- Fix EARL Turtle report.
- Test with karma.
- Test with Node.js 20.x.
- Add
inputFormat
option. Use "application/n-quads" for a N-Quads string that will be parsed. Omit option for a JSON dataset. This can simplify a common case of using the internal parser to generate a dataset.- NOTE: The
inputFormat
option was previously ignored and is now used. Any calling code that was passing in an incorrect value needs to be fixed.
- NOTE: The
- Add
signal
option to allow use of anAbortSignal
for complexity control. Enables the algorithm to abort after a timeout, manual abort, or other condition. - Add
maxWorkFactor
to calculate a deep iteration limit based on the number of non-unique blank nodes. This defaults to1
for roughly O(n) behavior and will handle common graphs. It must be adjusted to higher values if there is a need to process graphs with complex blank nodes or other "poison" graphs. It is recommended to use this parameter instead ofmaxDeepIterations
directly. IfmaxDeepIterations
is provided, thenmaxWorkFactor
will be ignored. - BREAKING: Check output
format
parameter. Must be omitted, falsy, or "application/n-quads". - Add EARL Turtle test result mode.
- Add
EARL_OFFICIAL
env flag to setup official test report mode. - Add
"react-native"
section topackage.json
(same as"browser"
), and instructions for how to use this library with React Native.
- BREAKING: Remove support for Node.js < 18. This is done to allow updates to tooling that no longer support older Node.js versions. The library code has not yet changed to be incompatible with older Node.js versions but it will no longer be tested and may become incompatible at any time.
- BREAKING: Change algorithm name from "URDNA2015" to "RDFC-1.0" to match
rdf-canon changes. Use of "URDNA2015" is now deprecated and an alias for
"RDFC-1.0". An API option
rejectURDNA2015
is available to disable "URDNA2015" support. A globalRDF_CANONIZE_TRACE_URDNA2015
is available to developers to trace calls that use "URDNA2015". See the README for important compatibility notes and API details. - BREAKING: Use latest rdf-canon N-Quads canonical form. This can
change the canonical output! There is an expanded set of literal string
control characters that are escaped as an
ECHAR
orUCHAR
instead of using a native representation.- Previously: the canonical N-Quads form used here was encoding
\u000A
(\n
),\u000D
(\r
),\u0022
("
), and\u005C
(\
) asECHARs
:\n
,\r
,\"
, and\\
, All other characters were represented as native Unicode. - Now: the output also encodes
\u0008
(\b
),\u0009
(\t
),\u000C
(\f
) asECHARs
\b
,\t
, and\f
, and encodes the "control" characters in the range of\u0000-\u001F
and\u007F
asUCHARs
\u00xx
. All other characters are represented as native Unicode.
- Previously: the canonical N-Quads form used here was encoding
- BREAKING: Use
globalThis
to accesscrypto
in browsers. Use a polyfill if your environment doesn't supportglobalThis
. - BREAKING: Change dataset handling of
BlankNodes
to match the RDF/JS: Data model specification. The_:
prefix is no longer used in theBlankNode
value
field. This should improve compatibility with other RDF/JS tooling but may cause compatibility issues with existing code. The previous behavior is historical and may predate the RDF/JS spec. - BREAKING: Change maximum deep iterations error text.
- Update tooling.
- Update for latest rdf-canon changes: test suite location, README, links, and identifiers.
- More closely align test code with the version in jsonld.js.
- Use combined test/benchmark system.
- Support running multiple test jobs in parallel.
- Refactor
MessageDigest-browser.js
toMessageDigest-webcrypto.js
so it can also be optionally used with Node.js. - Move platform specific support into
platform.js
andplatform-browser.js
. - Optimize WebCrypto bytes to hex conversion:
- Improvement depends on number of digests performed.
- Node.js using the improved browser algorithm can be ~4-9% faster overall.
- Node.js native
Buffer
conversion can be ~5-12% faster overall.
- Optimize a N-Quads serialization call.
- Optimize N-Quads escape/unescape calling replace:
- Run regex test before doing a replace call.
- Performance difference depends on data and how often escape/unescape would need to be called. A benchmark test data showed ~3-5% overall improvement.
- Optimize N-Quads escape replacement:
- Use a pre-computed map of replacement values.
- Performance difference depends on the number of replacements. The rdf-canon escaping test showed up to 15% improvement.
- Support generalized RDF
BlankNode
predicate during N-Quads serialization.
- Disable native lib tests in a browser.
- Disable sync tests in a browser. The sync code attempts to use the async webcrypto calls and produces invalid results. It is an error that this doesn't fail, but sync code is currently only for testing.
- Fix various testing and benchmark bugs.
- Escape and unescape all data.
- Support 8 hex char Unicode values.
- BREAKING: Remove URGNA2012 support. rdf-canon no longer supports or has a test suite for URGNA2012. URDNA2015 has been the preferred algorithm for many years.
- BREAKING: Remove deprecated support for legacy dataset format.
- Remove
benchmark/benchmark.js
tool in favor of combined test system and benchmarking control via environment vars.
- Allow
canonicalIdMap
to be passed tocanonize
which will be populated by the canonical identifier issuer with the bnode identifier mapping generated by the canonicalization algorithm. This feature is particularly useful when the resulting bnode labels need to be changed for use cases such as selective disclosure.
- Add optional
createMessageDigest
factory function for generating aMessageDigest
interface. This allows different hash implementations or even different hash algorithms, including HMACs to be used with URDNA2015. Note that using a different hash algorithm from SHA-256 will change the output.
- Fix typo in unsupported algorithm error.
- Test that input is not changed.
- Optimize quad processing.
- Allow a maximum number of iterations of the N-Degree Hash Quads algorithm to
be set, preventing unusual datasets (and likely meaningless or malicious) from
consuming unnecessary CPU cycles. If the set maximum is exceeded then an
error will be thrown, terminating the canonize process. This option has only
been added to URDNA2015. A future major breaking release is expected to set
the maximum number of iterations to a safe value by default; this release is
backwards compatible and therefore sets no default. A recommended value is
1
, which will cause, at most, each blank node to have the N-degree algorithm executed on it just once.
- BREAKING: Only support Node.js >= 12. Remove related tests, dependencies,
and generated
node6
output. - BREAKING: Remove browser bundles. Simplifies package and reduces install size. If you have a use case that requires the bundles, please file an issue.
- Fix browser override file path style.
- Use
setimmediate
package forsetImmediate
polyfill. The previous custom polyfill was removed. This should allow current projects using this package to stay the same and allow an easy future transition to webpack v5.
- BREAKING: Removed public API for
canonizeSync
. It is still available for testing purposes but does not run in the browser. - BREAKING: Removed dependency on
forge
which means that this library will only run in browsers that have support for the WebCrypto API (or an external polyfill for it). - BREAKING: Do not expose
existing
onIdentifierIssuer
. The old IDs can be retrieved in order viagetOldIds
.
- General optimizations and modernization of the library.
- Add
getOldIds
function toIdentifierIssuer
.
- Use node-forge@0.10.0.
- Optimize away length check on paths.
- Update node-forge dependency.
- Update semver dependency.
- Update node-forge dependency.
- Fix triple comparator in n-quads parser.
- Add eslint support.
- Remove use of deprecated
util.isUndefined()
. Avoids unneededutil
polyfill in webpack build.
- WARNING: This release has a BREAKING change that could cause the canonical N-Quads output to differ from previous releases. Specifically, tabs in literals are no longer escaped. No backwards compatibility mode is provided at this time but if you believe it is needed, please file an issue.
- If you wish to use the native bindings, you must now install
rdf-canonize-native
yourself. It is no longer a dependency. See below.
- BREAKING: N-Quad canonical serialized output.
- Only escape 4 chars.
- Now compatible with https://www.w3.org/TR/n-triples/#canonical-ntriples
- Improve N-Quads parsing.
- Unescape literals.
- Handle Unicode escapes.
- N-Quad serialization optimization.
- Varies based on input by roughly ~1-2x.
- BREAKING: Remove
rdf-canonize-native
as a dependency. The native bindings will still be used ifrdf-canonize-native
can be loaded. This means if you want the native bindings you must install them yourself. This change was done due to various problems caused by having any type of dependency involving the native code. With modern runtimes the JavaScript implementation is in many cases faster. The native bindings do have overhead but can be useful in cases where you need to offload canonizing into the background. It is recommended to perform benchmarks to determine which method works best in your case. - Update webpack and babel.
- BREAKING: Remove
usePureJavaScript
option and make the JavaScript implementation the default. Add explicituseNative
option to force the use of the native implementation fromrdf-canonize-native
. An error will be thrown if native bindings are not available.
- BREAKING: Move native support to optional
rdf-canonize-native
package. If native support is required in your environment then also add a dependency on therdf-canonize-native
package directly. This package only has an optional dependency on the native package to allow systems without native binding build tools to use the JavaScript implementation alone.
- Istanbul coverage support.
- Accept N-Quads upper case language tag.
- Improve acceptable N-Quads blank node labels.
- Update for Node.js 10 / OpenSSL 1.1 API.
- Update nan dependency for Node.js 10 support.
- Avoid variable length arrays. Not supported by some C++ compilers.
- Use const array initializer sizes.
- Comment out debug logging.
- Distribute
binding.gyp
.
- Benchmark tool using the same manifests as the test system.
- Support Node.js 6.
- Native Node.js addon support for URDNA2015. Improves performance.
usePureJavaScript
option to only use JavaScript.
- BREAKING: Remove Node.js 4.x testing and native support. Use a transpiler such as babel if you need further 4.x support.
- Expose
IdentifierIssuer
helper class.
- Fix build.
- Change internals to use ES6.
- Return Promise from API for async method.
- Move node-forge to dependencies.
- RDF Dataset Normalization async implementation from jsonld.js.
- webpack support.
- Split messageDigest into Node.js and browser files.
- Node.js file uses native crypto module.
- Browser file uses forge.
- See git history for all changes.