diff --git a/velox/common/encode/Base32.cpp b/velox/common/encode/Base32.cpp index 6705b0363e7c..de7e4765bf83 100644 --- a/velox/common/encode/Base32.cpp +++ b/velox/common/encode/Base32.cpp @@ -182,4 +182,119 @@ template } } +size_t Base32::calculateDecodedSize(const char* data, size_t& size) { + if (size == 0) { + return 0; + } + + // Check if the input data is padded + if (isPadded(data, size)) { + /// If padded, ensure that the string length is a multiple of the encoded + /// block size. + if (size % kEncodedBlockByteSize != 0) { + VELOX_USER_FAIL( + "Base32::decode() - invalid input string: " + "string length is not a multiple of 8."); + } + + auto needed = (size * kBinaryBlockByteSize) / kEncodedBlockByteSize; + auto padding = numPadding(data, size); + size -= padding; + + // Adjust the needed size by deducting the bytes corresponding to the + // padding from the calculated size. + return needed - + ((padding * kBinaryBlockByteSize) + (kEncodedBlockByteSize - 1)) / + kEncodedBlockByteSize; + } else { + // If not padded, calculate extra bytes, if any. + auto extra = size % kEncodedBlockByteSize; + auto needed = (size / kEncodedBlockByteSize) * kBinaryBlockByteSize; + + // Adjust the needed size for extra bytes, if present. + if (extra) { + if ((extra == 6) || (extra == 3) || (extra == 1)) { + VELOX_USER_FAIL( + "Base32::decode() - invalid input string: " + "string length cannot be 6, 3 or 1 more than a multiple of 8."); + } + needed += (extra * kBinaryBlockByteSize) / kEncodedBlockByteSize; + } + + return needed; + } +} + +size_t +Base32::decode(const char* src, size_t src_len, char* dst, size_t dst_len) { + return decodeImpl(src, src_len, dst, dst_len, kBase32ReverseIndexTable); +} + +size_t Base32::decodeImpl( + const char* src, + size_t src_len, + char* dst, + size_t dst_len, + const ReverseIndex& reverse_lookup) { + if (!src_len) { + return 0; + } + + auto needed = calculateDecodedSize(src, src_len); + if (dst_len < needed) { + VELOX_USER_FAIL( + "Base32::decode() - invalid output string: " + "output string is too small."); + } + + // Handle full groups of 8 characters. + for (; src_len > 8; src_len -= 8, src += 8, dst += 5) { + /// Each character of the 8 bytes encode 5 bits of the original, grab each + /// with the appropriate shifts to rebuild the original and then split that + /// back into the original 8 bit bytes. + uint64_t last = + (uint64_t(baseReverseLookup(kBase, src[0], reverse_lookup)) << 35) | + (uint64_t(baseReverseLookup(kBase, src[1], reverse_lookup)) << 30) | + (baseReverseLookup(kBase, src[2], reverse_lookup) << 25) | + (baseReverseLookup(kBase, src[3], reverse_lookup) << 20) | + (baseReverseLookup(kBase, src[4], reverse_lookup) << 15) | + (baseReverseLookup(kBase, src[5], reverse_lookup) << 10) | + (baseReverseLookup(kBase, src[6], reverse_lookup) << 5) | + baseReverseLookup(kBase, src[7], reverse_lookup); + dst[0] = (last >> 32) & 0xff; + dst[1] = (last >> 24) & 0xff; + dst[2] = (last >> 16) & 0xff; + dst[3] = (last >> 8) & 0xff; + dst[4] = last & 0xff; + } + + /// Handle the last 2, 4, 5, 7 or 8 characters. This is similar to the above, + /// but the last characters may or may not exist. + DCHECK(src_len >= 2); + uint64_t last = + (uint64_t(baseReverseLookup(kBase, src[0], reverse_lookup)) << 35) | + (uint64_t(baseReverseLookup(kBase, src[1], reverse_lookup)) << 30); + dst[0] = (last >> 32) & 0xff; + if (src_len > 2) { + last |= baseReverseLookup(kBase, src[2], reverse_lookup) << 25; + last |= baseReverseLookup(kBase, src[3], reverse_lookup) << 20; + dst[1] = (last >> 24) & 0xff; + if (src_len > 4) { + last |= baseReverseLookup(kBase, src[4], reverse_lookup) << 15; + dst[2] = (last >> 16) & 0xff; + if (src_len > 5) { + last |= baseReverseLookup(kBase, src[5], reverse_lookup) << 10; + last |= baseReverseLookup(kBase, src[6], reverse_lookup) << 5; + dst[3] = (last >> 8) & 0xff; + if (src_len > 7) { + last |= baseReverseLookup(kBase, src[7], reverse_lookup); + dst[4] = last & 0xff; + } + } + } + } + + return needed; +} + } // namespace facebook::velox::encoding diff --git a/velox/common/encode/Base32.h b/velox/common/encode/Base32.h index 72d223659e03..3266b32d8e1a 100644 --- a/velox/common/encode/Base32.h +++ b/velox/common/encode/Base32.h @@ -34,6 +34,16 @@ class Base32 { /// returned by the calculateEncodedSize(). static void encode(const char* data, size_t size, char* output); + /// Returns decoded size for the specified input. Adjusts the 'size' to + /// subtract the length of the padding, if exists. + static size_t calculateDecodedSize(const char* data, size_t& size); + + /// Decodes the specified number of characters from the 'src' and writes the + /// result to the 'dst'. The destination must have enough space, e.g. as + /// returned by the calculateDecodedSize(). + static size_t + decode(const char* src, size_t src_len, char* dst, size_t dst_len); + private: template static void encodeImpl( @@ -41,6 +51,15 @@ class Base32 { const Charset& charset, bool include_pad, char* out); + + /// Decodes the specified number of base 32 encoded characters from the 'src' + /// and writes to 'dst' + static size_t decodeImpl( + const char* src, + size_t src_len, + char* dst, + size_t dst_len, + const ReverseIndex& table); }; } // namespace facebook::velox::encoding diff --git a/velox/common/encode/tests/Base32Test.cpp b/velox/common/encode/tests/Base32Test.cpp index 8ed3beb1e874..99fd13f235fb 100644 --- a/velox/common/encode/tests/Base32Test.cpp +++ b/velox/common/encode/tests/Base32Test.cpp @@ -39,4 +39,35 @@ TEST_F(Base32Test, calculateEncodedSizeProperSize) { EXPECT_EQ(24, Base32::calculateEncodedSize(11, true)); } -} // namespace facebook::velox::encoding \ No newline at end of file +TEST_F(Base32Test, calculateDecodedSizeProperSize) { + struct TestCase { + std::string encoded; + size_t initial_size; + int expected_decoded; + size_t expected_size; + }; + + std::vector test_cases = { + {"ME======", 8, 1, 2}, + {"ME", 2, 1, 2}, + {"MFRA====", 8, 2, 4}, + {"MFRGG===", 8, 3, 5}, + {"NBSWY3DPEB3W64TMMQ======", 24, 11, 18}, + {"NBSWY3DPEB3W64TMMQ", 18, 11, 18}}; + + for (const auto& test : test_cases) { + size_t encoded_size = test.initial_size; + EXPECT_EQ( + test.expected_decoded, + Base32::calculateDecodedSize(test.encoded.c_str(), encoded_size)); + EXPECT_EQ(test.expected_size, encoded_size); + } +} + +TEST_F(Base32Test, errorWhenDecodedStringPartiallyPadded) { + size_t encoded_size = 9; + EXPECT_THROW( + Base32::calculateDecodedSize("MFRA====", encoded_size), VeloxUserError); +} + +} // namespace facebook::velox::encoding diff --git a/velox/common/encode/tests/CMakeLists.txt b/velox/common/encode/tests/CMakeLists.txt index ea2d2463b876..4d6b8035ad37 100644 --- a/velox/common/encode/tests/CMakeLists.txt +++ b/velox/common/encode/tests/CMakeLists.txt @@ -12,7 +12,8 @@ # See the License for the specific language governing permissions and # limitations under the License. -add_executable(velox_common_encode_test Base32Test.cpp Base64Test.cpp EncoderUtilsTests.cpp) +add_executable(velox_common_encode_test Base32Test.cpp Base64Test.cpp + EncoderUtilsTests.cpp) add_test(velox_common_encode_test velox_common_encode_test) target_link_libraries( velox_common_encode_test diff --git a/velox/docs/functions/presto/binary.rst b/velox/docs/functions/presto/binary.rst index ed08dde84f00..5f20a24533b3 100644 --- a/velox/docs/functions/presto/binary.rst +++ b/velox/docs/functions/presto/binary.rst @@ -41,6 +41,28 @@ Binary Functions Decodes ``bigint`` value from a 64-bit 2’s complement big endian ``binary``. +.. function:: from_base32(string) -> varbinary + + Decodes a Base32-encoded ``string`` back into its original binary form. + This function can handle both padded and non-padded Base32 encoded strings. Partially padded Base32 strings will result in an error. + + Examples + -------- + Query with padded Base32 string: + :: + SELECT from_base32('JBSWY3DPEBLW64TMMQ======'); -- [72, 101, 108, 108, 111, 32, 87, 111, 114, 108, 100] + + Query with non-padded Base32 string: + :: + SELECT from_base32('JBSWY3DPEBLW64TMMQ'); -- [72, 101, 108, 108, 111, 32, 87, 111, 114, 108, 100] + + Query with partially padded Base32 string: + :: + SELECT from_base32('JBSWY3DPEBLW64TM=='); -- Error: Base32::decode() - invalid input string: length is not a multiple of 8. + + In the examples above, both fully padded and non-padded Base32 strings ('JBSWY3DPEBLW64TMMQ======' and 'JBSWY3DPEBLW64TMMQ') decode to the binary representation of the text 'Hello World'. + The partially padded Base32 string 'JBSWY3DPEBLW64TM==' will lead to a decoding error. + .. function:: from_hex(string) -> varbinary Decodes binary data from the hex encoded ``string``. diff --git a/velox/functions/prestosql/BinaryFunctions.h b/velox/functions/prestosql/BinaryFunctions.h index 009790211ac2..be99ca604483 100644 --- a/velox/functions/prestosql/BinaryFunctions.h +++ b/velox/functions/prestosql/BinaryFunctions.h @@ -340,6 +340,22 @@ struct ToBase64UrlFunction { } }; +template +struct FromBase32Function { + VELOX_DEFINE_FUNCTION_TYPES(TExec); + + // T can be either arg_type or arg_type. These are the + // same, but hard-coding one of them might be confusing. + template + FOLLY_ALWAYS_INLINE void call(out_type& result, const T& input) { + auto inputSize = input.size(); + result.resize( + encoding::Base32::calculateDecodedSize(input.data(), inputSize)); + encoding::Base32::decode( + input.data(), inputSize, result.data(), result.size()); + } +}; + template struct ToBase32Function { VELOX_DEFINE_FUNCTION_TYPES(T); diff --git a/velox/functions/prestosql/registration/BinaryFunctionsRegistration.cpp b/velox/functions/prestosql/registration/BinaryFunctionsRegistration.cpp index 8722df517e44..6300d13e9b81 100644 --- a/velox/functions/prestosql/registration/BinaryFunctionsRegistration.cpp +++ b/velox/functions/prestosql/registration/BinaryFunctionsRegistration.cpp @@ -57,6 +57,8 @@ void registerSimpleFunctions(const std::string& prefix) { {prefix + "from_base64url"}); registerFunction( {prefix + "to_base32"}); + registerFunction( + {prefix + "from_base32"}); registerFunction( {prefix + "from_big_endian_32"}); diff --git a/velox/functions/prestosql/tests/BinaryFunctionsTest.cpp b/velox/functions/prestosql/tests/BinaryFunctionsTest.cpp index b61ec1cfbf12..ccbf8df6d88a 100644 --- a/velox/functions/prestosql/tests/BinaryFunctionsTest.cpp +++ b/velox/functions/prestosql/tests/BinaryFunctionsTest.cpp @@ -502,6 +502,42 @@ TEST_F(BinaryFunctionsTest, toBase32) { toBase32("Hello World from Velox!")); } +TEST_F(BinaryFunctionsTest, fromBase32) { + const auto fromBase32 = [&](std::optional value) { + return evaluateOnce("from_base32(c0)", value); + }; + + EXPECT_EQ(std::nullopt, fromBase32(std::nullopt)); + EXPECT_EQ("", fromBase32("")); + EXPECT_EQ("a", fromBase32("ME======")); + EXPECT_EQ("ab", fromBase32("MFRA====")); + EXPECT_EQ("abc", fromBase32("MFRGG===")); + EXPECT_EQ("db2", fromBase32("MRRDE===")); + EXPECT_EQ("abcd", fromBase32("MFRGGZA=")); + EXPECT_EQ("hello world", fromBase32("NBSWY3DPEB3W64TMMQ======")); + EXPECT_EQ( + "Hello World from Velox!", + fromBase32("JBSWY3DPEBLW64TMMQQGM4TPNUQFMZLMN54CC===")); + + // Try encoded strings without padding + EXPECT_EQ("a", fromBase32("ME")); + EXPECT_EQ("ab", fromBase32("MFRA")); + EXPECT_EQ("abc", fromBase32("MFRGG")); + EXPECT_EQ("db2", fromBase32("MRRDE")); + EXPECT_EQ("abcd", fromBase32("MFRGGZA")); + EXPECT_EQ("1234", fromBase32("GEZDGNA")); + EXPECT_EQ("abcde", fromBase32("MFRGGZDF")); + EXPECT_EQ("abcdef", fromBase32("MFRGGZDFMY")); + + // Check with invaild encoded strings + EXPECT_THROW(fromBase32("1="), VeloxUserError); + EXPECT_THROW(fromBase32("M1======"), VeloxUserError); + + VELOX_ASSERT_THROW( + fromBase32("J1======"), + "decode() - invalid input string: invalid characters"); +} + TEST_F(BinaryFunctionsTest, fromBigEndian32) { const auto fromBigEndian32 = [&](const std::optional& arg) { return evaluateOnce("from_big_endian_32(c0)", VARBINARY(), arg);