR(i) ){ + if( i==JSON_LOOKUP_NOTFOUND ){ + p->i = 0; + p->eType = 0; + p->iEnd = 0; + return SQLITE_OK; + } sqlite3_free(cur->pVtab->zErrMsg); - cur->pVtab->zErrMsg = jsonPathSyntaxError(zErr); + cur->pVtab->zErrMsg = jsonBadPathError(0, zRoot); jsonEachCursorReset(p); return cur->pVtab->zErrMsg ? SQLITE_ERROR : SQLITE_NOMEM; - }else if( pNode==0 ){ - return SQLITE_OK; } - }else{ - pNode = p->sParse.aNode; - } - p->iBegin = p->i = (int)(pNode - p->sParse.aNode); - p->eType = pNode->eType; - if( p->eType>=JSON_ARRAY ){ - assert( pNode->eU==0 ); - VVA( pNode->eU = 3 ); - pNode->u.iKey = 0; - p->iEnd = p->i + pNode->n + 1; - if( p->bRecursive ){ - p->eType = p->sParse.aNode[p->sParse.aUp[p->i]].eType; - if( p->i>0 && (p->sParse.aNode[p->i-1].jnFlags & JNODE_LABEL)!=0 ){ - p->i--; - } + if( p->sParse.iLabel ){ + p->i = p->sParse.iLabel; + p->eType = JSONB_OBJECT; }else{ - p->i++; - } - }else{ - p->iEnd = p->i+1; - } + p->i = i; + p->eType = JSONB_ARRAY; + } + } + jsonAppendRaw(&p->path, zRoot, p->nRoot); + }else{ + i = p->i = 0; + p->eType = 0; + p->nRoot = 1; + jsonAppendRaw(&p->path, "$", 1); + } + p->nParent = 0; + n = jsonbPayloadSize(&p->sParse, i, &sz); + p->iEnd = i+n+sz; + if( (p->sParse.aBlob[i] & 0x0f)>=JSONB_ARRAY && !p->bRecursive ){ + p->i = i + n; + p->eType = p->sParse.aBlob[i] & 0x0f; + p->aParent = sqlite3DbMallocZero(p->db, sizeof(JsonParent)); + if( p->aParent==0 ) return SQLITE_NOMEM; + p->nParent = 1; + p->nParentAlloc = 1; + p->aParent[0].iKey = 0; + p->aParent[0].iEnd = p->iEnd; + p->aParent[0].iHead = p->i; + p->aParent[0].iValue = i; } return SQLITE_OK; + +json_each_malformed_input: + sqlite3_free(cur->pVtab->zErrMsg); + cur->pVtab->zErrMsg = sqlite3_mprintf("malformed JSON"); + jsonEachCursorReset(p); + return cur->pVtab->zErrMsg ? SQLITE_ERROR : SQLITE_NOMEM; } /* The methods of the json_each virtual table */ @@ -206368,34 +208336,57 @@ static sqlite3_module jsonTreeModule = { SQLITE_PRIVATE void sqlite3RegisterJsonFunctions(void){ #ifndef SQLITE_OMIT_JSON static FuncDef aJsonFunc[] = { - JFUNCTION(json, 1, 0, jsonRemoveFunc), - JFUNCTION(json_array, -1, 0, jsonArrayFunc), - JFUNCTION(json_array_length, 1, 0, jsonArrayLengthFunc), - JFUNCTION(json_array_length, 2, 0, jsonArrayLengthFunc), - JFUNCTION(json_error_position,1, 0, jsonErrorFunc), - JFUNCTION(json_extract, -1, 0, jsonExtractFunc), - JFUNCTION(->, 2, JSON_JSON, jsonExtractFunc), - JFUNCTION(->>, 2, JSON_SQL, jsonExtractFunc), - JFUNCTION(json_insert, -1, 0, jsonSetFunc), - JFUNCTION(json_object, -1, 0, jsonObjectFunc), - JFUNCTION(json_patch, 2, 0, jsonPatchFunc), - JFUNCTION(json_quote, 1, 0, jsonQuoteFunc), - JFUNCTION(json_remove, -1, 0, jsonRemoveFunc), - JFUNCTION(json_replace, -1, 0, jsonReplaceFunc), - JFUNCTION(json_set, -1, JSON_ISSET, jsonSetFunc), - JFUNCTION(json_type, 1, 0, jsonTypeFunc), - JFUNCTION(json_type, 2, 0, jsonTypeFunc), - JFUNCTION(json_valid, 1, 0, jsonValidFunc), + /* sqlite3_result_subtype() ----, ,--- sqlite3_value_subtype() */ + /* | | */ + /* Uses cache ------, | | ,---- Returns JSONB */ + /* | | | | */ + /* Number of arguments ---, | | | | ,--- Flags */ + /* | | | | | | */ + JFUNCTION(json, 1,1,1, 0,0,0, jsonRemoveFunc), + JFUNCTION(jsonb, 1,1,0, 0,1,0, jsonRemoveFunc), + JFUNCTION(json_array, -1,0,1, 1,0,0, jsonArrayFunc), + JFUNCTION(jsonb_array, -1,0,1, 1,1,0, jsonArrayFunc), + JFUNCTION(json_array_length, 1,1,0, 0,0,0, jsonArrayLengthFunc), + JFUNCTION(json_array_length, 2,1,0, 0,0,0, jsonArrayLengthFunc), + JFUNCTION(json_error_position,1,1,0, 0,0,0, jsonErrorFunc), + JFUNCTION(json_extract, -1,1,1, 0,0,0, jsonExtractFunc), + JFUNCTION(jsonb_extract, -1,1,0, 0,1,0, jsonExtractFunc), + JFUNCTION(->, 2,1,1, 0,0,JSON_JSON, jsonExtractFunc), + JFUNCTION(->>, 2,1,0, 0,0,JSON_SQL, jsonExtractFunc), + JFUNCTION(json_insert, -1,1,1, 1,0,0, jsonSetFunc), + JFUNCTION(jsonb_insert, -1,1,0, 1,1,0, jsonSetFunc), + JFUNCTION(json_object, -1,0,1, 1,0,0, jsonObjectFunc), + JFUNCTION(jsonb_object, -1,0,1, 1,1,0, jsonObjectFunc), + JFUNCTION(json_patch, 2,1,1, 0,0,0, jsonPatchFunc), + JFUNCTION(jsonb_patch, 2,1,0, 0,1,0, jsonPatchFunc), + JFUNCTION(json_quote, 1,0,1, 1,0,0, jsonQuoteFunc), + JFUNCTION(json_remove, -1,1,1, 0,0,0, jsonRemoveFunc), + JFUNCTION(jsonb_remove, -1,1,0, 0,1,0, jsonRemoveFunc), + JFUNCTION(json_replace, -1,1,1, 1,0,0, jsonReplaceFunc), + JFUNCTION(jsonb_replace, -1,1,0, 1,1,0, jsonReplaceFunc), + JFUNCTION(json_set, -1,1,1, 1,0,JSON_ISSET, jsonSetFunc), + JFUNCTION(jsonb_set, -1,1,0, 1,1,JSON_ISSET, jsonSetFunc), + JFUNCTION(json_type, 1,1,0, 0,0,0, jsonTypeFunc), + JFUNCTION(json_type, 2,1,0, 0,0,0, jsonTypeFunc), + JFUNCTION(json_valid, 1,1,0, 0,0,0, jsonValidFunc), + JFUNCTION(json_valid, 2,1,0, 0,0,0, jsonValidFunc), #if SQLITE_DEBUG - JFUNCTION(json_parse, 1, 0, jsonParseFunc), - JFUNCTION(json_test1, 1, 0, jsonTest1Func), + JFUNCTION(json_parse, 1,1,0, 0,0,0, jsonParseFunc), #endif WAGGREGATE(json_group_array, 1, 0, 0, jsonArrayStep, jsonArrayFinal, jsonArrayValue, jsonGroupInverse, - SQLITE_SUBTYPE|SQLITE_UTF8|SQLITE_DETERMINISTIC), + SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8| + SQLITE_DETERMINISTIC), + WAGGREGATE(jsonb_group_array, 1, JSON_BLOB, 0, + jsonArrayStep, jsonArrayFinal, jsonArrayValue, jsonGroupInverse, + SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8|SQLITE_DETERMINISTIC), WAGGREGATE(json_group_object, 2, 0, 0, jsonObjectStep, jsonObjectFinal, jsonObjectValue, jsonGroupInverse, - SQLITE_SUBTYPE|SQLITE_UTF8|SQLITE_DETERMINISTIC) + SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8|SQLITE_DETERMINISTIC), + WAGGREGATE(jsonb_group_object,2, JSON_BLOB, 0, + jsonObjectStep, jsonObjectFinal, jsonObjectValue, jsonGroupInverse, + SQLITE_SUBTYPE|SQLITE_RESULT_SUBTYPE|SQLITE_UTF8| + SQLITE_DETERMINISTIC) }; sqlite3InsertBuiltinFuncs(aJsonFunc, ArraySize(aJsonFunc)); #endif @@ -207143,7 +209134,7 @@ static int nodeAcquire( ** increase its reference count and return it. */ if( (pNode = nodeHashLookup(pRtree, iNode))!=0 ){ - if( pParent && pParent!=pNode->pParent ){ + if( pParent && ALWAYS(pParent!=pNode->pParent) ){ RTREE_IS_CORRUPT(pRtree); return SQLITE_CORRUPT_VTAB; } @@ -209878,7 +211869,7 @@ static int rtreeSqlInit( } sqlite3_free(zSql); } - if( pRtree->nAux ){ + if( pRtree->nAux && rc!=SQLITE_NOMEM ){ pRtree->zReadAuxSql = sqlite3_mprintf( "SELECT * FROM \"%w\".\"%w_rowid\" WHERE rowid=?1", zDb, zPrefix); @@ -210567,15 +212558,13 @@ static int rtreeCheckTable( check.zTab = zTab; /* Find the number of auxiliary columns */ - if( check.rc==SQLITE_OK ){ - pStmt = rtreeCheckPrepare(&check, "SELECT * FROM %Q.'%q_rowid'", zDb, zTab); - if( pStmt ){ - nAux = sqlite3_column_count(pStmt) - 2; - sqlite3_finalize(pStmt); - }else - if( check.rc!=SQLITE_NOMEM ){ - check.rc = SQLITE_OK; - } + pStmt = rtreeCheckPrepare(&check, "SELECT * FROM %Q.'%q_rowid'", zDb, zTab); + if( pStmt ){ + nAux = sqlite3_column_count(pStmt) - 2; + sqlite3_finalize(pStmt); + }else + if( check.rc!=SQLITE_NOMEM ){ + check.rc = SQLITE_OK; } /* Find number of dimensions in the rtree table. */ @@ -210630,6 +212619,7 @@ static int rtreeIntegrity( if( rc==SQLITE_OK && *pzErr ){ *pzErr = sqlite3_mprintf("In RTree %s.%s:\n%z", pRtree->zDb, pRtree->zName, *pzErr); + if( (*pzErr)==0 ) rc = SQLITE_NOMEM; } return rc; } @@ -223300,9 +225290,7 @@ SQLITE_API void sqlite3session_delete(sqlite3_session *pSession){ ** associated hash-tables. */ sessionDeleteTable(pSession, pSession->pTable); - /* Assert that all allocations have been freed and then free the - ** session object itself. */ - // assert( pSession->nMalloc==0 ); + /* Free the session object. */ sqlite3_free(pSession); } @@ -227515,8 +229503,11 @@ struct Fts5PhraseIter { ** created with the "columnsize=0" option. ** ** xColumnText: -** This function attempts to retrieve the text of column iCol of the -** current document. If successful, (*pz) is set to point to a buffer +** If parameter iCol is less than zero, or greater than or equal to the +** number of columns in the table, SQLITE_RANGE is returned. +** +** Otherwise, this function attempts to retrieve the text of column iCol of +** the current document. If successful, (*pz) is set to point to a buffer ** containing the text in utf-8 encoding, (*pn) is set to the size in bytes ** (not characters) of the buffer and SQLITE_OK is returned. Otherwise, ** if an error occurs, an SQLite error code is returned and the final values @@ -227526,8 +229517,10 @@ struct Fts5PhraseIter { ** Returns the number of phrases in the current query expression. ** ** xPhraseSize: -** Returns the number of tokens in phrase iPhrase of the query. Phrases -** are numbered starting from zero. +** If parameter iCol is less than zero, or greater than or equal to the +** number of phrases in the current query, as returned by xPhraseCount, +** 0 is returned. Otherwise, this function returns the number of tokens in +** phrase iPhrase of the query. Phrases are numbered starting from zero. ** ** xInstCount: ** Set *pnInst to the total number of occurrences of all phrases within @@ -227543,12 +229536,13 @@ struct Fts5PhraseIter { ** Query for the details of phrase match iIdx within the current row. ** Phrase matches are numbered starting from zero, so the iIdx argument ** should be greater than or equal to zero and smaller than the value -** output by xInstCount(). +** output by xInstCount(). If iIdx is less than zero or greater than +** or equal to the value returned by xInstCount(), SQLITE_RANGE is returned. ** -** Usually, output parameter *piPhrase is set to the phrase number, *piCol +** Otherwise, output parameter *piPhrase is set to the phrase number, *piCol ** to the column in which it occurs and *piOff the token offset of the -** first token of the phrase. Returns SQLITE_OK if successful, or an error -** code (i.e. SQLITE_NOMEM) if an error occurs. +** first token of the phrase. SQLITE_OK is returned if successful, or an +** error code (i.e. SQLITE_NOMEM) if an error occurs. ** ** This API can be quite slow if used with an FTS5 table created with the ** "detail=none" or "detail=column" option. @@ -227574,6 +229568,10 @@ struct Fts5PhraseIter { ** Invoking Api.xUserData() returns a copy of the pointer passed as ** the third argument to pUserData. ** +** If parameter iPhrase is less than zero, or greater than or equal to +** the number of phrases in the query, as returned by xPhraseCount(), +** this function returns SQLITE_RANGE. +** ** If the callback function returns any value other than SQLITE_OK, the ** query is abandoned and the xQueryPhrase function returns immediately. ** If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK. @@ -227688,9 +229686,42 @@ struct Fts5PhraseIter { ** ** xPhraseNextColumn() ** See xPhraseFirstColumn above. +** +** xQueryToken(pFts5, iPhrase, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase iPhrase of the current +** query. Before returning, output parameter *ppToken is set to point +** to a buffer containing the requested token, and *pnToken to the +** size of this buffer in bytes. +** +** If iPhrase or iToken are less than zero, or if iPhrase is greater than +** or equal to the number of phrases in the query as reported by +** xPhraseCount(), or if iToken is equal to or greater than the number of +** tokens in the phrase, SQLITE_RANGE is returned and *ppToken and *pnToken + are both zeroed. +** +** The output text is not a copy of the query text that specified the +** token. It is the output of the tokenizer module. For tokendata=1 +** tables, this includes any embedded 0x00 and trailing data. +** +** xInstToken(pFts5, iIdx, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase hit iIdx within the +** current row. If iIdx is less than zero or greater than or equal to the +** value returned by xInstCount(), SQLITE_RANGE is returned. Otherwise, +** output variable (*ppToken) is set to point to a buffer containing the +** matching document token, and (*pnToken) to the size of that buffer in +** bytes. This API is not available if the specified token matches a +** prefix query term. In that case both output variables are always set +** to 0. +** +** The output text is not a copy of the document text that was tokenized. +** It is the output of the tokenizer module. For tokendata=1 tables, this +** includes any embedded 0x00 and trailing data. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. */ struct Fts5ExtensionApi { - int iVersion; /* Currently always set to 2 */ + int iVersion; /* Currently always set to 3 */ void *(*xUserData)(Fts5Context*); @@ -227725,6 +229756,13 @@ struct Fts5ExtensionApi { int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*); void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol); + + /* Below this point are iVersion>=3 only */ + int (*xQueryToken)(Fts5Context*, + int iPhrase, int iToken, + const char **ppToken, int *pnToken + ); + int (*xInstToken)(Fts5Context*, int iIdx, int iToken, const char**, int*); }; /* @@ -228199,6 +230237,7 @@ struct Fts5Config { char *zContent; /* content table */ char *zContentRowid; /* "content_rowid=" option value */ int bColumnsize; /* "columnsize=" option value (dflt==1) */ + int bTokendata; /* "tokendata=" option value (dflt==0) */ int eDetail; /* FTS5_DETAIL_XXX value */ char *zContentExprlist; Fts5Tokenizer *pTok; @@ -228387,17 +230426,19 @@ struct Fts5IndexIter { /* ** Values used as part of the flags argument passed to IndexQuery(). */ -#define FTS5INDEX_QUERY_PREFIX 0x0001 /* Prefix query */ -#define FTS5INDEX_QUERY_DESC 0x0002 /* Docs in descending rowid order */ -#define FTS5INDEX_QUERY_TEST_NOIDX 0x0004 /* Do not use prefix index */ -#define FTS5INDEX_QUERY_SCAN 0x0008 /* Scan query (fts5vocab) */ +#define FTS5INDEX_QUERY_PREFIX 0x0001 /* Prefix query */ +#define FTS5INDEX_QUERY_DESC 0x0002 /* Docs in descending rowid order */ +#define FTS5INDEX_QUERY_TEST_NOIDX 0x0004 /* Do not use prefix index */ +#define FTS5INDEX_QUERY_SCAN 0x0008 /* Scan query (fts5vocab) */ /* The following are used internally by the fts5_index.c module. They are ** defined here only to make it easier to avoid clashes with the flags ** above. */ -#define FTS5INDEX_QUERY_SKIPEMPTY 0x0010 -#define FTS5INDEX_QUERY_NOOUTPUT 0x0020 -#define FTS5INDEX_QUERY_SKIPHASH 0x0040 +#define FTS5INDEX_QUERY_SKIPEMPTY 0x0010 +#define FTS5INDEX_QUERY_NOOUTPUT 0x0020 +#define FTS5INDEX_QUERY_SKIPHASH 0x0040 +#define FTS5INDEX_QUERY_NOTOKENDATA 0x0080 +#define FTS5INDEX_QUERY_SCANONETERM 0x0100 /* ** Create/destroy an Fts5Index object. @@ -228466,6 +230507,10 @@ static void *sqlite3Fts5StructureRef(Fts5Index*); static void sqlite3Fts5StructureRelease(void*); static int sqlite3Fts5StructureTest(Fts5Index*, void*); +/* +** Used by xInstToken(): +*/ +static int sqlite3Fts5IterToken(Fts5IndexIter*, i64, int, int, const char**, int*); /* ** Insert or remove data to or from the index. Each time a document is @@ -228543,6 +230588,13 @@ static int sqlite3Fts5IndexLoadConfig(Fts5Index *p); static int sqlite3Fts5IndexGetOrigin(Fts5Index *p, i64 *piOrigin); static int sqlite3Fts5IndexContentlessDelete(Fts5Index *p, i64 iOrigin, i64 iRowid); +static void sqlite3Fts5IndexIterClearTokendata(Fts5IndexIter*); + +/* Used to populate hash tables for xInstToken in detail=none/column mode. */ +static int sqlite3Fts5IndexIterWriteTokendata( + Fts5IndexIter*, const char*, int, i64 iRowid, int iCol, int iOff +); + /* ** End of interface to code in fts5_index.c. **************************************************************************/ @@ -228648,6 +230700,7 @@ static void sqlite3Fts5HashScanNext(Fts5Hash*); static int sqlite3Fts5HashScanEof(Fts5Hash*); static void sqlite3Fts5HashScanEntry(Fts5Hash *, const char **pzTerm, /* OUT: term (nul-terminated) */ + int *pnTerm, /* OUT: Size of term in bytes */ const u8 **ppDoclist, /* OUT: pointer to doclist */ int *pnDoclist /* OUT: size of doclist in bytes */ ); @@ -228774,6 +230827,10 @@ static int sqlite3Fts5ExprClonePhrase(Fts5Expr*, int, Fts5Expr**); static int sqlite3Fts5ExprPhraseCollist(Fts5Expr *, int, const u8 **, int *); +static int sqlite3Fts5ExprQueryToken(Fts5Expr*, int, int, const char**, int*); +static int sqlite3Fts5ExprInstToken(Fts5Expr*, i64, int, int, int, int, const char**, int*); +static void sqlite3Fts5ExprClearTokens(Fts5Expr*); + /******************************************* ** The fts5_expr.c API above this point is used by the other hand-written ** C code in this module. The interfaces below this point are called by @@ -230589,6 +232646,14 @@ static int fts5HighlightCb( } if( iPos==p->iRangeEnd ){ + if( p->bOpen ){ + if( p->iter.iStart>=0 && iPos>=p->iter.iStart ){ + fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff); + p->iOff = iEndOff; + } + fts5HighlightAppend(&rc, p, p->zClose, -1); + p->bOpen = 0; + } fts5HighlightAppend(&rc, p, &p->zIn[p->iOff], iEndOff - p->iOff); p->iOff = iEndOff; } @@ -230622,8 +232687,10 @@ static void fts5HighlightFunction( ctx.zClose = (const char*)sqlite3_value_text(apVal[2]); ctx.iRangeEnd = -1; rc = pApi->xColumnText(pFts, iCol, &ctx.zIn, &ctx.nIn); - - if( ctx.zIn ){ + if( rc==SQLITE_RANGE ){ + sqlite3_result_text(pCtx, "", -1, SQLITE_STATIC); + rc = SQLITE_OK; + }else if( ctx.zIn ){ if( rc==SQLITE_OK ){ rc = fts5CInstIterInit(pApi, pFts, iCol, &ctx.iter); } @@ -231190,6 +233257,7 @@ static void sqlite3Fts5BufferAppendBlob( ){ if( nData ){ if( fts5BufferGrow(pRc, pBuf, nData) ) return; + assert( pBuf->p!=0 ); memcpy(&pBuf->p[pBuf->n], pData, nData); pBuf->n += nData; } @@ -231291,6 +233359,7 @@ static int sqlite3Fts5PoslistNext64( i64 *piOff /* IN/OUT: Current offset */ ){ int i = *pi; + assert( a!=0 || i==0 ); if( i>=n ){ /* EOF */ *piOff = -1; @@ -231298,6 +233367,7 @@ static int sqlite3Fts5PoslistNext64( }else{ i64 iOff = *piOff; u32 iVal; + assert( a!=0 ); fts5FastGetVarint32(a, i, iVal); if( iVal<=1 ){ if( iVal==0 ){ @@ -231929,6 +233999,16 @@ static int fts5ConfigParseSpecial( return rc; } + if( sqlite3_strnicmp("tokendata", zCmd, nCmd)==0 ){ + if( (zArg[0]!='0' && zArg[0]!='1') || zArg[1]!='\0' ){ + *pzErr = sqlite3_mprintf("malformed tokendata=... directive"); + rc = SQLITE_ERROR; + }else{ + pConfig->bTokendata = (zArg[0]=='1'); + } + return rc; + } + *pzErr = sqlite3_mprintf("unrecognized option: \"%.*s\"", nCmd, zCmd); return SQLITE_ERROR; } @@ -232662,7 +234742,9 @@ struct Fts5ExprNode { struct Fts5ExprTerm { u8 bPrefix; /* True for a prefix term */ u8 bFirst; /* True if token must be first in column */ - char *zTerm; /* nul-terminated term */ + char *pTerm; /* Term data */ + int nQueryTerm; /* Effective size of term in bytes */ + int nFullTerm; /* Size of term in bytes incl. tokendata */ Fts5IndexIter *pIter; /* Iterator for this term */ Fts5ExprTerm *pSynonym; /* Pointer to first in list of synonyms */ }; @@ -233529,7 +235611,7 @@ static int fts5ExprNearInitAll( p->pIter = 0; } rc = sqlite3Fts5IndexQuery( - pExpr->pIndex, p->zTerm, (int)strlen(p->zTerm), + pExpr->pIndex, p->pTerm, p->nQueryTerm, (pTerm->bPrefix ? FTS5INDEX_QUERY_PREFIX : 0) | (pExpr->bDesc ? FTS5INDEX_QUERY_DESC : 0), pNear->pColset, @@ -234166,7 +236248,7 @@ static void fts5ExprPhraseFree(Fts5ExprPhrase *pPhrase){ Fts5ExprTerm *pSyn; Fts5ExprTerm *pNext; Fts5ExprTerm *pTerm = &pPhrase->aTerm[i]; - sqlite3_free(pTerm->zTerm); + sqlite3_free(pTerm->pTerm); sqlite3Fts5IterClose(pTerm->pIter); for(pSyn=pTerm->pSynonym; pSyn; pSyn=pNext){ pNext = pSyn->pSynonym; @@ -234264,6 +236346,7 @@ static Fts5ExprNearset *sqlite3Fts5ParseNearset( typedef struct TokenCtx TokenCtx; struct TokenCtx { Fts5ExprPhrase *pPhrase; + Fts5Config *pConfig; int rc; }; @@ -234297,8 +236380,12 @@ static int fts5ParseTokenize( rc = SQLITE_NOMEM; }else{ memset(pSyn, 0, (size_t)nByte); - pSyn->zTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer); - memcpy(pSyn->zTerm, pToken, nToken); + pSyn->pTerm = ((char*)pSyn) + sizeof(Fts5ExprTerm) + sizeof(Fts5Buffer); + pSyn->nFullTerm = pSyn->nQueryTerm = nToken; + if( pCtx->pConfig->bTokendata ){ + pSyn->nQueryTerm = (int)strlen(pSyn->pTerm); + } + memcpy(pSyn->pTerm, pToken, nToken); pSyn->pSynonym = pPhrase->aTerm[pPhrase->nTerm-1].pSynonym; pPhrase->aTerm[pPhrase->nTerm-1].pSynonym = pSyn; } @@ -234323,7 +236410,11 @@ static int fts5ParseTokenize( if( rc==SQLITE_OK ){ pTerm = &pPhrase->aTerm[pPhrase->nTerm++]; memset(pTerm, 0, sizeof(Fts5ExprTerm)); - pTerm->zTerm = sqlite3Fts5Strndup(&rc, pToken, nToken); + pTerm->pTerm = sqlite3Fts5Strndup(&rc, pToken, nToken); + pTerm->nFullTerm = pTerm->nQueryTerm = nToken; + if( pCtx->pConfig->bTokendata && rc==SQLITE_OK ){ + pTerm->nQueryTerm = (int)strlen(pTerm->pTerm); + } } } @@ -234390,6 +236481,7 @@ static Fts5ExprPhrase *sqlite3Fts5ParseTerm( memset(&sCtx, 0, sizeof(TokenCtx)); sCtx.pPhrase = pAppend; + sCtx.pConfig = pConfig; rc = fts5ParseStringFromToken(pToken, &z); if( rc==SQLITE_OK ){ @@ -234437,12 +236529,15 @@ static int sqlite3Fts5ExprClonePhrase( Fts5Expr **ppNew ){ int rc = SQLITE_OK; /* Return code */ - Fts5ExprPhrase *pOrig; /* The phrase extracted from pExpr */ + Fts5ExprPhrase *pOrig = 0; /* The phrase extracted from pExpr */ Fts5Expr *pNew = 0; /* Expression to return via *ppNew */ - TokenCtx sCtx = {0,0}; /* Context object for fts5ParseTokenize */ - - pOrig = pExpr->apExprPhrase[iPhrase]; - pNew = (Fts5Expr*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Expr)); + TokenCtx sCtx = {0,0,0}; /* Context object for fts5ParseTokenize */ + if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){ + rc = SQLITE_RANGE; + }else{ + pOrig = pExpr->apExprPhrase[iPhrase]; + pNew = (Fts5Expr*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5Expr)); + } if( rc==SQLITE_OK ){ pNew->apExprPhrase = (Fts5ExprPhrase**)sqlite3Fts5MallocZero(&rc, sizeof(Fts5ExprPhrase*)); @@ -234455,7 +236550,7 @@ static int sqlite3Fts5ExprClonePhrase( pNew->pRoot->pNear = (Fts5ExprNearset*)sqlite3Fts5MallocZero(&rc, sizeof(Fts5ExprNearset) + sizeof(Fts5ExprPhrase*)); } - if( rc==SQLITE_OK ){ + if( rc==SQLITE_OK && ALWAYS(pOrig!=0) ){ Fts5Colset *pColsetOrig = pOrig->pNode->pNear->pColset; if( pColsetOrig ){ sqlite3_int64 nByte; @@ -234469,26 +236564,27 @@ static int sqlite3Fts5ExprClonePhrase( } } - if( pOrig->nTerm ){ - int i; /* Used to iterate through phrase terms */ - for(i=0; rc==SQLITE_OK && inTerm; i++){ - int tflags = 0; - Fts5ExprTerm *p; - for(p=&pOrig->aTerm[i]; p && rc==SQLITE_OK; p=p->pSynonym){ - const char *zTerm = p->zTerm; - rc = fts5ParseTokenize((void*)&sCtx, tflags, zTerm, (int)strlen(zTerm), - 0, 0); - tflags = FTS5_TOKEN_COLOCATED; - } - if( rc==SQLITE_OK ){ - sCtx.pPhrase->aTerm[i].bPrefix = pOrig->aTerm[i].bPrefix; - sCtx.pPhrase->aTerm[i].bFirst = pOrig->aTerm[i].bFirst; + if( rc==SQLITE_OK ){ + if( pOrig->nTerm ){ + int i; /* Used to iterate through phrase terms */ + sCtx.pConfig = pExpr->pConfig; + for(i=0; rc==SQLITE_OK && inTerm; i++){ + int tflags = 0; + Fts5ExprTerm *p; + for(p=&pOrig->aTerm[i]; p && rc==SQLITE_OK; p=p->pSynonym){ + rc = fts5ParseTokenize((void*)&sCtx,tflags,p->pTerm,p->nFullTerm,0,0); + tflags = FTS5_TOKEN_COLOCATED; + } + if( rc==SQLITE_OK ){ + sCtx.pPhrase->aTerm[i].bPrefix = pOrig->aTerm[i].bPrefix; + sCtx.pPhrase->aTerm[i].bFirst = pOrig->aTerm[i].bFirst; + } } + }else{ + /* This happens when parsing a token or quoted phrase that contains + ** no token characters at all. (e.g ... MATCH '""'). */ + sCtx.pPhrase = sqlite3Fts5MallocZero(&rc, sizeof(Fts5ExprPhrase)); } - }else{ - /* This happens when parsing a token or quoted phrase that contains - ** no token characters at all. (e.g ... MATCH '""'). */ - sCtx.pPhrase = sqlite3Fts5MallocZero(&rc, sizeof(Fts5ExprPhrase)); } if( rc==SQLITE_OK && ALWAYS(sCtx.pPhrase) ){ @@ -234858,11 +236954,13 @@ static Fts5ExprNode *fts5ParsePhraseToAnd( if( parseGrowPhraseArray(pParse) ){ fts5ExprPhraseFree(pPhrase); }else{ + Fts5ExprTerm *p = &pNear->apPhrase[0]->aTerm[ii]; + Fts5ExprTerm *pTo = &pPhrase->aTerm[0]; pParse->apPhrase[pParse->nPhrase++] = pPhrase; pPhrase->nTerm = 1; - pPhrase->aTerm[0].zTerm = sqlite3Fts5Strndup( - &pParse->rc, pNear->apPhrase[0]->aTerm[ii].zTerm, -1 - ); + pTo->pTerm = sqlite3Fts5Strndup(&pParse->rc, p->pTerm, p->nFullTerm); + pTo->nQueryTerm = p->nQueryTerm; + pTo->nFullTerm = p->nFullTerm; pRet->apChild[ii] = sqlite3Fts5ParseNode(pParse, FTS5_STRING, 0, 0, sqlite3Fts5ParseNearset(pParse, 0, pPhrase) ); @@ -235047,16 +237145,17 @@ static char *fts5ExprTermPrint(Fts5ExprTerm *pTerm){ /* Determine the maximum amount of space required. */ for(p=pTerm; p; p=p->pSynonym){ - nByte += (int)strlen(pTerm->zTerm) * 2 + 3 + 2; + nByte += pTerm->nQueryTerm * 2 + 3 + 2; } zQuoted = sqlite3_malloc64(nByte); if( zQuoted ){ int i = 0; for(p=pTerm; p; p=p->pSynonym){ - char *zIn = p->zTerm; + char *zIn = p->pTerm; + char *zEnd = &zIn[p->nQueryTerm]; zQuoted[i++] = '"'; - while( *zIn ){ + while( zInnTerm; iTerm++){ - char *zTerm = pPhrase->aTerm[iTerm].zTerm; - zRet = fts5PrintfAppend(zRet, "%s%s", iTerm==0?"":" ", zTerm); + Fts5ExprTerm *p = &pPhrase->aTerm[iTerm]; + zRet = fts5PrintfAppend(zRet, "%s%.*s", iTerm==0?"":" ", + p->nQueryTerm, p->pTerm + ); if( pPhrase->aTerm[iTerm].bPrefix ){ zRet = fts5PrintfAppend(zRet, "*"); } @@ -235536,6 +237637,17 @@ static int fts5ExprColsetTest(Fts5Colset *pColset, int iCol){ return 0; } +/* +** pToken is a buffer nToken bytes in size that may or may not contain +** an embedded 0x00 byte. If it does, return the number of bytes in +** the buffer before the 0x00. If it does not, return nToken. +*/ +static int fts5QueryTerm(const char *pToken, int nToken){ + int ii; + for(ii=0; iipExpr; int i; + int nQuery = nToken; + i64 iRowid = pExpr->pRoot->iRowid; UNUSED_PARAM2(iUnused1, iUnused2); - if( nToken>FTS5_MAX_TOKEN_SIZE ) nToken = FTS5_MAX_TOKEN_SIZE; + if( nQuery>FTS5_MAX_TOKEN_SIZE ) nQuery = FTS5_MAX_TOKEN_SIZE; + if( pExpr->pConfig->bTokendata ){ + nQuery = fts5QueryTerm(pToken, nQuery); + } if( (tflags & FTS5_TOKEN_COLOCATED)==0 ) p->iOff++; for(i=0; inPhrase; i++){ - Fts5ExprTerm *pTerm; + Fts5ExprTerm *pT; if( p->aPopulator[i].bOk==0 ) continue; - for(pTerm=&pExpr->apExprPhrase[i]->aTerm[0]; pTerm; pTerm=pTerm->pSynonym){ - int nTerm = (int)strlen(pTerm->zTerm); - if( (nTerm==nToken || (nTermbPrefix)) - && memcmp(pTerm->zTerm, pToken, nTerm)==0 + for(pT=&pExpr->apExprPhrase[i]->aTerm[0]; pT; pT=pT->pSynonym){ + if( (pT->nQueryTerm==nQuery || (pT->nQueryTermbPrefix)) + && memcmp(pT->pTerm, pToken, pT->nQueryTerm)==0 ){ int rc = sqlite3Fts5PoslistWriterAppend( &pExpr->apExprPhrase[i]->poslist, &p->aPopulator[i].writer, p->iOff ); + if( rc==SQLITE_OK && pExpr->pConfig->bTokendata && !pT->bPrefix ){ + int iCol = p->iOff>>32; + int iTokOff = p->iOff & 0x7FFFFFFF; + rc = sqlite3Fts5IndexIterWriteTokendata( + pT->pIter, pToken, nToken, iRowid, iCol, iTokOff + ); + } if( rc ) return rc; break; } @@ -235698,6 +237821,83 @@ static int sqlite3Fts5ExprPhraseCollist( return rc; } +/* +** Does the work of the fts5_api.xQueryToken() API method. +*/ +static int sqlite3Fts5ExprQueryToken( + Fts5Expr *pExpr, + int iPhrase, + int iToken, + const char **ppOut, + int *pnOut +){ + Fts5ExprPhrase *pPhrase = 0; + + if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){ + return SQLITE_RANGE; + } + pPhrase = pExpr->apExprPhrase[iPhrase]; + if( iToken<0 || iToken>=pPhrase->nTerm ){ + return SQLITE_RANGE; + } + + *ppOut = pPhrase->aTerm[iToken].pTerm; + *pnOut = pPhrase->aTerm[iToken].nFullTerm; + return SQLITE_OK; +} + +/* +** Does the work of the fts5_api.xInstToken() API method. +*/ +static int sqlite3Fts5ExprInstToken( + Fts5Expr *pExpr, + i64 iRowid, + int iPhrase, + int iCol, + int iOff, + int iToken, + const char **ppOut, + int *pnOut +){ + Fts5ExprPhrase *pPhrase = 0; + Fts5ExprTerm *pTerm = 0; + int rc = SQLITE_OK; + + if( iPhrase<0 || iPhrase>=pExpr->nPhrase ){ + return SQLITE_RANGE; + } + pPhrase = pExpr->apExprPhrase[iPhrase]; + if( iToken<0 || iToken>=pPhrase->nTerm ){ + return SQLITE_RANGE; + } + pTerm = &pPhrase->aTerm[iToken]; + if( pTerm->bPrefix==0 ){ + if( pExpr->pConfig->bTokendata ){ + rc = sqlite3Fts5IterToken( + pTerm->pIter, iRowid, iCol, iOff+iToken, ppOut, pnOut + ); + }else{ + *ppOut = pTerm->pTerm; + *pnOut = pTerm->nFullTerm; + } + } + return rc; +} + +/* +** Clear the token mappings for all Fts5IndexIter objects mannaged by +** the expression passed as the only argument. +*/ +static void sqlite3Fts5ExprClearTokens(Fts5Expr *pExpr){ + int ii; + for(ii=0; iinPhrase; ii++){ + Fts5ExprTerm *pT; + for(pT=&pExpr->apExprPhrase[ii]->aTerm[0]; pT; pT=pT->pSynonym){ + sqlite3Fts5IndexIterClearTokendata(pT->pIter); + } + } +} + /* ** 2014 August 11 ** @@ -235736,10 +237936,15 @@ struct Fts5Hash { /* ** Each entry in the hash table is represented by an object of the -** following type. Each object, its key (a nul-terminated string) and -** its current data are stored in a single memory allocation. The -** key immediately follows the object in memory. The position list -** data immediately follows the key data in memory. +** following type. Each object, its key, and its current data are stored +** in a single memory allocation. The key immediately follows the object +** in memory. The position list data immediately follows the key data +** in memory. +** +** The key is Fts5HashEntry.nKey bytes in size. It consists of a single +** byte identifying the index (either the main term index or a prefix-index), +** followed by the term data. For example: "0token". There is no +** nul-terminator - in this case nKey=6. ** ** The data that follows the key is in a similar, but not identical format ** to the doclist data stored in the database. It is: @@ -235874,8 +238079,7 @@ static int fts5HashResize(Fts5Hash *pHash){ unsigned int iHash; Fts5HashEntry *p = apOld[i]; apOld[i] = p->pHashNext; - iHash = fts5HashKey(nNew, (u8*)fts5EntryKey(p), - (int)strlen(fts5EntryKey(p))); + iHash = fts5HashKey(nNew, (u8*)fts5EntryKey(p), p->nKey); p->pHashNext = apNew[iHash]; apNew[iHash] = p; } @@ -235959,7 +238163,7 @@ static int sqlite3Fts5HashWrite( for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ char *zKey = fts5EntryKey(p); if( zKey[0]==bByte - && p->nKey==nToken + && p->nKey==nToken+1 && memcmp(&zKey[1], pToken, nToken)==0 ){ break; @@ -235989,9 +238193,9 @@ static int sqlite3Fts5HashWrite( zKey[0] = bByte; memcpy(&zKey[1], pToken, nToken); assert( iHash==fts5HashKey(pHash->nSlot, (u8*)zKey, nToken+1) ); - p->nKey = nToken; + p->nKey = nToken+1; zKey[nToken+1] = '\0'; - p->nData = nToken+1 + 1 + sizeof(Fts5HashEntry); + p->nData = nToken+1 + sizeof(Fts5HashEntry); p->pHashNext = pHash->aSlot[iHash]; pHash->aSlot[iHash] = p; pHash->nEntry++; @@ -236108,12 +238312,17 @@ static Fts5HashEntry *fts5HashEntryMerge( *ppOut = p1; p1 = 0; }else{ - int i = 0; char *zKey1 = fts5EntryKey(p1); char *zKey2 = fts5EntryKey(p2); - while( zKey1[i]==zKey2[i] ) i++; + int nMin = MIN(p1->nKey, p2->nKey); - if( ((u8)zKey1[i])>((u8)zKey2[i]) ){ + int cmp = memcmp(zKey1, zKey2, nMin); + if( cmp==0 ){ + cmp = p1->nKey - p2->nKey; + } + assert( cmp!=0 ); + + if( cmp>0 ){ /* p2 is smaller */ *ppOut = p2; ppOut = &p2->pScanNext; @@ -236132,10 +238341,8 @@ static Fts5HashEntry *fts5HashEntryMerge( } /* -** Extract all tokens from hash table iHash and link them into a list -** in sorted order. The hash table is cleared before returning. It is -** the responsibility of the caller to free the elements of the returned -** list. +** Link all tokens from hash table iHash into a list in sorted order. The +** tokens are not removed from the hash table. */ static int fts5HashEntrySort( Fts5Hash *pHash, @@ -236157,7 +238364,7 @@ static int fts5HashEntrySort( Fts5HashEntry *pIter; for(pIter=pHash->aSlot[iSlot]; pIter; pIter=pIter->pHashNext){ if( pTerm==0 - || (pIter->nKey+1>=nTerm && 0==memcmp(fts5EntryKey(pIter), pTerm, nTerm)) + || (pIter->nKey>=nTerm && 0==memcmp(fts5EntryKey(pIter), pTerm, nTerm)) ){ Fts5HashEntry *pEntry = pIter; pEntry->pScanNext = 0; @@ -236196,12 +238403,11 @@ static int sqlite3Fts5HashQuery( for(p=pHash->aSlot[iHash]; p; p=p->pHashNext){ zKey = fts5EntryKey(p); - assert( p->nKey+1==(int)strlen(zKey) ); - if( nTerm==p->nKey+1 && memcmp(zKey, pTerm, nTerm)==0 ) break; + if( nTerm==p->nKey && memcmp(zKey, pTerm, nTerm)==0 ) break; } if( p ){ - int nHashPre = sizeof(Fts5HashEntry) + nTerm + 1; + int nHashPre = sizeof(Fts5HashEntry) + nTerm; int nList = p->nData - nHashPre; u8 *pRet = (u8*)(*ppOut = sqlite3_malloc64(nPre + nList + 10)); if( pRet ){ @@ -236262,19 +238468,22 @@ static int sqlite3Fts5HashScanEof(Fts5Hash *p){ static void sqlite3Fts5HashScanEntry( Fts5Hash *pHash, const char **pzTerm, /* OUT: term (nul-terminated) */ + int *pnTerm, /* OUT: Size of term in bytes */ const u8 **ppDoclist, /* OUT: pointer to doclist */ int *pnDoclist /* OUT: size of doclist in bytes */ ){ Fts5HashEntry *p; if( (p = pHash->pScan) ){ char *zKey = fts5EntryKey(p); - int nTerm = (int)strlen(zKey); + int nTerm = p->nKey; fts5HashAddPoslistSize(pHash, p, 0); *pzTerm = zKey; - *ppDoclist = (const u8*)&zKey[nTerm+1]; - *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm + 1); + *pnTerm = nTerm; + *ppDoclist = (const u8*)&zKey[nTerm]; + *pnDoclist = p->nData - (sizeof(Fts5HashEntry) + nTerm); }else{ *pzTerm = 0; + *pnTerm = 0; *ppDoclist = 0; *pnDoclist = 0; } @@ -236605,6 +238814,9 @@ typedef struct Fts5SegWriter Fts5SegWriter; typedef struct Fts5Structure Fts5Structure; typedef struct Fts5StructureLevel Fts5StructureLevel; typedef struct Fts5StructureSegment Fts5StructureSegment; +typedef struct Fts5TokenDataIter Fts5TokenDataIter; +typedef struct Fts5TokenDataMap Fts5TokenDataMap; +typedef struct Fts5TombstoneArray Fts5TombstoneArray; struct Fts5Data { u8 *p; /* Pointer to buffer containing record */ @@ -236639,6 +238851,7 @@ struct Fts5Index { /* Error state. */ int rc; /* Current error code */ + int flushRc; /* State used by the fts5DataXXX() functions. */ sqlite3_blob *pReader; /* RO incr-blob open on %_data table */ @@ -236647,6 +238860,7 @@ struct Fts5Index { sqlite3_stmt *pIdxWriter; /* "INSERT ... %_idx VALUES(?,?,?,?)" */ sqlite3_stmt *pIdxDeleter; /* "DELETE FROM %_idx WHERE segid=?" */ sqlite3_stmt *pIdxSelect; + sqlite3_stmt *pIdxNextSelect; int nRead; /* Total number of blocks read */ sqlite3_stmt *pDeleteFromIdx; @@ -236800,8 +239014,7 @@ struct Fts5SegIter { Fts5Data *pLeaf; /* Current leaf data */ Fts5Data *pNextLeaf; /* Leaf page (iLeafPgno+1) */ i64 iLeafOffset; /* Byte offset within current leaf */ - Fts5Data **apTombstone; /* Array of tombstone pages */ - int nTombstone; + Fts5TombstoneArray *pTombArray; /* Array of tombstone pages */ /* Next method */ void (*xNext)(Fts5Index*, Fts5SegIter*, int*); @@ -236828,6 +239041,15 @@ struct Fts5SegIter { u8 bDel; /* True if the delete flag is set */ }; +/* +** Array of tombstone pages. Reference counted. +*/ +struct Fts5TombstoneArray { + int nRef; /* Number of pointers to this object */ + int nTombstone; + Fts5Data *apTombstone[1]; /* Array of tombstone pages */ +}; + /* ** Argument is a pointer to an Fts5Data structure that contains a ** leaf page. @@ -236872,9 +239094,16 @@ struct Fts5SegIter { ** poslist: ** Used by sqlite3Fts5IterPoslist() when the poslist needs to be buffered. ** There is no way to tell if this is populated or not. +** +** pColset: +** If not NULL, points to an object containing a set of column indices. +** Only matches that occur in one of these columns will be returned. +** The Fts5Iter does not own the Fts5Colset object, and so it is not +** freed when the iterator is closed - it is owned by the upper layer. */ struct Fts5Iter { Fts5IndexIter base; /* Base class containing output vars */ + Fts5TokenDataIter *pTokenDataIter; Fts5Index *pIndex; /* Index that owns this iterator */ Fts5Buffer poslist; /* Buffer containing current poslist */ @@ -236892,7 +239121,6 @@ struct Fts5Iter { Fts5SegIter aSeg[1]; /* Array of segment iterators */ }; - /* ** An instance of the following type is used to iterate through the contents ** of a doclist-index record. @@ -237810,9 +240038,9 @@ static int fts5DlidxLvlNext(Fts5DlidxLvl *pLvl){ } if( iOffnn ){ - i64 iVal; + u64 iVal; pLvl->iLeafPgno += (iOff - pLvl->iOff) + 1; - iOff += fts5GetVarint(&pData->p[iOff], (u64*)&iVal); + iOff += fts5GetVarint(&pData->p[iOff], &iVal); pLvl->iRowid += iVal; pLvl->iOff = iOff; }else{ @@ -238191,18 +240419,20 @@ static void fts5SegIterSetNext(Fts5Index *p, Fts5SegIter *pIter){ } /* -** Allocate a tombstone hash page array (pIter->apTombstone) for the -** iterator passed as the second argument. If an OOM error occurs, leave -** an error in the Fts5Index object. +** Allocate a tombstone hash page array object (pIter->pTombArray) for +** the iterator passed as the second argument. If an OOM error occurs, +** leave an error in the Fts5Index object. */ static void fts5SegIterAllocTombstone(Fts5Index *p, Fts5SegIter *pIter){ const int nTomb = pIter->pSeg->nPgTombstone; if( nTomb>0 ){ - Fts5Data **apTomb = 0; - apTomb = (Fts5Data**)sqlite3Fts5MallocZero(&p->rc, sizeof(Fts5Data)*nTomb); - if( apTomb ){ - pIter->apTombstone = apTomb; - pIter->nTombstone = nTomb; + int nByte = nTomb * sizeof(Fts5Data*) + sizeof(Fts5TombstoneArray); + Fts5TombstoneArray *pNew; + pNew = (Fts5TombstoneArray*)sqlite3Fts5MallocZero(&p->rc, nByte); + if( pNew ){ + pNew->nTombstone = nTomb; + pNew->nRef = 1; + pIter->pTombArray = pNew; } } } @@ -238459,15 +240689,16 @@ static void fts5SegIterNext_None( }else{ const u8 *pList = 0; const char *zTerm = 0; + int nTerm = 0; int nList; sqlite3Fts5HashScanNext(p->pHash); - sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &pList, &nList); + sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &nTerm, &pList, &nList); if( pList==0 ) goto next_none_eof; pIter->pLeaf->p = (u8*)pList; pIter->pLeaf->nn = nList; pIter->pLeaf->szLeaf = nList; pIter->iEndofDoclist = nList; - sqlite3Fts5BufferSet(&p->rc,&pIter->term, (int)strlen(zTerm), (u8*)zTerm); + sqlite3Fts5BufferSet(&p->rc,&pIter->term, nTerm, (u8*)zTerm); pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid); } @@ -238533,11 +240764,12 @@ static void fts5SegIterNext( }else if( pIter->pSeg==0 ){ const u8 *pList = 0; const char *zTerm = 0; + int nTerm = 0; int nList = 0; assert( (pIter->flags & FTS5_SEGITER_ONETERM) || pbNewTerm ); if( 0==(pIter->flags & FTS5_SEGITER_ONETERM) ){ sqlite3Fts5HashScanNext(p->pHash); - sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &pList, &nList); + sqlite3Fts5HashScanEntry(p->pHash, &zTerm, &nTerm, &pList, &nList); } if( pList==0 ){ fts5DataRelease(pIter->pLeaf); @@ -238547,8 +240779,7 @@ static void fts5SegIterNext( pIter->pLeaf->nn = nList; pIter->pLeaf->szLeaf = nList; pIter->iEndofDoclist = nList+1; - sqlite3Fts5BufferSet(&p->rc, &pIter->term, (int)strlen(zTerm), - (u8*)zTerm); + sqlite3Fts5BufferSet(&p->rc, &pIter->term, nTerm, (u8*)zTerm); pIter->iLeafOffset = fts5GetVarint(pList, (u64*)&pIter->iRowid); *pbNewTerm = 1; } @@ -238934,7 +241165,7 @@ static void fts5SegIterSeekInit( fts5LeafSeek(p, bGe, pIter, pTerm, nTerm); } - if( p->rc==SQLITE_OK && bGe==0 ){ + if( p->rc==SQLITE_OK && (bGe==0 || (flags & FTS5INDEX_QUERY_SCANONETERM)) ){ pIter->flags |= FTS5_SEGITER_ONETERM; if( pIter->pLeaf ){ if( flags & FTS5INDEX_QUERY_DESC ){ @@ -238950,7 +241181,9 @@ static void fts5SegIterSeekInit( } fts5SegIterSetNext(p, pIter); - fts5SegIterAllocTombstone(p, pIter); + if( 0==(flags & FTS5INDEX_QUERY_SCANONETERM) ){ + fts5SegIterAllocTombstone(p, pIter); + } /* Either: ** @@ -238967,6 +241200,79 @@ static void fts5SegIterSeekInit( ); } + +/* +** SQL used by fts5SegIterNextInit() to find the page to open. +*/ +static sqlite3_stmt *fts5IdxNextStmt(Fts5Index *p){ + if( p->pIdxNextSelect==0 ){ + Fts5Config *pConfig = p->pConfig; + fts5IndexPrepareStmt(p, &p->pIdxNextSelect, sqlite3_mprintf( + "SELECT pgno FROM '%q'.'%q_idx' WHERE " + "segid=? AND term>? ORDER BY term ASC LIMIT 1", + pConfig->zDb, pConfig->zName + )); + + } + return p->pIdxNextSelect; +} + +/* +** This is similar to fts5SegIterSeekInit(), except that it initializes +** the segment iterator to point to the first term following the page +** with pToken/nToken on it. +*/ +static void fts5SegIterNextInit( + Fts5Index *p, + const char *pTerm, int nTerm, + Fts5StructureSegment *pSeg, /* Description of segment */ + Fts5SegIter *pIter /* Object to populate */ +){ + int iPg = -1; /* Page of segment to open */ + int bDlidx = 0; + sqlite3_stmt *pSel = 0; /* SELECT to find iPg */ + + pSel = fts5IdxNextStmt(p); + if( pSel ){ + assert( p->rc==SQLITE_OK ); + sqlite3_bind_int(pSel, 1, pSeg->iSegid); + sqlite3_bind_blob(pSel, 2, pTerm, nTerm, SQLITE_STATIC); + + if( sqlite3_step(pSel)==SQLITE_ROW ){ + i64 val = sqlite3_column_int64(pSel, 0); + iPg = (int)(val>>1); + bDlidx = (val & 0x0001); + } + p->rc = sqlite3_reset(pSel); + sqlite3_bind_null(pSel, 2); + if( p->rc ) return; + } + + memset(pIter, 0, sizeof(*pIter)); + pIter->pSeg = pSeg; + pIter->flags |= FTS5_SEGITER_ONETERM; + if( iPg>=0 ){ + pIter->iLeafPgno = iPg - 1; + fts5SegIterNextPage(p, pIter); + fts5SegIterSetNext(p, pIter); + } + if( pIter->pLeaf ){ + const u8 *a = pIter->pLeaf->p; + int iTermOff = 0; + + pIter->iPgidxOff = pIter->pLeaf->szLeaf; + pIter->iPgidxOff += fts5GetVarint32(&a[pIter->iPgidxOff], iTermOff); + pIter->iLeafOffset = iTermOff; + fts5SegIterLoadTerm(p, pIter, 0); + fts5SegIterLoadNPos(p, pIter); + if( bDlidx ) fts5SegIterLoadDlidx(p, pIter); + + assert( p->rc!=SQLITE_OK || + fts5BufferCompareBlob(&pIter->term, (const u8*)pTerm, nTerm)>0 + ); + } +} + /* ** Initialize the object pIter to point to term pTerm/nTerm within the ** in-memory hash table. If there is no such term in the hash-table, the @@ -238993,14 +241299,21 @@ static void fts5SegIterHashInit( const u8 *pList = 0; p->rc = sqlite3Fts5HashScanInit(p->pHash, (const char*)pTerm, nTerm); - sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &pList, &nList); - n = (z ? (int)strlen((const char*)z) : 0); + sqlite3Fts5HashScanEntry(p->pHash, (const char**)&z, &n, &pList, &nList); if( pList ){ pLeaf = fts5IdxMalloc(p, sizeof(Fts5Data)); if( pLeaf ){ pLeaf->p = (u8*)pList; } } + + /* The call to sqlite3Fts5HashScanInit() causes the hash table to + ** fill the size field of all existing position lists. This means they + ** can no longer be appended to. Since the only scenario in which they + ** can be appended to is if the previous operation on this table was + ** a DELETE, by clearing the Fts5Index.bDelete flag we can avoid this + ** possibility altogether. */ + p->bDelete = 0; }else{ p->rc = sqlite3Fts5HashQuery(p->pHash, sizeof(Fts5Data), (const char*)pTerm, nTerm, (void**)&pLeaf, &nList @@ -239045,6 +241358,23 @@ static void fts5IndexFreeArray(Fts5Data **ap, int n){ } } +/* +** Decrement the ref-count of the object passed as the only argument. If it +** reaches 0, free it and its contents. +*/ +static void fts5TombstoneArrayDelete(Fts5TombstoneArray *p){ + if( p ){ + p->nRef--; + if( p->nRef<=0 ){ + int ii; + for(ii=0; iinTombstone; ii++){ + fts5DataRelease(p->apTombstone[ii]); + } + sqlite3_free(p); + } + } +} + /* ** Zero the iterator passed as the only argument. */ @@ -239052,7 +241382,7 @@ static void fts5SegIterClear(Fts5SegIter *pIter){ fts5BufferFree(&pIter->term); fts5DataRelease(pIter->pLeaf); fts5DataRelease(pIter->pNextLeaf); - fts5IndexFreeArray(pIter->apTombstone, pIter->nTombstone); + fts5TombstoneArrayDelete(pIter->pTombArray); fts5DlidxIterFree(pIter->pDlidx); sqlite3_free(pIter->aRowidOffset); memset(pIter, 0, sizeof(Fts5SegIter)); @@ -239297,7 +241627,6 @@ static void fts5SegIterNextFrom( }while( p->rc==SQLITE_OK ); } - /* ** Free the iterator object passed as the second argument. */ @@ -239442,24 +241771,25 @@ static int fts5IndexTombstoneQuery( static int fts5MultiIterIsDeleted(Fts5Iter *pIter){ int iFirst = pIter->aFirst[1].iFirst; Fts5SegIter *pSeg = &pIter->aSeg[iFirst]; + Fts5TombstoneArray *pArray = pSeg->pTombArray; - if( pSeg->pLeaf && pSeg->nTombstone ){ + if( pSeg->pLeaf && pArray ){ /* Figure out which page the rowid might be present on. */ - int iPg = ((u64)pSeg->iRowid) % pSeg->nTombstone; + int iPg = ((u64)pSeg->iRowid) % pArray->nTombstone; assert( iPg>=0 ); /* If tombstone hash page iPg has not yet been loaded from the ** database, load it now. */ - if( pSeg->apTombstone[iPg]==0 ){ - pSeg->apTombstone[iPg] = fts5DataRead(pIter->pIndex, + if( pArray->apTombstone[iPg]==0 ){ + pArray->apTombstone[iPg] = fts5DataRead(pIter->pIndex, FTS5_TOMBSTONE_ROWID(pSeg->pSeg->iSegid, iPg) ); - if( pSeg->apTombstone[iPg]==0 ) return 0; + if( pArray->apTombstone[iPg]==0 ) return 0; } return fts5IndexTombstoneQuery( - pSeg->apTombstone[iPg], - pSeg->nTombstone, + pArray->apTombstone[iPg], + pArray->nTombstone, pSeg->iRowid ); } @@ -239998,6 +242328,32 @@ static void fts5IterSetOutputCb(int *pRc, Fts5Iter *pIter){ } } +/* +** All the component segment-iterators of pIter have been set up. This +** functions finishes setup for iterator pIter itself. +*/ +static void fts5MultiIterFinishSetup(Fts5Index *p, Fts5Iter *pIter){ + int iIter; + for(iIter=pIter->nSeg-1; iIter>0; iIter--){ + int iEq; + if( (iEq = fts5MultiIterDoCompare(pIter, iIter)) ){ + Fts5SegIter *pSeg = &pIter->aSeg[iEq]; + if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0); + fts5MultiIterAdvanced(p, pIter, iEq, iIter); + } + } + fts5MultiIterSetEof(pIter); + fts5AssertMultiIterSetup(p, pIter); + + if( (pIter->bSkipEmpty && fts5MultiIterIsEmpty(p, pIter)) + || fts5MultiIterIsDeleted(pIter) + ){ + fts5MultiIterNext(p, pIter, 0, 0); + }else if( pIter->base.bEof==0 ){ + Fts5SegIter *pSeg = &pIter->aSeg[pIter->aFirst[1].iFirst]; + pIter->xSetOutputs(pIter, pSeg); + } +} /* ** Allocate a new Fts5Iter object. @@ -240079,31 +242435,12 @@ static void fts5MultiIterNew( assert( iIter==nSeg ); } - /* If the above was successful, each component iterators now points + /* If the above was successful, each component iterator now points ** to the first entry in its segment. In this case initialize the ** aFirst[] array. Or, if an error has occurred, free the iterator ** object and set the output variable to NULL. */ if( p->rc==SQLITE_OK ){ - for(iIter=pNew->nSeg-1; iIter>0; iIter--){ - int iEq; - if( (iEq = fts5MultiIterDoCompare(pNew, iIter)) ){ - Fts5SegIter *pSeg = &pNew->aSeg[iEq]; - if( p->rc==SQLITE_OK ) pSeg->xNext(p, pSeg, 0); - fts5MultiIterAdvanced(p, pNew, iEq, iIter); - } - } - fts5MultiIterSetEof(pNew); - fts5AssertMultiIterSetup(p, pNew); - - if( (pNew->bSkipEmpty && fts5MultiIterIsEmpty(p, pNew)) - || fts5MultiIterIsDeleted(pNew) - ){ - fts5MultiIterNext(p, pNew, 0, 0); - }else if( pNew->base.bEof==0 ){ - Fts5SegIter *pSeg = &pNew->aSeg[pNew->aFirst[1].iFirst]; - pNew->xSetOutputs(pNew, pSeg); - } - + fts5MultiIterFinishSetup(p, pNew); }else{ fts5MultiIterFree(pNew); *ppOut = 0; @@ -240128,7 +242465,6 @@ static void fts5MultiIterNew2( pNew = fts5MultiIterAlloc(p, 2); if( pNew ){ Fts5SegIter *pIter = &pNew->aSeg[1]; - pIter->flags = FTS5_SEGITER_ONETERM; if( pData->szLeaf>0 ){ pIter->pLeaf = pData; @@ -240276,6 +242612,7 @@ static void fts5IndexDiscardData(Fts5Index *p){ sqlite3Fts5HashClear(p->pHash); p->nPendingData = 0; p->nPendingRow = 0; + p->flushRc = SQLITE_OK; } p->nContentlessDelete = 0; } @@ -240491,7 +242828,7 @@ static void fts5WriteDlidxAppend( } if( pDlidx->bPrevValid ){ - iVal = iRowid - pDlidx->iPrev; + iVal = (u64)iRowid - (u64)pDlidx->iPrev; }else{ i64 iPgno = (i==0 ? pWriter->writer.pgno : pDlidx[-1].pgno); assert( pDlidx->buf.n==0 ); @@ -240678,7 +243015,7 @@ static void fts5WriteAppendPoslistData( const u8 *a = aData; int n = nData; - assert( p->pConfig->pgsz>0 ); + assert( p->pConfig->pgsz>0 || p->rc!=SQLITE_OK ); while( p->rc==SQLITE_OK && (pPage->buf.n + pPage->pgidx.n + n)>=p->pConfig->pgsz ){ @@ -241411,18 +243748,24 @@ static void fts5DoSecureDelete( iOff = iStart; - /* Set variable bLastInDoclist to true if this entry happens to be - ** the last rowid in the doclist for its term. */ + /* If the position-list for the entry being removed flows over past + ** the end of this page, delete the portion of the position-list on the + ** next page and beyond. + ** + ** Set variable bLastInDoclist to true if this entry happens + ** to be the last rowid in the doclist for its term. */ + if( iNextOff>=iPgIdx ){ + int pgno = pSeg->iLeafPgno+1; + fts5SecureDeleteOverflow(p, pSeg->pSeg, pgno, &bLastInDoclist); + iNextOff = iPgIdx; + } + if( pSeg->bDel==0 ){ - if( iNextOff>=iPgIdx ){ - int pgno = pSeg->iLeafPgno+1; - fts5SecureDeleteOverflow(p, pSeg->pSeg, pgno, &bLastInDoclist); - iNextOff = iPgIdx; - }else{ + if( iNextOff!=iPgIdx ){ /* Loop through the page-footer. If iNextOff (offset of the ** entry following the one we are removing) is equal to the ** offset of a key on this page, then the entry is the last - ** in its doclist. */ + ** in its doclist. */ int iKeyOff = 0; for(iIdx=0; iIdxrc!=SQLITE_OK ) break; @@ -241716,7 +244058,7 @@ static void fts5FlushOneHash(Fts5Index *p){ if( bSecureDelete ){ if( eDetail==FTS5_DETAIL_NONE ){ if( iOffrc!=SQLITE_OK || pDoclist[iOff]==0x01 ){ iOff++; continue; @@ -241852,6 +244194,10 @@ static void fts5FlushOneHash(Fts5Index *p){ */ static void fts5IndexFlush(Fts5Index *p){ /* Unless it is empty, flush the hash table to disk */ + if( p->flushRc ){ + p->rc = p->flushRc; + return; + } if( p->nPendingData || p->nContentlessDelete ){ assert( p->pHash ); fts5FlushOneHash(p); @@ -241860,6 +244206,8 @@ static void fts5IndexFlush(Fts5Index *p){ p->nPendingData = 0; p->nPendingRow = 0; p->nContentlessDelete = 0; + }else if( p->nPendingData || p->nContentlessDelete ){ + p->flushRc = p->rc; } } } @@ -241938,8 +244286,9 @@ static int sqlite3Fts5IndexOptimize(Fts5Index *p){ assert( p->rc==SQLITE_OK ); fts5IndexFlush(p); - assert( p->nContentlessDelete==0 ); + assert( p->rc!=SQLITE_OK || p->nContentlessDelete==0 ); pStruct = fts5StructureRead(p); + assert( p->rc!=SQLITE_OK || pStruct!=0 ); fts5StructureInvalidate(p); if( pStruct ){ @@ -242345,7 +244694,7 @@ static void fts5SetupPrefixIter( u8 *pToken, /* Buffer containing prefix to match */ int nToken, /* Size of buffer pToken in bytes */ Fts5Colset *pColset, /* Restrict matches to these columns */ - Fts5Iter **ppIter /* OUT: New iterator */ + Fts5Iter **ppIter /* OUT: New iterator */ ){ Fts5Structure *pStruct; Fts5Buffer *aBuf; @@ -242366,8 +244715,9 @@ static void fts5SetupPrefixIter( aBuf = (Fts5Buffer*)fts5IdxMalloc(p, sizeof(Fts5Buffer)*nBuf); pStruct = fts5StructureRead(p); + assert( p->rc!=SQLITE_OK || (aBuf && pStruct) ); - if( aBuf && pStruct ){ + if( p->rc==SQLITE_OK ){ const int flags = FTS5INDEX_QUERY_SCAN | FTS5INDEX_QUERY_SKIPEMPTY | FTS5INDEX_QUERY_NOOUTPUT; @@ -242379,6 +244729,12 @@ static void fts5SetupPrefixIter( int bNewTerm = 1; memset(&doclist, 0, sizeof(doclist)); + + /* If iIdx is non-zero, then it is the number of a prefix-index for + ** prefixes 1 character longer than the prefix being queried for. That + ** index contains all the doclists required, except for the one + ** corresponding to the prefix itself. That one is extracted from the + ** main term index here. */ if( iIdx!=0 ){ int dummy = 0; const int f2 = FTS5INDEX_QUERY_SKIPEMPTY|FTS5INDEX_QUERY_NOOUTPUT; @@ -242402,6 +244758,7 @@ static void fts5SetupPrefixIter( pToken[0] = FTS5_MAIN_PREFIX + iIdx; fts5MultiIterNew(p, pStruct, flags, pColset, pToken, nToken, -1, 0, &p1); fts5IterSetOutputCb(&p->rc, p1); + for( /* no-op */ ; fts5MultiIterEof(p, p1)==0; fts5MultiIterNext2(p, p1, &bNewTerm) @@ -242417,7 +244774,6 @@ static void fts5SetupPrefixIter( } if( p1->base.nData==0 ) continue; - if( p1->base.iRowid<=iLastRowid && doclist.n>0 ){ for(i=0; p->rc==SQLITE_OK && doclist.n; i++){ int i1 = i*nMerge; @@ -242456,7 +244812,7 @@ static void fts5SetupPrefixIter( } fts5MultiIterFree(p1); - pData = fts5IdxMalloc(p, sizeof(Fts5Data)+doclist.n+FTS5_DATA_ZERO_PADDING); + pData = fts5IdxMalloc(p, sizeof(*pData)+doclist.n+FTS5_DATA_ZERO_PADDING); if( pData ){ pData->p = (u8*)&pData[1]; pData->nn = pData->szLeaf = doclist.n; @@ -242599,6 +244955,7 @@ static int sqlite3Fts5IndexClose(Fts5Index *p){ sqlite3_finalize(p->pIdxWriter); sqlite3_finalize(p->pIdxDeleter); sqlite3_finalize(p->pIdxSelect); + sqlite3_finalize(p->pIdxNextSelect); sqlite3_finalize(p->pDataVersion); sqlite3_finalize(p->pDeleteFromIdx); sqlite3Fts5HashFree(p->pHash); @@ -242694,6 +245051,454 @@ static int sqlite3Fts5IndexWrite( return rc; } +/* +** pToken points to a buffer of size nToken bytes containing a search +** term, including the index number at the start, used on a tokendata=1 +** table. This function returns true if the term in buffer pBuf matches +** token pToken/nToken. +*/ +static int fts5IsTokendataPrefix( + Fts5Buffer *pBuf, + const u8 *pToken, + int nToken +){ + return ( + pBuf->n>=nToken + && 0==memcmp(pBuf->p, pToken, nToken) + && (pBuf->n==nToken || pBuf->p[nToken]==0x00) + ); +} + +/* +** Ensure the segment-iterator passed as the only argument points to EOF. +*/ +static void fts5SegIterSetEOF(Fts5SegIter *pSeg){ + fts5DataRelease(pSeg->pLeaf); + pSeg->pLeaf = 0; +} + +/* +** Usually, a tokendata=1 iterator (struct Fts5TokenDataIter) accumulates an +** array of these for each row it visits. Or, for an iterator used by an +** "ORDER BY rank" query, it accumulates an array of these for the entire +** query. +** +** Each instance in the array indicates the iterator (and therefore term) +** associated with position iPos of rowid iRowid. This is used by the +** xInstToken() API. +*/ +struct Fts5TokenDataMap { + i64 iRowid; /* Row this token is located in */ + i64 iPos; /* Position of token */ + int iIter; /* Iterator token was read from */ +}; + +/* +** An object used to supplement Fts5Iter for tokendata=1 iterators. +*/ +struct Fts5TokenDataIter { + int nIter; + int nIterAlloc; + + int nMap; + int nMapAlloc; + Fts5TokenDataMap *aMap; + + Fts5PoslistReader *aPoslistReader; + int *aPoslistToIter; + Fts5Iter *apIter[1]; +}; + +/* +** This function appends iterator pAppend to Fts5TokenDataIter pIn and +** returns the result. +*/ +static Fts5TokenDataIter *fts5AppendTokendataIter( + Fts5Index *p, /* Index object (for error code) */ + Fts5TokenDataIter *pIn, /* Current Fts5TokenDataIter struct */ + Fts5Iter *pAppend /* Append this iterator */ +){ + Fts5TokenDataIter *pRet = pIn; + + if( p->rc==SQLITE_OK ){ + if( pIn==0 || pIn->nIter==pIn->nIterAlloc ){ + int nAlloc = pIn ? pIn->nIterAlloc*2 : 16; + int nByte = nAlloc * sizeof(Fts5Iter*) + sizeof(Fts5TokenDataIter); + Fts5TokenDataIter *pNew = (Fts5TokenDataIter*)sqlite3_realloc(pIn, nByte); + + if( pNew==0 ){ + p->rc = SQLITE_NOMEM; + }else{ + if( pIn==0 ) memset(pNew, 0, nByte); + pRet = pNew; + pNew->nIterAlloc = nAlloc; + } + } + } + if( p->rc ){ + sqlite3Fts5IterClose((Fts5IndexIter*)pAppend); + }else{ + pRet->apIter[pRet->nIter++] = pAppend; + } + assert( pRet==0 || pRet->nIter<=pRet->nIterAlloc ); + + return pRet; +} + +/* +** Delete an Fts5TokenDataIter structure and its contents. +*/ +static void fts5TokendataIterDelete(Fts5TokenDataIter *pSet){ + if( pSet ){ + int ii; + for(ii=0; iinIter; ii++){ + fts5MultiIterFree(pSet->apIter[ii]); + } + sqlite3_free(pSet->aPoslistReader); + sqlite3_free(pSet->aMap); + sqlite3_free(pSet); + } +} + +/* +** Append a mapping to the token-map belonging to object pT. +*/ +static void fts5TokendataIterAppendMap( + Fts5Index *p, + Fts5TokenDataIter *pT, + int iIter, + i64 iRowid, + i64 iPos +){ + if( p->rc==SQLITE_OK ){ + if( pT->nMap==pT->nMapAlloc ){ + int nNew = pT->nMapAlloc ? pT->nMapAlloc*2 : 64; + int nByte = nNew * sizeof(Fts5TokenDataMap); + Fts5TokenDataMap *aNew; + + aNew = (Fts5TokenDataMap*)sqlite3_realloc(pT->aMap, nByte); + if( aNew==0 ){ + p->rc = SQLITE_NOMEM; + return; + } + + pT->aMap = aNew; + pT->nMapAlloc = nNew; + } + + pT->aMap[pT->nMap].iRowid = iRowid; + pT->aMap[pT->nMap].iPos = iPos; + pT->aMap[pT->nMap].iIter = iIter; + pT->nMap++; + } +} + +/* +** The iterator passed as the only argument must be a tokendata=1 iterator +** (pIter->pTokenDataIter!=0). This function sets the iterator output +** variables (pIter->base.*) according to the contents of the current +** row. +*/ +static void fts5IterSetOutputsTokendata(Fts5Iter *pIter){ + int ii; + int nHit = 0; + i64 iRowid = SMALLEST_INT64; + int iMin = 0; + + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + + pIter->base.nData = 0; + pIter->base.pData = 0; + + for(ii=0; iinIter; ii++){ + Fts5Iter *p = pT->apIter[ii]; + if( p->base.bEof==0 ){ + if( nHit==0 || p->base.iRowidbase.iRowid; + nHit = 1; + pIter->base.pData = p->base.pData; + pIter->base.nData = p->base.nData; + iMin = ii; + }else if( p->base.iRowid==iRowid ){ + nHit++; + } + } + } + + if( nHit==0 ){ + pIter->base.bEof = 1; + }else{ + int eDetail = pIter->pIndex->pConfig->eDetail; + pIter->base.bEof = 0; + pIter->base.iRowid = iRowid; + + if( nHit==1 && eDetail==FTS5_DETAIL_FULL ){ + fts5TokendataIterAppendMap(pIter->pIndex, pT, iMin, iRowid, -1); + }else + if( nHit>1 && eDetail!=FTS5_DETAIL_NONE ){ + int nReader = 0; + int nByte = 0; + i64 iPrev = 0; + + /* Allocate array of iterators if they are not already allocated. */ + if( pT->aPoslistReader==0 ){ + pT->aPoslistReader = (Fts5PoslistReader*)sqlite3Fts5MallocZero( + &pIter->pIndex->rc, + pT->nIter * (sizeof(Fts5PoslistReader) + sizeof(int)) + ); + if( pT->aPoslistReader==0 ) return; + pT->aPoslistToIter = (int*)&pT->aPoslistReader[pT->nIter]; + } + + /* Populate an iterator for each poslist that will be merged */ + for(ii=0; iinIter; ii++){ + Fts5Iter *p = pT->apIter[ii]; + if( iRowid==p->base.iRowid ){ + pT->aPoslistToIter[nReader] = ii; + sqlite3Fts5PoslistReaderInit( + p->base.pData, p->base.nData, &pT->aPoslistReader[nReader++] + ); + nByte += p->base.nData; + } + } + + /* Ensure the output buffer is large enough */ + if( fts5BufferGrow(&pIter->pIndex->rc, &pIter->poslist, nByte+nHit*10) ){ + return; + } + + /* Ensure the token-mapping is large enough */ + if( eDetail==FTS5_DETAIL_FULL && pT->nMapAlloc<(pT->nMap + nByte) ){ + int nNew = (pT->nMapAlloc + nByte) * 2; + Fts5TokenDataMap *aNew = (Fts5TokenDataMap*)sqlite3_realloc( + pT->aMap, nNew*sizeof(Fts5TokenDataMap) + ); + if( aNew==0 ){ + pIter->pIndex->rc = SQLITE_NOMEM; + return; + } + pT->aMap = aNew; + pT->nMapAlloc = nNew; + } + + pIter->poslist.n = 0; + + while( 1 ){ + i64 iMinPos = LARGEST_INT64; + + /* Find smallest position */ + iMin = 0; + for(ii=0; iiaPoslistReader[ii]; + if( pReader->bEof==0 ){ + if( pReader->iPosiPos; + iMin = ii; + } + } + } + + /* If all readers were at EOF, break out of the loop. */ + if( iMinPos==LARGEST_INT64 ) break; + + sqlite3Fts5PoslistSafeAppend(&pIter->poslist, &iPrev, iMinPos); + sqlite3Fts5PoslistReaderNext(&pT->aPoslistReader[iMin]); + + if( eDetail==FTS5_DETAIL_FULL ){ + pT->aMap[pT->nMap].iPos = iMinPos; + pT->aMap[pT->nMap].iIter = pT->aPoslistToIter[iMin]; + pT->aMap[pT->nMap].iRowid = iRowid; + pT->nMap++; + } + } + + pIter->base.pData = pIter->poslist.p; + pIter->base.nData = pIter->poslist.n; + } + } +} + +/* +** The iterator passed as the only argument must be a tokendata=1 iterator +** (pIter->pTokenDataIter!=0). This function advances the iterator. If +** argument bFrom is false, then the iterator is advanced to the next +** entry. Or, if bFrom is true, it is advanced to the first entry with +** a rowid of iFrom or greater. +*/ +static void fts5TokendataIterNext(Fts5Iter *pIter, int bFrom, i64 iFrom){ + int ii; + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + + for(ii=0; iinIter; ii++){ + Fts5Iter *p = pT->apIter[ii]; + if( p->base.bEof==0 + && (p->base.iRowid==pIter->base.iRowid || (bFrom && p->base.iRowidpIndex, p, bFrom, iFrom); + while( bFrom && p->base.bEof==0 + && p->base.iRowidpIndex->rc==SQLITE_OK + ){ + fts5MultiIterNext(p->pIndex, p, 0, 0); + } + } + } + + fts5IterSetOutputsTokendata(pIter); +} + +/* +** If the segment-iterator passed as the first argument is at EOF, then +** set pIter->term to a copy of buffer pTerm. +*/ +static void fts5TokendataSetTermIfEof(Fts5Iter *pIter, Fts5Buffer *pTerm){ + if( pIter && pIter->aSeg[0].pLeaf==0 ){ + fts5BufferSet(&pIter->pIndex->rc, &pIter->aSeg[0].term, pTerm->n, pTerm->p); + } +} + +/* +** This function sets up an iterator to use for a non-prefix query on a +** tokendata=1 table. +*/ +static Fts5Iter *fts5SetupTokendataIter( + Fts5Index *p, /* FTS index to query */ + const u8 *pToken, /* Buffer containing query term */ + int nToken, /* Size of buffer pToken in bytes */ + Fts5Colset *pColset /* Colset to filter on */ +){ + Fts5Iter *pRet = 0; + Fts5TokenDataIter *pSet = 0; + Fts5Structure *pStruct = 0; + const int flags = FTS5INDEX_QUERY_SCANONETERM | FTS5INDEX_QUERY_SCAN; + + Fts5Buffer bSeek = {0, 0, 0}; + Fts5Buffer *pSmall = 0; + + fts5IndexFlush(p); + pStruct = fts5StructureRead(p); + + while( p->rc==SQLITE_OK ){ + Fts5Iter *pPrev = pSet ? pSet->apIter[pSet->nIter-1] : 0; + Fts5Iter *pNew = 0; + Fts5SegIter *pNewIter = 0; + Fts5SegIter *pPrevIter = 0; + + int iLvl, iSeg, ii; + + pNew = fts5MultiIterAlloc(p, pStruct->nSegment); + if( pSmall ){ + fts5BufferSet(&p->rc, &bSeek, pSmall->n, pSmall->p); + fts5BufferAppendBlob(&p->rc, &bSeek, 1, (const u8*)"\0"); + }else{ + fts5BufferSet(&p->rc, &bSeek, nToken, pToken); + } + if( p->rc ){ + sqlite3Fts5IterClose((Fts5IndexIter*)pNew); + break; + } + + pNewIter = &pNew->aSeg[0]; + pPrevIter = (pPrev ? &pPrev->aSeg[0] : 0); + for(iLvl=0; iLvlnLevel; iLvl++){ + for(iSeg=pStruct->aLevel[iLvl].nSeg-1; iSeg>=0; iSeg--){ + Fts5StructureSegment *pSeg = &pStruct->aLevel[iLvl].aSeg[iSeg]; + int bDone = 0; + + if( pPrevIter ){ + if( fts5BufferCompare(pSmall, &pPrevIter->term) ){ + memcpy(pNewIter, pPrevIter, sizeof(Fts5SegIter)); + memset(pPrevIter, 0, sizeof(Fts5SegIter)); + bDone = 1; + }else if( pPrevIter->iEndofDoclist>pPrevIter->pLeaf->szLeaf ){ + fts5SegIterNextInit(p,(const char*)bSeek.p,bSeek.n-1,pSeg,pNewIter); + bDone = 1; + } + } + + if( bDone==0 ){ + fts5SegIterSeekInit(p, bSeek.p, bSeek.n, flags, pSeg, pNewIter); + } + + if( pPrevIter ){ + if( pPrevIter->pTombArray ){ + pNewIter->pTombArray = pPrevIter->pTombArray; + pNewIter->pTombArray->nRef++; + } + }else{ + fts5SegIterAllocTombstone(p, pNewIter); + } + + pNewIter++; + if( pPrevIter ) pPrevIter++; + if( p->rc ) break; + } + } + fts5TokendataSetTermIfEof(pPrev, pSmall); + + pNew->bSkipEmpty = 1; + pNew->pColset = pColset; + fts5IterSetOutputCb(&p->rc, pNew); + + /* Loop through all segments in the new iterator. Find the smallest + ** term that any segment-iterator points to. Iterator pNew will be + ** used for this term. Also, set any iterator that points to a term that + ** does not match pToken/nToken to point to EOF */ + pSmall = 0; + for(ii=0; iinSeg; ii++){ + Fts5SegIter *pII = &pNew->aSeg[ii]; + if( 0==fts5IsTokendataPrefix(&pII->term, pToken, nToken) ){ + fts5SegIterSetEOF(pII); + } + if( pII->pLeaf && (!pSmall || fts5BufferCompare(pSmall, &pII->term)>0) ){ + pSmall = &pII->term; + } + } + + /* If pSmall is still NULL at this point, then the new iterator does + ** not point to any terms that match the query. So delete it and break + ** out of the loop - all required iterators have been collected. */ + if( pSmall==0 ){ + sqlite3Fts5IterClose((Fts5IndexIter*)pNew); + break; + } + + /* Append this iterator to the set and continue. */ + pSet = fts5AppendTokendataIter(p, pSet, pNew); + } + + if( p->rc==SQLITE_OK && pSet ){ + int ii; + for(ii=0; iinIter; ii++){ + Fts5Iter *pIter = pSet->apIter[ii]; + int iSeg; + for(iSeg=0; iSegnSeg; iSeg++){ + pIter->aSeg[iSeg].flags |= FTS5_SEGITER_ONETERM; + } + fts5MultiIterFinishSetup(p, pIter); + } + } + + if( p->rc==SQLITE_OK ){ + pRet = fts5MultiIterAlloc(p, 0); + } + if( pRet ){ + pRet->pTokenDataIter = pSet; + if( pSet ){ + fts5IterSetOutputsTokendata(pRet); + }else{ + pRet->base.bEof = 1; + } + }else{ + fts5TokendataIterDelete(pSet); + } + + fts5StructureRelease(pStruct); + fts5BufferFree(&bSeek); + return pRet; +} + + /* ** Open a new iterator to iterate though all rowid that match the ** specified token or token prefix. @@ -242715,8 +245520,13 @@ static int sqlite3Fts5IndexQuery( if( sqlite3Fts5BufferSize(&p->rc, &buf, nToken+1)==0 ){ int iIdx = 0; /* Index to search */ int iPrefixIdx = 0; /* +1 prefix index */ + int bTokendata = pConfig->bTokendata; if( nToken>0 ) memcpy(&buf.p[1], pToken, nToken); + if( flags & (FTS5INDEX_QUERY_NOTOKENDATA|FTS5INDEX_QUERY_SCAN) ){ + bTokendata = 0; + } + /* Figure out which index to search and set iIdx accordingly. If this ** is a prefix query for which there is no prefix index, set iIdx to ** greater than pConfig->nPrefix to indicate that the query will be @@ -242742,7 +245552,10 @@ static int sqlite3Fts5IndexQuery( } } - if( iIdx<=pConfig->nPrefix ){ + if( bTokendata && iIdx==0 ){ + buf.p[0] = '0'; + pRet = fts5SetupTokendataIter(p, buf.p, nToken+1, pColset); + }else if( iIdx<=pConfig->nPrefix ){ /* Straight index lookup */ Fts5Structure *pStruct = fts5StructureRead(p); buf.p[0] = (u8)(FTS5_MAIN_PREFIX + iIdx); @@ -242789,7 +245602,11 @@ static int sqlite3Fts5IndexQuery( static int sqlite3Fts5IterNext(Fts5IndexIter *pIndexIter){ Fts5Iter *pIter = (Fts5Iter*)pIndexIter; assert( pIter->pIndex->rc==SQLITE_OK ); - fts5MultiIterNext(pIter->pIndex, pIter, 0, 0); + if( pIter->pTokenDataIter ){ + fts5TokendataIterNext(pIter, 0, 0); + }else{ + fts5MultiIterNext(pIter->pIndex, pIter, 0, 0); + } return fts5IndexReturn(pIter->pIndex); } @@ -242822,7 +245639,11 @@ static int sqlite3Fts5IterNextScan(Fts5IndexIter *pIndexIter){ */ static int sqlite3Fts5IterNextFrom(Fts5IndexIter *pIndexIter, i64 iMatch){ Fts5Iter *pIter = (Fts5Iter*)pIndexIter; - fts5MultiIterNextFrom(pIter->pIndex, pIter, iMatch); + if( pIter->pTokenDataIter ){ + fts5TokendataIterNext(pIter, 1, iMatch); + }else{ + fts5MultiIterNextFrom(pIter->pIndex, pIter, iMatch); + } return fts5IndexReturn(pIter->pIndex); } @@ -242837,6 +245658,99 @@ static const char *sqlite3Fts5IterTerm(Fts5IndexIter *pIndexIter, int *pn){ return (z ? &z[1] : 0); } +/* +** This is used by xInstToken() to access the token at offset iOff, column +** iCol of row iRowid. The token is returned via output variables *ppOut +** and *pnOut. The iterator passed as the first argument must be a tokendata=1 +** iterator (pIter->pTokenDataIter!=0). +*/ +static int sqlite3Fts5IterToken( + Fts5IndexIter *pIndexIter, + i64 iRowid, + int iCol, + int iOff, + const char **ppOut, int *pnOut +){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + Fts5TokenDataMap *aMap = pT->aMap; + i64 iPos = (((i64)iCol)<<32) + iOff; + + int i1 = 0; + int i2 = pT->nMap; + int iTest = 0; + + while( i2>i1 ){ + iTest = (i1 + i2) / 2; + + if( aMap[iTest].iRowidiRowid ){ + i2 = iTest; + }else{ + if( aMap[iTest].iPosiPos ){ + i2 = iTest; + }else{ + break; + } + } + } + + if( i2>i1 ){ + Fts5Iter *pMap = pT->apIter[aMap[iTest].iIter]; + *ppOut = (const char*)pMap->aSeg[0].term.p+1; + *pnOut = pMap->aSeg[0].term.n-1; + } + + return SQLITE_OK; +} + +/* +** Clear any existing entries from the token-map associated with the +** iterator passed as the only argument. +*/ +static void sqlite3Fts5IndexIterClearTokendata(Fts5IndexIter *pIndexIter){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + if( pIter && pIter->pTokenDataIter ){ + pIter->pTokenDataIter->nMap = 0; + } +} + +/* +** Set a token-mapping for the iterator passed as the first argument. This +** is used in detail=column or detail=none mode when a token is requested +** using the xInstToken() API. In this case the caller tokenizers the +** current row and configures the token-mapping via multiple calls to this +** function. +*/ +static int sqlite3Fts5IndexIterWriteTokendata( + Fts5IndexIter *pIndexIter, + const char *pToken, int nToken, + i64 iRowid, int iCol, int iOff +){ + Fts5Iter *pIter = (Fts5Iter*)pIndexIter; + Fts5TokenDataIter *pT = pIter->pTokenDataIter; + Fts5Index *p = pIter->pIndex; + int ii; + + assert( p->pConfig->eDetail!=FTS5_DETAIL_FULL ); + assert( pIter->pTokenDataIter ); + + for(ii=0; iinIter; ii++){ + Fts5Buffer *pTerm = &pT->apIter[ii]->aSeg[0].term; + if( nToken==pTerm->n-1 && memcmp(pToken, pTerm->p+1, nToken)==0 ) break; + } + if( iinIter ){ + fts5TokendataIterAppendMap(p, pT, ii, iRowid, (((i64)iCol)<<32) + iOff); + } + return fts5IndexReturn(p); +} + /* ** Close an iterator opened by an earlier call to sqlite3Fts5IndexQuery(). */ @@ -242844,6 +245758,7 @@ static void sqlite3Fts5IterClose(Fts5IndexIter *pIndexIter){ if( pIndexIter ){ Fts5Iter *pIter = (Fts5Iter*)pIndexIter; Fts5Index *pIndex = pIter->pIndex; + fts5TokendataIterDelete(pIter->pTokenDataIter); fts5MultiIterFree(pIter); sqlite3Fts5IndexCloseReader(pIndex); } @@ -243351,7 +246266,9 @@ static int fts5QueryCksum( int eDetail = p->pConfig->eDetail; u64 cksum = *pCksum; Fts5IndexIter *pIter = 0; - int rc = sqlite3Fts5IndexQuery(p, z, n, flags, 0, &pIter); + int rc = sqlite3Fts5IndexQuery( + p, z, n, (flags | FTS5INDEX_QUERY_NOTOKENDATA), 0, &pIter + ); while( rc==SQLITE_OK && ALWAYS(pIter!=0) && 0==sqlite3Fts5IterEof(pIter) ){ i64 rowid = pIter->iRowid; @@ -243518,7 +246435,7 @@ static void fts5IndexIntegrityCheckEmpty( } static void fts5IntegrityCheckPgidx(Fts5Index *p, Fts5Data *pLeaf){ - int iTermOff = 0; + i64 iTermOff = 0; int ii; Fts5Buffer buf1 = {0,0,0}; @@ -243527,7 +246444,7 @@ static void fts5IntegrityCheckPgidx(Fts5Index *p, Fts5Data *pLeaf){ ii = pLeaf->szLeaf; while( iinn && p->rc==SQLITE_OK ){ int res; - int iOff; + i64 iOff; int nIncr; ii += fts5GetVarint32(&pLeaf->p[ii], nIncr); @@ -244049,6 +246966,24 @@ static void fts5DecodeRowidList( } #endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */ +#if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG) +static void fts5BufferAppendTerm(int *pRc, Fts5Buffer *pBuf, Fts5Buffer *pTerm){ + int ii; + fts5BufferGrow(pRc, pBuf, pTerm->n*2 + 1); + if( *pRc==SQLITE_OK ){ + for(ii=0; iin; ii++){ + if( pTerm->p[ii]==0x00 ){ + pBuf->p[pBuf->n++] = '\\'; + pBuf->p[pBuf->n++] = '0'; + }else{ + pBuf->p[pBuf->n++] = pTerm->p[ii]; + } + } + pBuf->p[pBuf->n] = 0x00; + } +} +#endif /* SQLITE_TEST || SQLITE_FTS5_DEBUG */ + #if defined(SQLITE_TEST) || defined(SQLITE_FTS5_DEBUG) /* ** The implementation of user-defined scalar function fts5_decode(). @@ -244156,9 +247091,8 @@ static void fts5DecodeFunction( iOff += fts5GetVarint32(&a[iOff], nAppend); term.n = nKeep; fts5BufferAppendBlob(&rc, &term, nAppend, &a[iOff]); - sqlite3Fts5BufferAppendPrintf( - &rc, &s, " term=%.*s", term.n, (const char*)term.p - ); + sqlite3Fts5BufferAppendPrintf(&rc, &s, " term="); + fts5BufferAppendTerm(&rc, &s, &term); iOff += nAppend; /* Figure out where the doclist for this term ends */ @@ -244266,9 +247200,8 @@ static void fts5DecodeFunction( fts5BufferAppendBlob(&rc, &term, nByte, &a[iOff]); iOff += nByte; - sqlite3Fts5BufferAppendPrintf( - &rc, &s, " term=%.*s", term.n, (const char*)term.p - ); + sqlite3Fts5BufferAppendPrintf(&rc, &s, " term="); + fts5BufferAppendTerm(&rc, &s, &term); iOff += fts5DecodeDoclist(&rc, &s, &a[iOff], iEnd-iOff); } @@ -244743,7 +247676,7 @@ struct Fts5FullTable { Fts5Global *pGlobal; /* Global (connection wide) data */ Fts5Cursor *pSortCsr; /* Sort data from this cursor */ int iSavepoint; /* Successful xSavepoint()+1 */ - int bInSavepoint; + #ifdef SQLITE_DEBUG struct Fts5TransactionState ts; #endif @@ -245281,12 +248214,15 @@ static int fts5BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){ } idxStr[iIdxStr] = '\0'; - /* Set idxFlags flags for the ORDER BY clause */ + /* Set idxFlags flags for the ORDER BY clause + ** + ** Note that tokendata=1 tables cannot currently handle "ORDER BY rowid DESC". + */ if( pInfo->nOrderBy==1 ){ int iSort = pInfo->aOrderBy[0].iColumn; if( iSort==(pConfig->nCol+1) && bSeenMatch ){ idxFlags |= FTS5_BI_ORDER_RANK; - }else if( iSort==-1 ){ + }else if( iSort==-1 && (!pInfo->aOrderBy[0].desc || !pConfig->bTokendata) ){ idxFlags |= FTS5_BI_ORDER_ROWID; } if( BitFlagTest(idxFlags, FTS5_BI_ORDER_RANK|FTS5_BI_ORDER_ROWID) ){ @@ -245538,6 +248474,16 @@ static int fts5NextMethod(sqlite3_vtab_cursor *pCursor){ ); assert( !CsrFlagTest(pCsr, FTS5CSR_EOF) ); + /* If this cursor uses FTS5_PLAN_MATCH and this is a tokendata=1 table, + ** clear any token mappings accumulated at the fts5_index.c level. In + ** other cases, specifically FTS5_PLAN_SOURCE and FTS5_PLAN_SORTED_MATCH, + ** we need to retain the mappings for the entire query. */ + if( pCsr->ePlan==FTS5_PLAN_MATCH + && ((Fts5Table*)pCursor->pVtab)->pConfig->bTokendata + ){ + sqlite3Fts5ExprClearTokens(pCsr->pExpr); + } + if( pCsr->ePlan<3 ){ int bSkip = 0; if( (rc = fts5CursorReseek(pCsr, &bSkip)) || bSkip ) return rc; @@ -246198,7 +249144,10 @@ static int fts5SpecialInsert( }else if( 0==sqlite3_stricmp("flush", zCmd) ){ rc = sqlite3Fts5FlushToDisk(&pTab->p); }else{ - rc = sqlite3Fts5IndexLoadConfig(pTab->p.pIndex); + rc = sqlite3Fts5FlushToDisk(&pTab->p); + if( rc==SQLITE_OK ){ + rc = sqlite3Fts5IndexLoadConfig(pTab->p.pIndex); + } if( rc==SQLITE_OK ){ rc = sqlite3Fts5ConfigSetValue(pTab->p.pConfig, zCmd, pVal, &bError); } @@ -246523,7 +249472,10 @@ static int fts5ApiColumnText( ){ int rc = SQLITE_OK; Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; - if( fts5IsContentless((Fts5FullTable*)(pCsr->base.pVtab)) + Fts5Table *pTab = (Fts5Table*)(pCsr->base.pVtab); + if( iCol<0 || iCol>=pTab->pConfig->nCol ){ + rc = SQLITE_RANGE; + }else if( fts5IsContentless((Fts5FullTable*)(pCsr->base.pVtab)) || pCsr->ePlan==FTS5_PLAN_SPECIAL ){ *pz = 0; @@ -246548,8 +249500,9 @@ static int fts5CsrPoslist( int rc = SQLITE_OK; int bLive = (pCsr->pSorter==0); - if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_POSLIST) ){ - + if( iPhrase<0 || iPhrase>=sqlite3Fts5ExprPhraseCount(pCsr->pExpr) ){ + rc = SQLITE_RANGE; + }else if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_POSLIST) ){ if( pConfig->eDetail!=FTS5_DETAIL_FULL ){ Fts5PoslistPopulator *aPopulator; int i; @@ -246573,15 +249526,21 @@ static int fts5CsrPoslist( CsrFlagClear(pCsr, FTS5CSR_REQUIRE_POSLIST); } - if( pCsr->pSorter && pConfig->eDetail==FTS5_DETAIL_FULL ){ - Fts5Sorter *pSorter = pCsr->pSorter; - int i1 = (iPhrase==0 ? 0 : pSorter->aIdx[iPhrase-1]); - *pn = pSorter->aIdx[iPhrase] - i1; - *pa = &pSorter->aPoslist[i1]; + if( rc==SQLITE_OK ){ + if( pCsr->pSorter && pConfig->eDetail==FTS5_DETAIL_FULL ){ + Fts5Sorter *pSorter = pCsr->pSorter; + int i1 = (iPhrase==0 ? 0 : pSorter->aIdx[iPhrase-1]); + *pn = pSorter->aIdx[iPhrase] - i1; + *pa = &pSorter->aPoslist[i1]; + }else{ + *pn = sqlite3Fts5ExprPoslist(pCsr->pExpr, iPhrase, pa); + } }else{ - *pn = sqlite3Fts5ExprPoslist(pCsr->pExpr, iPhrase, pa); + *pa = 0; + *pn = 0; } + return rc; } @@ -246688,12 +249647,6 @@ static int fts5ApiInst( ){ if( iIdx<0 || iIdx>=pCsr->nInstCount ){ rc = SQLITE_RANGE; -#if 0 - }else if( fts5IsOffsetless((Fts5Table*)pCsr->base.pVtab) ){ - *piPhrase = pCsr->aInst[iIdx*3]; - *piCol = pCsr->aInst[iIdx*3 + 2]; - *piOff = -1; -#endif }else{ *piPhrase = pCsr->aInst[iIdx*3]; *piCol = pCsr->aInst[iIdx*3 + 1]; @@ -246948,13 +249901,56 @@ static int fts5ApiPhraseFirstColumn( return rc; } +/* +** xQueryToken() API implemenetation. +*/ +static int fts5ApiQueryToken( + Fts5Context* pCtx, + int iPhrase, + int iToken, + const char **ppOut, + int *pnOut +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + return sqlite3Fts5ExprQueryToken(pCsr->pExpr, iPhrase, iToken, ppOut, pnOut); +} + +/* +** xInstToken() API implemenetation. +*/ +static int fts5ApiInstToken( + Fts5Context *pCtx, + int iIdx, + int iToken, + const char **ppOut, int *pnOut +){ + Fts5Cursor *pCsr = (Fts5Cursor*)pCtx; + int rc = SQLITE_OK; + if( CsrFlagTest(pCsr, FTS5CSR_REQUIRE_INST)==0 + || SQLITE_OK==(rc = fts5CacheInstArray(pCsr)) + ){ + if( iIdx<0 || iIdx>=pCsr->nInstCount ){ + rc = SQLITE_RANGE; + }else{ + int iPhrase = pCsr->aInst[iIdx*3]; + int iCol = pCsr->aInst[iIdx*3 + 1]; + int iOff = pCsr->aInst[iIdx*3 + 2]; + i64 iRowid = fts5CursorRowid(pCsr); + rc = sqlite3Fts5ExprInstToken( + pCsr->pExpr, iRowid, iPhrase, iCol, iOff, iToken, ppOut, pnOut + ); + } + } + return rc; +} + static int fts5ApiQueryPhrase(Fts5Context*, int, void*, int(*)(const Fts5ExtensionApi*, Fts5Context*, void*) ); static const Fts5ExtensionApi sFts5Api = { - 2, /* iVersion */ + 3, /* iVersion */ fts5ApiUserData, fts5ApiColumnCount, fts5ApiRowCount, @@ -246974,6 +249970,8 @@ static const Fts5ExtensionApi sFts5Api = { fts5ApiPhraseNext, fts5ApiPhraseFirstColumn, fts5ApiPhraseNextColumn, + fts5ApiQueryToken, + fts5ApiInstToken }; /* @@ -247240,9 +250238,7 @@ static int fts5RenameMethod( ){ int rc; Fts5FullTable *pTab = (Fts5FullTable*)pVtab; - pTab->bInSavepoint = 1; rc = sqlite3Fts5StorageRename(pTab->pStorage, zName); - pTab->bInSavepoint = 0; return rc; } @@ -247259,26 +250255,12 @@ static int sqlite3Fts5FlushToDisk(Fts5Table *pTab){ static int fts5SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){ Fts5FullTable *pTab = (Fts5FullTable*)pVtab; int rc = SQLITE_OK; - char *zSql = 0; - fts5CheckTransactionState(pTab, FTS5_SAVEPOINT, iSavepoint); - if( pTab->bInSavepoint==0 ){ - zSql = sqlite3_mprintf("INSERT INTO %Q.%Q(%Q) VALUES('flush')", - pTab->p.pConfig->zDb, pTab->p.pConfig->zName, pTab->p.pConfig->zName - ); - if( zSql ){ - pTab->bInSavepoint = 1; - rc = sqlite3_exec(pTab->p.pConfig->db, zSql, 0, 0, 0); - pTab->bInSavepoint = 0; - sqlite3_free(zSql); - }else{ - rc = SQLITE_NOMEM; - } - if( rc==SQLITE_OK ){ - pTab->iSavepoint = iSavepoint+1; - } + fts5CheckTransactionState(pTab, FTS5_SAVEPOINT, iSavepoint); + rc = sqlite3Fts5FlushToDisk((Fts5Table*)pVtab); + if( rc==SQLITE_OK ){ + pTab->iSavepoint = iSavepoint+1; } - return rc; } @@ -247310,8 +250292,8 @@ static int fts5RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){ int rc = SQLITE_OK; fts5CheckTransactionState(pTab, FTS5_ROLLBACKTO, iSavepoint); fts5TripCursors(pTab); - pTab->p.pConfig->pgsz = 0; if( (iSavepoint+1)<=pTab->iSavepoint ){ + pTab->p.pConfig->pgsz = 0; rc = sqlite3Fts5StorageRollback(pTab->pStorage); } return rc; @@ -247516,7 +250498,7 @@ static void fts5SourceIdFunc( ){ assert( nArg==0 ); UNUSED_PARAM2(nArg, apUnused); - sqlite3_result_text(pCtx, "fts5: 2023-11-01 11:23:50 17129ba1ff7f0daf37100ee82d507aef7827cf38de1866e2633096ae6ad81301", -1, SQLITE_TRANSIENT); + sqlite3_result_text(pCtx, "fts5: 2024-01-15 17:01:13 1066602b2b1976fe58b5150777cced894af17c803e068f5918390d6915b46e1d", -1, SQLITE_TRANSIENT); } /* @@ -247539,7 +250521,7 @@ static int fts5ShadowName(const char *zName){ ** if anything is found amiss. Return a NULL pointer if everything is ** OK. */ -static int fts5Integrity( +static int fts5IntegrityMethod( sqlite3_vtab *pVtab, /* the FTS5 virtual table to check */ const char *zSchema, /* Name of schema in which this table lives */ const char *zTabname, /* Name of the table itself */ @@ -247597,7 +250579,7 @@ static int fts5Init(sqlite3 *db){ /* xRelease */ fts5ReleaseMethod, /* xRollbackTo */ fts5RollbackToMethod, /* xShadowName */ fts5ShadowName, - /* xIntegrity */ fts5Integrity + /* xIntegrity */ fts5IntegrityMethod }; int rc; @@ -248363,7 +251345,7 @@ static int sqlite3Fts5StorageRebuild(Fts5Storage *p){ } if( rc==SQLITE_OK ){ - rc = fts5StorageGetStmt(p, FTS5_STMT_SCAN, &pScan, 0); + rc = fts5StorageGetStmt(p, FTS5_STMT_SCAN, &pScan, pConfig->pzErrmsg); } while( rc==SQLITE_OK && SQLITE_ROW==sqlite3_step(pScan) ){ @@ -249150,6 +252132,12 @@ static const unsigned char sqlite3Utf8Trans1[] = { #endif /* ifndef SQLITE_AMALGAMATION */ +#define FTS5_SKIP_UTF8(zIn) { \ + if( ((unsigned char)(*(zIn++)))>=0xc0 ){ \ + while( (((unsigned char)*zIn) & 0xc0)==0x80 ){ zIn++; } \ + } \ +} + typedef struct Unicode61Tokenizer Unicode61Tokenizer; struct Unicode61Tokenizer { unsigned char aTokenChar[128]; /* ASCII range token characters */ @@ -250185,6 +253173,7 @@ static int fts5PorterTokenize( typedef struct TrigramTokenizer TrigramTokenizer; struct TrigramTokenizer { int bFold; /* True to fold to lower-case */ + int iFoldParam; /* Parameter to pass to Fts5UnicodeFold() */ }; /* @@ -250211,6 +253200,7 @@ static int fts5TriCreate( }else{ int i; pNew->bFold = 1; + pNew->iFoldParam = 0; for(i=0; rc==SQLITE_OK && ibFold = (zArg[0]=='0'); } + }else if( 0==sqlite3_stricmp(azArg[i], "remove_diacritics") ){ + if( (zArg[0]!='0' && zArg[0]!='1' && zArg[0]!='2') || zArg[1] ){ + rc = SQLITE_ERROR; + }else{ + pNew->iFoldParam = (zArg[0]!='0') ? 2 : 0; + } }else{ rc = SQLITE_ERROR; } } + + if( pNew->iFoldParam!=0 && pNew->bFold==0 ){ + rc = SQLITE_ERROR; + } + if( rc!=SQLITE_OK ){ fts5TriDelete((Fts5Tokenizer*)pNew); pNew = 0; @@ -250245,40 +253246,62 @@ static int fts5TriTokenize( TrigramTokenizer *p = (TrigramTokenizer*)pTok; int rc = SQLITE_OK; char aBuf[32]; + char *zOut = aBuf; + int ii; const unsigned char *zIn = (const unsigned char*)pText; const unsigned char *zEof = &zIn[nText]; u32 iCode; + int aStart[3]; /* Input offset of each character in aBuf[] */ UNUSED_PARAM(unusedFlags); - while( 1 ){ - char *zOut = aBuf; - int iStart = zIn - (const unsigned char*)pText; - const unsigned char *zNext; - - READ_UTF8(zIn, zEof, iCode); - if( iCode==0 ) break; - zNext = zIn; - if( zInbFold ) iCode = sqlite3Fts5UnicodeFold(iCode, 0); - WRITE_UTF8(zOut, iCode); + + /* Populate aBuf[] with the characters for the first trigram. */ + for(ii=0; ii<3; ii++){ + do { + aStart[ii] = zIn - (const unsigned char*)pText; READ_UTF8(zIn, zEof, iCode); - if( iCode==0 ) break; - }else{ - break; - } - if( zInbFold ) iCode = sqlite3Fts5UnicodeFold(iCode, 0); - WRITE_UTF8(zOut, iCode); + if( iCode==0 ) return SQLITE_OK; + if( p->bFold ) iCode = sqlite3Fts5UnicodeFold(iCode, p->iFoldParam); + }while( iCode==0 ); + WRITE_UTF8(zOut, iCode); + } + + /* At the start of each iteration of this loop: + ** + ** aBuf: Contains 3 characters. The 3 characters of the next trigram. + ** zOut: Points to the byte following the last character in aBuf. + ** aStart[3]: Contains the byte offset in the input text corresponding + ** to the start of each of the three characters in the buffer. + */ + assert( zIn<=zEof ); + while( 1 ){ + int iNext; /* Start of character following current tri */ + const char *z1; + + /* Read characters from the input up until the first non-diacritic */ + do { + iNext = zIn - (const unsigned char*)pText; READ_UTF8(zIn, zEof, iCode); if( iCode==0 ) break; - if( p->bFold ) iCode = sqlite3Fts5UnicodeFold(iCode, 0); - WRITE_UTF8(zOut, iCode); - }else{ - break; - } - rc = xToken(pCtx, 0, aBuf, zOut-aBuf, iStart, iStart + zOut-aBuf); - if( rc!=SQLITE_OK ) break; - zIn = zNext; + if( p->bFold ) iCode = sqlite3Fts5UnicodeFold(iCode, p->iFoldParam); + }while( iCode==0 ); + + /* Pass the current trigram back to fts5 */ + rc = xToken(pCtx, 0, aBuf, zOut-aBuf, aStart[0], iNext); + if( iCode==0 || rc!=SQLITE_OK ) break; + + /* Remove the first character from buffer aBuf[]. Append the character + ** with codepoint iCode. */ + z1 = aBuf; + FTS5_SKIP_UTF8(z1); + memmove(aBuf, z1, zOut - z1); + zOut -= (z1 - aBuf); + WRITE_UTF8(zOut, iCode); + + /* Update the aStart[] array */ + aStart[0] = aStart[1]; + aStart[1] = aStart[2]; + aStart[2] = iNext; } return rc; @@ -250301,7 +253324,9 @@ static int sqlite3Fts5TokenizerPattern( ){ if( xCreate==fts5TriCreate ){ TrigramTokenizer *p = (TrigramTokenizer*)pTok; - return p->bFold ? FTS5_PATTERN_LIKE : FTS5_PATTERN_GLOB; + if( p->iFoldParam==0 ){ + return p->bFold ? FTS5_PATTERN_LIKE : FTS5_PATTERN_GLOB; + } } return FTS5_PATTERN_NONE; } @@ -252090,7 +255115,7 @@ static int fts5VocabFilterMethod( if( pEq ){ zTerm = (const char *)sqlite3_value_text(pEq); nTerm = sqlite3_value_bytes(pEq); - f = 0; + f = FTS5INDEX_QUERY_NOTOKENDATA; }else{ if( pGe ){ zTerm = (const char *)sqlite3_value_text(pGe); diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.h b/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.h index 4eff82d9d..ff58e2e73 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.h +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3-binding.h @@ -147,9 +147,9 @@ extern "C" { ** [sqlite3_libversion_number()], [sqlite3_sourceid()], ** [sqlite_version()] and [sqlite_source_id()]. */ -#define SQLITE_VERSION "3.44.0" -#define SQLITE_VERSION_NUMBER 3044000 -#define SQLITE_SOURCE_ID "2023-11-01 11:23:50 17129ba1ff7f0daf37100ee82d507aef7827cf38de1866e2633096ae6ad81301" +#define SQLITE_VERSION "3.45.0" +#define SQLITE_VERSION_NUMBER 3045000 +#define SQLITE_SOURCE_ID "2024-01-15 17:01:13 1066602b2b1976fe58b5150777cced894af17c803e068f5918390d6915b46e1d" /* ** CAPI3REF: Run-Time Library Version Numbers @@ -3955,15 +3955,17 @@ SQLITE_API void sqlite3_free_filename(sqlite3_filename); ** ** ** ^The sqlite3_errmsg() and sqlite3_errmsg16() return English-language -** text that describes the error, as either UTF-8 or UTF-16 respectively. +** text that describes the error, as either UTF-8 or UTF-16 respectively, +** or NULL if no error message is available. ** (See how SQLite handles [invalid UTF] for exceptions to this rule.) ** ^(Memory to hold the error message string is managed internally. ** The application does not need to worry about freeing the result. ** However, the error string might be overwritten or deallocated by ** subsequent calls to other SQLite interface functions.)^ ** -** ^The sqlite3_errstr() interface returns the English-language text -** that describes the [result code], as UTF-8. +** ^The sqlite3_errstr(E) interface returns the English-language text +** that describes the [result code] E, as UTF-8, or NULL if E is not an +** result code for which a text error message is available. ** ^(Memory to hold the error message string is managed internally ** and must not be freed by the application)^. ** @@ -5574,13 +5576,27 @@ SQLITE_API int sqlite3_create_window_function( ** ** ** [[SQLITE_SUBTYPE]]
    SQLITE_SUBTYPE
    -** The SQLITE_SUBTYPE flag indicates to SQLite that a function may call +** The SQLITE_SUBTYPE flag indicates to SQLite that a function might call ** [sqlite3_value_subtype()] to inspect the sub-types of its arguments. -** Specifying this flag makes no difference for scalar or aggregate user -** functions. However, if it is not specified for a user-defined window -** function, then any sub-types belonging to arguments passed to the window -** function may be discarded before the window function is called (i.e. -** sqlite3_value_subtype() will always return 0). +** This flag instructs SQLite to omit some corner-case optimizations that +** might disrupt the operation of the [sqlite3_value_subtype()] function, +** causing it to return zero rather than the correct subtype(). +** SQL functions that invokes [sqlite3_value_subtype()] should have this +** property. If the SQLITE_SUBTYPE property is omitted, then the return +** value from [sqlite3_value_subtype()] might sometimes be zero even though +** a non-zero subtype was specified by the function argument expression. +** +** [[SQLITE_RESULT_SUBTYPE]]
    SQLITE_RESULT_SUBTYPE
    +** The SQLITE_RESULT_SUBTYPE flag indicates to SQLite that a function might call +** [sqlite3_result_subtype()] to cause a sub-type to be associated with its +** result. +** Every function that invokes [sqlite3_result_subtype()] should have this +** property. If it does not, then the call to [sqlite3_result_subtype()] +** might become a no-op if the function is used as term in an +** [expression index]. On the other hand, SQL functions that never invoke +** [sqlite3_result_subtype()] should avoid setting this property, as the +** purpose of this property is to disable certain optimizations that are +** incompatible with subtypes. **
    ** */ @@ -5588,6 +5604,7 @@ SQLITE_API int sqlite3_create_window_function( #define SQLITE_DIRECTONLY 0x000080000 #define SQLITE_SUBTYPE 0x000100000 #define SQLITE_INNOCUOUS 0x000200000 +#define SQLITE_RESULT_SUBTYPE 0x001000000 /* ** CAPI3REF: Deprecated Functions @@ -5784,6 +5801,12 @@ SQLITE_API int sqlite3_value_encoding(sqlite3_value*); ** information can be used to pass a limited amount of context from ** one SQL function to another. Use the [sqlite3_result_subtype()] ** routine to set the subtype for the return value of an SQL function. +** +** Every [application-defined SQL function] that invoke this interface +** should include the [SQLITE_SUBTYPE] property in the text +** encoding argument when the function is [sqlite3_create_function|registered]. +** If the [SQLITE_SUBTYPE] property is omitted, then sqlite3_value_subtype() +** might return zero instead of the upstream subtype in some corner cases. */ SQLITE_API unsigned int sqlite3_value_subtype(sqlite3_value*); @@ -5914,14 +5937,22 @@ SQLITE_API sqlite3 *sqlite3_context_db_handle(sqlite3_context*); **
  • ^(when sqlite3_set_auxdata() is invoked again on the same ** parameter)^, or **
  • ^(during the original sqlite3_set_auxdata() call when a memory -** allocation error occurs.)^ +** allocation error occurs.)^ +**
  • ^(during the original sqlite3_set_auxdata() call if the function +** is evaluated during query planning instead of during query execution, +** as sometimes happens with [SQLITE_ENABLE_STAT4].)^ ** -** Note the last bullet in particular. The destructor X in +** Note the last two bullets in particular. The destructor X in ** sqlite3_set_auxdata(C,N,P,X) might be called immediately, before the ** sqlite3_set_auxdata() interface even returns. Hence sqlite3_set_auxdata() ** should be called near the end of the function implementation and the ** function implementation should not make any use of P after -** sqlite3_set_auxdata() has been called. +** sqlite3_set_auxdata() has been called. Furthermore, a call to +** sqlite3_get_auxdata() that occurs immediately after a corresponding call +** to sqlite3_set_auxdata() might still return NULL if an out-of-memory +** condition occurred during the sqlite3_set_auxdata() call or if the +** function is being evaluated during query planning rather than during +** query execution. ** ** ^(In practice, auxiliary data is preserved between function calls for ** function parameters that are compile-time constants, including literal @@ -6195,6 +6226,20 @@ SQLITE_API int sqlite3_result_zeroblob64(sqlite3_context*, sqlite3_uint64 n); ** higher order bits are discarded. ** The number of subtype bytes preserved by SQLite might increase ** in future releases of SQLite. +** +** Every [application-defined SQL function] that invokes this interface +** should include the [SQLITE_RESULT_SUBTYPE] property in its +** text encoding argument when the SQL function is +** [sqlite3_create_function|registered]. If the [SQLITE_RESULT_SUBTYPE] +** property is omitted from the function that invokes sqlite3_result_subtype(), +** then in some cases the sqlite3_result_subtype() might fail to set +** the result subtype. +** +** If SQLite is compiled with -DSQLITE_STRICT_SUBTYPE=1, then any +** SQL function that invokes the sqlite3_result_subtype() interface +** and that does not have the SQLITE_RESULT_SUBTYPE property will raise +** an error. Future versions of SQLite might enable -DSQLITE_STRICT_SUBTYPE=1 +** by default. */ SQLITE_API void sqlite3_result_subtype(sqlite3_context*,unsigned int); @@ -7995,9 +8040,11 @@ SQLITE_API int sqlite3_vfs_unregister(sqlite3_vfs*); ** ** ^(Some systems (for example, Windows 95) do not support the operation ** implemented by sqlite3_mutex_try(). On those systems, sqlite3_mutex_try() -** will always return SQLITE_BUSY. The SQLite core only ever uses -** sqlite3_mutex_try() as an optimization so this is acceptable -** behavior.)^ +** will always return SQLITE_BUSY. In most cases the SQLite core only uses +** sqlite3_mutex_try() as an optimization, so this is acceptable +** behavior. The exceptions are unix builds that set the +** SQLITE_ENABLE_SETLK_TIMEOUT build option. In that case a working +** sqlite3_mutex_try() is required.)^ ** ** ^The sqlite3_mutex_leave() routine exits a mutex that was ** previously entered by the same thread. The behavior @@ -8256,6 +8303,7 @@ SQLITE_API int sqlite3_test_control(int op, ...); #define SQLITE_TESTCTRL_ASSERT 12 #define SQLITE_TESTCTRL_ALWAYS 13 #define SQLITE_TESTCTRL_RESERVE 14 /* NOT USED */ +#define SQLITE_TESTCTRL_JSON_SELFCHECK 14 #define SQLITE_TESTCTRL_OPTIMIZATIONS 15 #define SQLITE_TESTCTRL_ISKEYWORD 16 /* NOT USED */ #define SQLITE_TESTCTRL_SCRATCHMALLOC 17 /* NOT USED */ @@ -12769,8 +12817,11 @@ struct Fts5PhraseIter { ** created with the "columnsize=0" option. ** ** xColumnText: -** This function attempts to retrieve the text of column iCol of the -** current document. If successful, (*pz) is set to point to a buffer +** If parameter iCol is less than zero, or greater than or equal to the +** number of columns in the table, SQLITE_RANGE is returned. +** +** Otherwise, this function attempts to retrieve the text of column iCol of +** the current document. If successful, (*pz) is set to point to a buffer ** containing the text in utf-8 encoding, (*pn) is set to the size in bytes ** (not characters) of the buffer and SQLITE_OK is returned. Otherwise, ** if an error occurs, an SQLite error code is returned and the final values @@ -12780,8 +12831,10 @@ struct Fts5PhraseIter { ** Returns the number of phrases in the current query expression. ** ** xPhraseSize: -** Returns the number of tokens in phrase iPhrase of the query. Phrases -** are numbered starting from zero. +** If parameter iCol is less than zero, or greater than or equal to the +** number of phrases in the current query, as returned by xPhraseCount, +** 0 is returned. Otherwise, this function returns the number of tokens in +** phrase iPhrase of the query. Phrases are numbered starting from zero. ** ** xInstCount: ** Set *pnInst to the total number of occurrences of all phrases within @@ -12797,12 +12850,13 @@ struct Fts5PhraseIter { ** Query for the details of phrase match iIdx within the current row. ** Phrase matches are numbered starting from zero, so the iIdx argument ** should be greater than or equal to zero and smaller than the value -** output by xInstCount(). +** output by xInstCount(). If iIdx is less than zero or greater than +** or equal to the value returned by xInstCount(), SQLITE_RANGE is returned. ** -** Usually, output parameter *piPhrase is set to the phrase number, *piCol +** Otherwise, output parameter *piPhrase is set to the phrase number, *piCol ** to the column in which it occurs and *piOff the token offset of the -** first token of the phrase. Returns SQLITE_OK if successful, or an error -** code (i.e. SQLITE_NOMEM) if an error occurs. +** first token of the phrase. SQLITE_OK is returned if successful, or an +** error code (i.e. SQLITE_NOMEM) if an error occurs. ** ** This API can be quite slow if used with an FTS5 table created with the ** "detail=none" or "detail=column" option. @@ -12828,6 +12882,10 @@ struct Fts5PhraseIter { ** Invoking Api.xUserData() returns a copy of the pointer passed as ** the third argument to pUserData. ** +** If parameter iPhrase is less than zero, or greater than or equal to +** the number of phrases in the query, as returned by xPhraseCount(), +** this function returns SQLITE_RANGE. +** ** If the callback function returns any value other than SQLITE_OK, the ** query is abandoned and the xQueryPhrase function returns immediately. ** If the returned value is SQLITE_DONE, xQueryPhrase returns SQLITE_OK. @@ -12942,9 +13000,42 @@ struct Fts5PhraseIter { ** ** xPhraseNextColumn() ** See xPhraseFirstColumn above. +** +** xQueryToken(pFts5, iPhrase, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase iPhrase of the current +** query. Before returning, output parameter *ppToken is set to point +** to a buffer containing the requested token, and *pnToken to the +** size of this buffer in bytes. +** +** If iPhrase or iToken are less than zero, or if iPhrase is greater than +** or equal to the number of phrases in the query as reported by +** xPhraseCount(), or if iToken is equal to or greater than the number of +** tokens in the phrase, SQLITE_RANGE is returned and *ppToken and *pnToken + are both zeroed. +** +** The output text is not a copy of the query text that specified the +** token. It is the output of the tokenizer module. For tokendata=1 +** tables, this includes any embedded 0x00 and trailing data. +** +** xInstToken(pFts5, iIdx, iToken, ppToken, pnToken) +** This is used to access token iToken of phrase hit iIdx within the +** current row. If iIdx is less than zero or greater than or equal to the +** value returned by xInstCount(), SQLITE_RANGE is returned. Otherwise, +** output variable (*ppToken) is set to point to a buffer containing the +** matching document token, and (*pnToken) to the size of that buffer in +** bytes. This API is not available if the specified token matches a +** prefix query term. In that case both output variables are always set +** to 0. +** +** The output text is not a copy of the document text that was tokenized. +** It is the output of the tokenizer module. For tokendata=1 tables, this +** includes any embedded 0x00 and trailing data. +** +** This API can be quite slow if used with an FTS5 table created with the +** "detail=none" or "detail=column" option. */ struct Fts5ExtensionApi { - int iVersion; /* Currently always set to 2 */ + int iVersion; /* Currently always set to 3 */ void *(*xUserData)(Fts5Context*); @@ -12979,6 +13070,13 @@ struct Fts5ExtensionApi { int (*xPhraseFirstColumn)(Fts5Context*, int iPhrase, Fts5PhraseIter*, int*); void (*xPhraseNextColumn)(Fts5Context*, Fts5PhraseIter*, int *piCol); + + /* Below this point are iVersion>=3 only */ + int (*xQueryToken)(Fts5Context*, + int iPhrase, int iToken, + const char **ppToken, int *pnToken + ); + int (*xInstToken)(Fts5Context*, int iIdx, int iToken, const char**, int*); }; /* diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3.go b/vendor/github.com/mattn/go-sqlite3/sqlite3.go index 576418233..a16d8abf3 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3.go @@ -607,10 +607,9 @@ func (c *SQLiteConn) RegisterAuthorizer(callback func(int, string, string, strin // RegisterFunc makes a Go function available as a SQLite function. // // The Go function can have arguments of the following types: any -// numeric type except complex, bool, []byte, string and -// interface{}. interface{} arguments are given the direct translation -// of the SQLite data type: int64 for INTEGER, float64 for FLOAT, -// []byte for BLOB, string for TEXT. +// numeric type except complex, bool, []byte, string and any. +// any arguments are given the direct translation of the SQLite data type: +// int64 for INTEGER, float64 for FLOAT, []byte for BLOB, string for TEXT. // // The function can additionally be variadic, as long as the type of // the variadic argument is one of the above. @@ -620,7 +619,7 @@ func (c *SQLiteConn) RegisterAuthorizer(callback func(int, string, string, strin // optimizations in its queries. // // See _example/go_custom_funcs for a detailed example. -func (c *SQLiteConn) RegisterFunc(name string, impl interface{}, pure bool) error { +func (c *SQLiteConn) RegisterFunc(name string, impl any, pure bool) error { var fi functionInfo fi.f = reflect.ValueOf(impl) t := fi.f.Type() @@ -702,7 +701,7 @@ func sqlite3CreateFunction(db *C.sqlite3, zFunctionName *C.char, nArg C.int, eTe // return an error in addition to their other return values. // // See _example/go_custom_funcs for a detailed example. -func (c *SQLiteConn) RegisterAggregator(name string, impl interface{}, pure bool) error { +func (c *SQLiteConn) RegisterAggregator(name string, impl any, pure bool) error { var ai aggInfo ai.constructor = reflect.ValueOf(impl) t := ai.constructor.Type() @@ -976,103 +975,104 @@ func (c *SQLiteConn) begin(ctx context.Context) (driver.Tx, error) { // The argument is may be either in parentheses or it may be separated from // the pragma name by an equal sign. The two syntaxes yield identical results. // In many pragmas, the argument is a boolean. The boolean can be one of: -// 1 yes true on -// 0 no false off +// +// 1 yes true on +// 0 no false off // // You can specify a DSN string using a URI as the filename. -// test.db -// file:test.db?cache=shared&mode=memory -// :memory: -// file::memory: // -// mode -// Access mode of the database. -// https://www.sqlite.org/c3ref/open.html -// Values: -// - ro -// - rw -// - rwc -// - memory +// test.db +// file:test.db?cache=shared&mode=memory +// :memory: +// file::memory: // -// cache -// SQLite Shared-Cache Mode -// https://www.sqlite.org/sharedcache.html -// Values: -// - shared -// - private +// mode +// Access mode of the database. +// https://www.sqlite.org/c3ref/open.html +// Values: +// - ro +// - rw +// - rwc +// - memory // -// immutable=Boolean -// The immutable parameter is a boolean query parameter that indicates -// that the database file is stored on read-only media. When immutable is set, -// SQLite assumes that the database file cannot be changed, -// even by a process with higher privilege, -// and so the database is opened read-only and all locking and change detection is disabled. -// Caution: Setting the immutable property on a database file that -// does in fact change can result in incorrect query results and/or SQLITE_CORRUPT errors. +// cache +// SQLite Shared-Cache Mode +// https://www.sqlite.org/sharedcache.html +// Values: +// - shared +// - private // -// go-sqlite3 adds the following query parameters to those used by SQLite: -// _loc=XXX -// Specify location of time format. It's possible to specify "auto". +// immutable=Boolean +// The immutable parameter is a boolean query parameter that indicates +// that the database file is stored on read-only media. When immutable is set, +// SQLite assumes that the database file cannot be changed, +// even by a process with higher privilege, +// and so the database is opened read-only and all locking and change detection is disabled. +// Caution: Setting the immutable property on a database file that +// does in fact change can result in incorrect query results and/or SQLITE_CORRUPT errors. // -// _mutex=XXX -// Specify mutex mode. XXX can be "no", "full". +// go-sqlite3 adds the following query parameters to those used by SQLite: // -// _txlock=XXX -// Specify locking behavior for transactions. XXX can be "immediate", -// "deferred", "exclusive". +// _loc=XXX +// Specify location of time format. It's possible to specify "auto". // -// _auto_vacuum=X | _vacuum=X -// 0 | none - Auto Vacuum disabled -// 1 | full - Auto Vacuum FULL -// 2 | incremental - Auto Vacuum Incremental +// _mutex=XXX +// Specify mutex mode. XXX can be "no", "full". // -// _busy_timeout=XXX"| _timeout=XXX -// Specify value for sqlite3_busy_timeout. +// _txlock=XXX +// Specify locking behavior for transactions. XXX can be "immediate", +// "deferred", "exclusive". // -// _case_sensitive_like=Boolean | _cslike=Boolean -// https://www.sqlite.org/pragma.html#pragma_case_sensitive_like -// Default or disabled the LIKE operation is case-insensitive. -// When enabling this options behaviour of LIKE will become case-sensitive. +// _auto_vacuum=X | _vacuum=X +// 0 | none - Auto Vacuum disabled +// 1 | full - Auto Vacuum FULL +// 2 | incremental - Auto Vacuum Incremental // -// _defer_foreign_keys=Boolean | _defer_fk=Boolean -// Defer Foreign Keys until outermost transaction is committed. +// _busy_timeout=XXX"| _timeout=XXX +// Specify value for sqlite3_busy_timeout. // -// _foreign_keys=Boolean | _fk=Boolean -// Enable or disable enforcement of foreign keys. +// _case_sensitive_like=Boolean | _cslike=Boolean +// https://www.sqlite.org/pragma.html#pragma_case_sensitive_like +// Default or disabled the LIKE operation is case-insensitive. +// When enabling this options behaviour of LIKE will become case-sensitive. // -// _ignore_check_constraints=Boolean -// This pragma enables or disables the enforcement of CHECK constraints. -// The default setting is off, meaning that CHECK constraints are enforced by default. +// _defer_foreign_keys=Boolean | _defer_fk=Boolean +// Defer Foreign Keys until outermost transaction is committed. // -// _journal_mode=MODE | _journal=MODE -// Set journal mode for the databases associated with the current connection. -// https://www.sqlite.org/pragma.html#pragma_journal_mode +// _foreign_keys=Boolean | _fk=Boolean +// Enable or disable enforcement of foreign keys. // -// _locking_mode=X | _locking=X -// Sets the database connection locking-mode. -// The locking-mode is either NORMAL or EXCLUSIVE. -// https://www.sqlite.org/pragma.html#pragma_locking_mode +// _ignore_check_constraints=Boolean +// This pragma enables or disables the enforcement of CHECK constraints. +// The default setting is off, meaning that CHECK constraints are enforced by default. // -// _query_only=Boolean -// The query_only pragma prevents all changes to database files when enabled. +// _journal_mode=MODE | _journal=MODE +// Set journal mode for the databases associated with the current connection. +// https://www.sqlite.org/pragma.html#pragma_journal_mode // -// _recursive_triggers=Boolean | _rt=Boolean -// Enable or disable recursive triggers. +// _locking_mode=X | _locking=X +// Sets the database connection locking-mode. +// The locking-mode is either NORMAL or EXCLUSIVE. +// https://www.sqlite.org/pragma.html#pragma_locking_mode // -// _secure_delete=Boolean|FAST -// When secure_delete is on, SQLite overwrites deleted content with zeros. -// https://www.sqlite.org/pragma.html#pragma_secure_delete +// _query_only=Boolean +// The query_only pragma prevents all changes to database files when enabled. // -// _synchronous=X | _sync=X -// Change the setting of the "synchronous" flag. -// https://www.sqlite.org/pragma.html#pragma_synchronous +// _recursive_triggers=Boolean | _rt=Boolean +// Enable or disable recursive triggers. // -// _writable_schema=Boolean -// When this pragma is on, the SQLITE_MASTER tables in which database -// can be changed using ordinary UPDATE, INSERT, and DELETE statements. -// Warning: misuse of this pragma can easily result in a corrupt database file. +// _secure_delete=Boolean|FAST +// When secure_delete is on, SQLite overwrites deleted content with zeros. +// https://www.sqlite.org/pragma.html#pragma_secure_delete // +// _synchronous=X | _sync=X +// Change the setting of the "synchronous" flag. +// https://www.sqlite.org/pragma.html#pragma_synchronous // +// _writable_schema=Boolean +// When this pragma is on, the SQLITE_MASTER tables in which database +// can be changed using ordinary UPDATE, INSERT, and DELETE statements. +// Warning: misuse of this pragma can easily result in a corrupt database file. func (d *SQLiteDriver) Open(dsn string) (driver.Conn, error) { if C.sqlite3_threadsafe() == 0 { return nil, errors.New("sqlite library was not compiled for thread-safe operation") diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_func_crypt.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_func_crypt.go index afd93333d..bd9a3bc09 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_func_crypt.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_func_crypt.go @@ -50,15 +50,15 @@ import ( // perhaps using a cryptographic hash function like SHA1. // CryptEncoderSHA1 encodes a password with SHA1 -func CryptEncoderSHA1(pass []byte, hash interface{}) []byte { +func CryptEncoderSHA1(pass []byte, hash any) []byte { h := sha1.Sum(pass) return h[:] } // CryptEncoderSSHA1 encodes a password with SHA1 with the // configured salt. -func CryptEncoderSSHA1(salt string) func(pass []byte, hash interface{}) []byte { - return func(pass []byte, hash interface{}) []byte { +func CryptEncoderSSHA1(salt string) func(pass []byte, hash any) []byte { + return func(pass []byte, hash any) []byte { s := []byte(salt) p := append(pass, s...) h := sha1.Sum(p) @@ -67,15 +67,15 @@ func CryptEncoderSSHA1(salt string) func(pass []byte, hash interface{}) []byte { } // CryptEncoderSHA256 encodes a password with SHA256 -func CryptEncoderSHA256(pass []byte, hash interface{}) []byte { +func CryptEncoderSHA256(pass []byte, hash any) []byte { h := sha256.Sum256(pass) return h[:] } // CryptEncoderSSHA256 encodes a password with SHA256 // with the configured salt -func CryptEncoderSSHA256(salt string) func(pass []byte, hash interface{}) []byte { - return func(pass []byte, hash interface{}) []byte { +func CryptEncoderSSHA256(salt string) func(pass []byte, hash any) []byte { + return func(pass []byte, hash any) []byte { s := []byte(salt) p := append(pass, s...) h := sha256.Sum256(p) @@ -84,15 +84,15 @@ func CryptEncoderSSHA256(salt string) func(pass []byte, hash interface{}) []byte } // CryptEncoderSHA384 encodes a password with SHA384 -func CryptEncoderSHA384(pass []byte, hash interface{}) []byte { +func CryptEncoderSHA384(pass []byte, hash any) []byte { h := sha512.Sum384(pass) return h[:] } // CryptEncoderSSHA384 encodes a password with SHA384 // with the configured salt -func CryptEncoderSSHA384(salt string) func(pass []byte, hash interface{}) []byte { - return func(pass []byte, hash interface{}) []byte { +func CryptEncoderSSHA384(salt string) func(pass []byte, hash any) []byte { + return func(pass []byte, hash any) []byte { s := []byte(salt) p := append(pass, s...) h := sha512.Sum384(p) @@ -101,15 +101,15 @@ func CryptEncoderSSHA384(salt string) func(pass []byte, hash interface{}) []byte } // CryptEncoderSHA512 encodes a password with SHA512 -func CryptEncoderSHA512(pass []byte, hash interface{}) []byte { +func CryptEncoderSHA512(pass []byte, hash any) []byte { h := sha512.Sum512(pass) return h[:] } // CryptEncoderSSHA512 encodes a password with SHA512 // with the configured salt -func CryptEncoderSSHA512(salt string) func(pass []byte, hash interface{}) []byte { - return func(pass []byte, hash interface{}) []byte { +func CryptEncoderSSHA512(salt string) func(pass []byte, hash any) []byte { + return func(pass []byte, hash any) []byte { s := []byte(salt) p := append(pass, s...) h := sha512.Sum512(p) diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_go18.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_go18.go index 514fd7ecb..34cad08e4 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_go18.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_go18.go @@ -3,8 +3,8 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. -// +build cgo -// +build go1.8 +//go:build cgo && go1.8 +// +build cgo,go1.8 package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_libsqlite3.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_libsqlite3.go index ac609c956..95cc7c0b6 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_libsqlite3.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_libsqlite3.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build libsqlite3 // +build libsqlite3 package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension.go index 9433fea82..03cbc8b68 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build !sqlite_omit_load_extension // +build !sqlite_omit_load_extension package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension_omit.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension_omit.go index 8c75f9bda..d4f8ce651 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension_omit.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_load_extension_omit.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_omit_load_extension // +build sqlite_omit_load_extension package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_allow_uri_authority.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_allow_uri_authority.go index 8c4d4d20d..51240cbf6 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_allow_uri_authority.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_allow_uri_authority.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_allow_uri_authority // +build sqlite_allow_uri_authority package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_app_armor.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_app_armor.go index 63c80cfea..565dbc298 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_app_armor.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_app_armor.go @@ -4,8 +4,8 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. -// +build !windows -// +build sqlite_app_armor +//go:build !windows && sqlite_app_armor +// +build !windows,sqlite_app_armor package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_column_metadata.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_column_metadata.go index c67fa82b1..63659b46b 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_column_metadata.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_column_metadata.go @@ -1,3 +1,4 @@ +//go:build sqlite_column_metadata // +build sqlite_column_metadata package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_foreign_keys.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_foreign_keys.go index a676e097a..82c944e1b 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_foreign_keys.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_foreign_keys.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_foreign_keys // +build sqlite_foreign_keys package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_fts5.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_fts5.go index 0f38df75d..2645f284b 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_fts5.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_fts5.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_fts5 || fts5 // +build sqlite_fts5 fts5 package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_icu.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_icu.go index f82bdd0d4..2d47827be 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_icu.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_icu.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_icu || icu // +build sqlite_icu icu package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_introspect.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_introspect.go index 6512b2b3b..cd2e54011 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_introspect.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_introspect.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_introspect // +build sqlite_introspect package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_math_functions.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_math_functions.go index 7cd68d3f5..bd62d9a2a 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_math_functions.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_math_functions.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_math_functions // +build sqlite_math_functions package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate.go index cea032e3e..ed725eeb9 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build cgo // +build cgo package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_hook.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_hook.go index b43e4821b..8cce278fd 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_hook.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_hook.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_preupdate_hook // +build sqlite_preupdate_hook package sqlite3 @@ -54,10 +55,10 @@ func (d *SQLitePreUpdateData) Count() int { return int(C.sqlite3_preupdate_count(d.Conn.db)) } -func (d *SQLitePreUpdateData) row(dest []interface{}, new bool) error { +func (d *SQLitePreUpdateData) row(dest []any, new bool) error { for i := 0; i < d.Count() && i < len(dest); i++ { var val *C.sqlite3_value - var src interface{} + var src any // Initially I tried making this just a function pointer argument, but // it's absurdly complicated to pass C function pointers. @@ -95,7 +96,7 @@ func (d *SQLitePreUpdateData) row(dest []interface{}, new bool) error { // Old populates dest with the row data to be replaced. This works similar to // database/sql's Rows.Scan() -func (d *SQLitePreUpdateData) Old(dest ...interface{}) error { +func (d *SQLitePreUpdateData) Old(dest ...any) error { if d.Op == SQLITE_INSERT { return errors.New("There is no old row for INSERT operations") } @@ -104,7 +105,7 @@ func (d *SQLitePreUpdateData) Old(dest ...interface{}) error { // New populates dest with the replacement row data. This works similar to // database/sql's Rows.Scan() -func (d *SQLitePreUpdateData) New(dest ...interface{}) error { +func (d *SQLitePreUpdateData) New(dest ...any) error { if d.Op == SQLITE_DELETE { return errors.New("There is no new row for DELETE operations") } diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_omit.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_omit.go index c510a15b4..f60da6c16 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_omit.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_preupdate_omit.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build !sqlite_preupdate_hook && cgo // +build !sqlite_preupdate_hook,cgo package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete.go index 934fa6b8e..6bb05b843 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_secure_delete // +build sqlite_secure_delete package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete_fast.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete_fast.go index b0de130ff..982020aeb 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete_fast.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_secure_delete_fast.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_secure_delete_fast // +build sqlite_secure_delete_fast package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize.go index 2560c43a7..f1710c1c3 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize.go @@ -1,3 +1,4 @@ +//go:build !libsqlite3 || sqlite_serialize // +build !libsqlite3 sqlite_serialize package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize_omit.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize_omit.go index b154dd34d..d00ead0b6 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize_omit.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_serialize_omit.go @@ -1,3 +1,4 @@ +//go:build libsqlite3 && !sqlite_serialize // +build libsqlite3,!sqlite_serialize package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_stat4.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_stat4.go index d4d30f0dd..799fbb0fc 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_stat4.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_stat4.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_stat4 // +build sqlite_stat4 package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_unlock_notify.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_unlock_notify.go index adfa26c54..76f7bbfb6 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_unlock_notify.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_unlock_notify.go @@ -3,8 +3,8 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. -// +build cgo -// +build sqlite_unlock_notify +//go:build cgo && sqlite_unlock_notify +// +build cgo,sqlite_unlock_notify package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth.go index b62b60840..de9630c27 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_userauth // +build sqlite_userauth package sqlite3 @@ -79,7 +80,7 @@ var ( // If a database contains the SQLITE_USER table, then the // call to Authenticate must be invoked with an // appropriate username and password prior to enable read and write -//access to the database. +// access to the database. // // Return SQLITE_OK on success or SQLITE_ERROR if the username/password // combination is incorrect or unknown. @@ -103,9 +104,10 @@ func (c *SQLiteConn) Authenticate(username, password string) error { // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authenticate(username, password string) int { // Allocate C Variables cuser := C.CString(username) @@ -155,9 +157,10 @@ func (c *SQLiteConn) AuthUserAdd(username, password string, admin bool) error { // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authUserAdd(username, password string, admin int) int { // Allocate C Variables cuser := C.CString(username) @@ -207,9 +210,10 @@ func (c *SQLiteConn) AuthUserChange(username, password string, admin bool) error // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authUserChange(username, password string, admin int) int { // Allocate C Variables cuser := C.CString(username) @@ -249,9 +253,10 @@ func (c *SQLiteConn) AuthUserDelete(username string) error { // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authUserDelete(username string) int { // Allocate C Variables cuser := C.CString(username) @@ -280,8 +285,9 @@ func (c *SQLiteConn) AuthEnabled() (exists bool) { // It is however exported for usage within SQL by the user. // // Returns: -// 0 - Disabled -// 1 - Enabled +// +// 0 - Disabled +// 1 - Enabled func (c *SQLiteConn) authEnabled() int { return int(C._sqlite3_auth_enabled(c.db)) } diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth_omit.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth_omit.go index 302cd57a9..15370df70 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth_omit.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_userauth_omit.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build !sqlite_userauth // +build !sqlite_userauth package sqlite3 @@ -17,7 +18,7 @@ import ( // If a database contains the SQLITE_USER table, then the // call to Authenticate must be invoked with an // appropriate username and password prior to enable read and write -//access to the database. +// access to the database. // // Return SQLITE_OK on success or SQLITE_ERROR if the username/password // combination is incorrect or unknown. @@ -34,9 +35,10 @@ func (c *SQLiteConn) Authenticate(username, password string) error { // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authenticate(username, password string) int { // NOOP return 0 @@ -65,9 +67,10 @@ func (c *SQLiteConn) AuthUserAdd(username, password string, admin bool) error { // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authUserAdd(username, password string, admin int) int { // NOOP return 0 @@ -96,9 +99,10 @@ func (c *SQLiteConn) AuthUserChange(username, password string, admin bool) error // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authUserChange(username, password string, admin int) int { // NOOP return 0 @@ -122,9 +126,10 @@ func (c *SQLiteConn) AuthUserDelete(username string) error { // It is however exported for usage within SQL by the user. // // Returns: -// C.SQLITE_OK (0) -// C.SQLITE_ERROR (1) -// C.SQLITE_AUTH (23) +// +// C.SQLITE_OK (0) +// C.SQLITE_ERROR (1) +// C.SQLITE_AUTH (23) func (c *SQLiteConn) authUserDelete(username string) int { // NOOP return 0 @@ -142,8 +147,9 @@ func (c *SQLiteConn) AuthEnabled() (exists bool) { // It is however exported for usage within SQL by the user. // // Returns: -// 0 - Disabled -// 1 - Enabled +// +// 0 - Disabled +// 1 - Enabled func (c *SQLiteConn) authEnabled() int { // NOOP return 0 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_full.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_full.go index 5185a96d0..df13c9d2d 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_full.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_full.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_vacuum_full // +build sqlite_vacuum_full package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_incr.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_incr.go index a9d8a185c..a2e48814b 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_incr.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vacuum_incr.go @@ -4,6 +4,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_vacuum_incr // +build sqlite_vacuum_incr package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vtable.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vtable.go index 4a93c4652..9b164b3e0 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vtable.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_opt_vtable.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_vtable || vtable // +build sqlite_vtable vtable package sqlite3 @@ -516,7 +517,7 @@ func goMDestroy(pClientData unsafe.Pointer) { func goVFilter(pCursor unsafe.Pointer, idxNum C.int, idxName *C.char, argc C.int, argv **C.sqlite3_value) *C.char { vtc := lookupHandle(pCursor).(*sqliteVTabCursor) args := (*[(math.MaxInt32 - 1) / unsafe.Sizeof((*C.sqlite3_value)(nil))]*C.sqlite3_value)(unsafe.Pointer(argv))[:argc:argc] - vals := make([]interface{}, 0, argc) + vals := make([]any, 0, argc) for _, v := range args { conv, err := callbackArgGeneric(v) if err != nil { @@ -588,7 +589,7 @@ func goVUpdate(pVTab unsafe.Pointer, argc C.int, argv **C.sqlite3_value, pRowid if v, ok := vt.vTab.(VTabUpdater); ok { // convert argv args := (*[(math.MaxInt32 - 1) / unsafe.Sizeof((*C.sqlite3_value)(nil))]*C.sqlite3_value)(unsafe.Pointer(argv))[:argc:argc] - vals := make([]interface{}, 0, argc) + vals := make([]any, 0, argc) for _, v := range args { conv, err := callbackArgGeneric(v) if err != nil { @@ -662,9 +663,9 @@ type VTab interface { // deleted. // See: https://sqlite.org/vtab.html#xupdate type VTabUpdater interface { - Delete(interface{}) error - Insert(interface{}, []interface{}) (int64, error) - Update(interface{}, []interface{}) error + Delete(any) error + Insert(any, []any) (int64, error) + Update(any, []any) error } // VTabCursor describes cursors that point into the virtual table and are used @@ -673,7 +674,7 @@ type VTabCursor interface { // http://sqlite.org/vtab.html#xclose Close() error // http://sqlite.org/vtab.html#xfilter - Filter(idxNum int, idxStr string, vals []interface{}) error + Filter(idxNum int, idxStr string, vals []any) error // http://sqlite.org/vtab.html#xnext Next() error // http://sqlite.org/vtab.html#xeof diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_other.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_other.go index 077d3c644..1f9a75506 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_other.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_other.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build !windows // +build !windows package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_solaris.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_solaris.go index 102f90c9a..fb4d32517 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_solaris.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_solaris.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build solaris // +build solaris package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_trace.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_trace.go index 56bb91490..6c47cce19 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_trace.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_trace.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build sqlite_trace || trace // +build sqlite_trace trace package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_type.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_type.go index 0fd8210bb..20537a097 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_type.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_type.go @@ -74,7 +74,7 @@ func scanType(cdt string) reflect.Type { case SQLITE_TIME: return reflect.TypeOf(sql.NullTime{}) } - return reflect.TypeOf(new(interface{})) + return reflect.TypeOf(new(any)) } func databaseTypeConvSqlite(t string) int { diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_usleep_windows.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_usleep_windows.go index b6739bf62..6527f6fd9 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_usleep_windows.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_usleep_windows.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build cgo // +build cgo package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/sqlite3_windows.go b/vendor/github.com/mattn/go-sqlite3/sqlite3_windows.go index 81aa2abde..f863bcd36 100644 --- a/vendor/github.com/mattn/go-sqlite3/sqlite3_windows.go +++ b/vendor/github.com/mattn/go-sqlite3/sqlite3_windows.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build windows // +build windows package sqlite3 diff --git a/vendor/github.com/mattn/go-sqlite3/static_mock.go b/vendor/github.com/mattn/go-sqlite3/static_mock.go index f19e842ff..d2c5a2760 100644 --- a/vendor/github.com/mattn/go-sqlite3/static_mock.go +++ b/vendor/github.com/mattn/go-sqlite3/static_mock.go @@ -3,6 +3,7 @@ // Use of this source code is governed by an MIT-style // license that can be found in the LICENSE file. +//go:build !cgo // +build !cgo package sqlite3 @@ -28,10 +29,10 @@ type ( ) func (SQLiteDriver) Open(s string) (driver.Conn, error) { return nil, errorMsg } -func (c *SQLiteConn) RegisterAggregator(string, interface{}, bool) error { return errorMsg } +func (c *SQLiteConn) RegisterAggregator(string, any, bool) error { return errorMsg } func (c *SQLiteConn) RegisterAuthorizer(func(int, string, string, string) int) {} func (c *SQLiteConn) RegisterCollation(string, func(string, string) int) error { return errorMsg } func (c *SQLiteConn) RegisterCommitHook(func() int) {} -func (c *SQLiteConn) RegisterFunc(string, interface{}, bool) error { return errorMsg } +func (c *SQLiteConn) RegisterFunc(string, any, bool) error { return errorMsg } func (c *SQLiteConn) RegisterRollbackHook(func()) {} func (c *SQLiteConn) RegisterUpdateHook(func(int, string, string, int64)) {} diff --git a/vendor/modules.txt b/vendor/modules.txt index 9a7eaa0e2..e0681a2b2 100644 --- a/vendor/modules.txt +++ b/vendor/modules.txt @@ -68,8 +68,8 @@ github.com/kylelemons/go-gypsy/yaml github.com/lib/pq github.com/lib/pq/oid github.com/lib/pq/scram -# github.com/mattn/go-sqlite3 v1.14.19 -## explicit; go 1.16 +# github.com/mattn/go-sqlite3 v1.14.20 +## explicit; go 1.19 github.com/mattn/go-sqlite3 # github.com/matttproud/golang_protobuf_extensions/v2 v2.0.0 ## explicit; go 1.19