* [C++] Removed most heap allocations in builder
* Updated API docs to indicate heap usage
* Override assertion for heap allocation in parser
* Cleaned up implemenations, enable heap alloc for tests
* Generalized CreateVectorOfStrings
* remove cmake option for heap alloc. reverted two heap removals
* Only use scratch space for vector of strings
* Override Windows SCL warning
* Changed std::transform to for loop to avoid MSCV warnings
* switched to const iterators
* Replaced iterator with for loop
* remove std::to_string in test to be compatible
* [idl_parser] Check structs and enums do not clash in a namespace
Uses fully qualified names to check for clashes within a given namespace whether explicitly defined or in the global namespace.
* [idl_parser] Move type name clash check to ParseEnum and ParseDecl
Change point at which parsing error is returned to ensure error is caught more generally. This change means that the error is returned after parsing the entirety of the offending duplicate rather than at the start when parsing it's name.
* [idl_parser] Add single and multi file type name clash tests
Adds a selection of tests for valid single file schemas with types that have the same name but are in different namespaces.
Adds a test for an a valid schema that spans two files with two types that have the same name but are in different namespaces.
Adds a test for an an invalid schema that spans two files with two types that have the same name and are in the same namespace.
* Implement Serialize for flexbuffer::Reader
* bump version
* Remove use of experimantal or-patterns
* Remove more use of experimantal or-patterns
Co-authored-by: Casper Neo <cneo@google.com>
This commit fixes two serious issues connected with reverse iterators:
1. Vector's rbegin/rend was incompatible with std::reverse_iterator::base();
2. Array's rend() was incorrect, it based on end() instead of begin().
This adds basic support for different Lua versions.
For Lua 5.2 and Lua 5.3, both the Bit32 and Compat53 Lua modules must be
installed for it to work. You can typically get these on Linux using
apt install lua-compat53 lua-bit32
For Lua 5.4, it should work as is, as it is a clean superset of Lua 5.3,
which is what the original Lua Flatbuffers supported.
LuaJIT as installed by apt doesn't include the COMPAT mode for using
Lua5.2 features. So this change just makes the flatbuffer code use a
common implementation.
* [Lua] Add LuaJIT support
Here is the output of LuaTest.sh:
Run with LuaJIT:
built 100 512-byte flatbuffers in 0.16sec: 0.63/msec, 0.31MB/sec
built 1000 512-byte flatbuffers in 0.08sec: 12.06/msec, 5.89MB/sec
built 10000 512-byte flatbuffers in 0.80sec: 12.44/msec, 6.07MB/sec
built 10000 512-byte flatbuffers in 0.33sec: 30.58/msec, 14.93MB/sec
traversed 100 592-byte flatbuffers in 0.04sec: 2.51/msec, 1.42MB/sec
traversed 1000 592-byte flatbuffers in 0.03sec: 31.52/msec, 17.79MB/sec
traversed 10000 592-byte flatbuffers in 0.21sec: 48.77/msec, 27.53MB/sec
Run with Lua 5.3:
built 100 512-byte flatbuffers in 0.02sec: 5.44/msec, 2.66MB/sec
built 1000 512-byte flatbuffers in 0.17sec: 5.74/msec, 2.80MB/sec
built 10000 512-byte flatbuffers in 1.75sec: 5.72/msec, 2.79MB/sec
built 10000 512-byte flatbuffers in 1.38sec: 7.26/msec, 3.55MB/sec
traversed 100 592-byte flatbuffers in 0.00sec: 27.64/msec, 15.60MB/sec
traversed 1000 592-byte flatbuffers in 0.03sec: 30.46/msec, 17.20MB/sec
traversed 10000 592-byte flatbuffers in 0.34sec: 29.62/msec, 16.72MB/sec
* [Lua] Better usage description
* update according to the review
* Add advance feature indicators to reflection
* deserialize too
* model advanced features as bitflags
* use uint64_t instead of AdvancedFeatures
* git clang format
* initialize advanced_features_
* remove whitespace
Co-authored-by: Casper Neo <cneo@google.com>
* Fix Miri flag passing and bump Rust version.
* Fix Miri problems from Arrays PR.
SafeSliceAccess was removed for Arrays. It's kind of unsound.
It has two properties:
1. EndianSafe
2. Alignment 1
We only need 1. in create_vector_direct to memcpy data.
We both 1. and 2. for accessing things with slices as buffers are built on &[u8]
which is unaligned. Conditional compilation implements
SafeSliceAccess for >1byte scalars (like f32) on LittleEndian machines
which is wrong since they don't satisfy 2.
This UB is still accessible for Vectors (though not exercised our
tests) as it implements SafeSliceAccess. I'll fix this later by
splitting SafeSliceAccess into its 2 properties.
Co-authored-by: Casper Neo <cneo@google.com>
* Mark endian_scalar as unsafe.
Also
- removed the deprecated flexbuffer slice from example
- fixed some cargo warnings
* Assertions and read_scalar made unsafe
* Clippy lints
* Add to Safety
Co-authored-by: Casper Neo <cneo@google.com>
* Add support for fixed size arrays
* clang-format
* Update rust image to 1.51 to support const generics
* Handle correctly big endian
* Add fuzz tests and clean code
* Add struct fuzz test and optimize struct arrays for api
* Bump flatbuffers crate version
* Add Build TS to CI jobs
* Add npm compile step
* Fix syntax
* Add required code gen for TS
* Exit on failure
* Make TS code gen identical to test run
* Remove duplicate TS code gen artifacts
* Remove bad gitignore
* Forgot parts of generate_code and make sure same settings for test run
* Don't forget about the bat
* Try and fix flatc args
* Another attempt to fix args
* Fix typo
* Commit up to date code gen
* Another attempt to fix sh/bat
* Move -o param to after -I
* Commit missing code gen file
* Fix grpc code gen and test
* Another grpc code gen fix
* Rework to not use ts folder
* Fix env vars in bat and make less churn
* Move TS code gen to dedicated folder
* Fix transpilation output folder and module paths
* Fixes to code gen
* Include generated js
* Moved ts code gen
* Remove test ts code gen folder
* disable clippy
* Vector of enum default
* swift and tests
* git clang format
* Rewrite enum parser checks
* Remove Voids from more_defaults
* vector enum swift
* remove vector accessor from swift
* clang format
Co-authored-by: Casper Neo <cneo@google.com>
* [C++] #6501 - Problem when mapping a native type multiple times
- idl.h:
added "native_type_pack_name"
- flatbuffers.h:
added CreateVectorOfNativeStructs variants which receive a pointer to the serialization function
- idl_gen_cpp.cpp:
adapted code generation in case "native_type_pack_name" attribute is present
- extended tests & docs; improved surrounding native_type docs a little
* integrated review feedback
* [idl_parser] Add kTokenNumericConstant token
This commit adds the new token for correct parsing of signed numeric constants.
Before this expressions `-nan` or `-inf` were treated as kTokenStringConstant.
This was ambiguous if a real string field parsed.
For example, `{ "text_field" : -name }` was accepted by the parser as valid JSON object.
Related oss-fuzz issue: 6200301176619008
* Add additional positive tests fo 'inf' and 'nan' as identifiers
* Rebase to HEAD
* Move processing of signed constants to ParseSingleValue method.
* Add missed `--cpp-static-reflection` (#6324) to pass CI
* Remove `flatbuffers.pc` from repository to unblock CI (#6455).
Probably the generated flatbuffers.pc should not be a part of repo.
* Fix FieldIdentifierTest()
* [C++17] Add compile-time reflection for fields.
Included in this commit is the following:
- The C++ generator has been modified so that,
when in C++17 mode, it will emit Table and
Struct field traits that can be used at com-
pile time as a form of static reflection. This
includes field types, field names, and a tuple
of field getter results.
- Diffs to the cpp17 generated files. No other
generated files are affected.
- A unit test that also serves as an example. It
demonstrates how to use the full power of this
reflection to implement a full recursive
JSON-like stringifier for Flatbuffers types,
but without needing any runtime access to the
*.fbs definition files; the computation is
done using only static reflection.
Tested on Linux with gcc 10.2.0.
Fixes#6285.
* Fix int-conversion warning on MSVC.
* Try to fix std::to_string ambiguity on MSVC.
* Fix clang-format diffs.
* Fix more clang-format diffs.
* Fix last clang-format diff.
* Enable C++17 build/test for VC 19 platform in CI.
* Forgot to add value to cmake command line variable.
* Various fixes/changes in response to @vglavnyy's feedback.
* Replace "fields pack" with index-based getters.
* Fix MSVC error.
* Fix clang-format diffs.
* getter_for method returns result instead of address-of-getter.
* Next round of reviewer suggestions.
* Use type instead of hardcoded struct name.
* Fix clang-format diff.
* Add test for FieldType since it is not used in the stringify test.
* Add fields_number field to Traits struct.
* Add --cpp-static-reflection flag and put those features behind it.
* Fix clang-format diffs.
* Remove <tuple> include.
* [idl_gen] Delete ts::GenPrefixedImport()
I don't know what this is for, but it's the only piece of code external
to idl_parser.cpp that expects the key of Parser::included_files_ to be
a path. And it appears to be unused.
* [idl_parser] Track included files by hash
Parser::included_files_ is a map whose main purpose is to keep track of
which files have already been parsed in order to protect against
multiple inclusion. Its key is the path that the file was found at
during parsing (or, if it's an in-memory file, just its name).
This commit changes the key to be the 64 bit FNV-1a hash of the file's
name (just the name, not the complete path) xor'd with the hash of the
file's contents (unless it's an in-memory file, then we only hash the
name.)
This allows multiple include protection to function even in the face of
unique per-file include paths (fixes#6425).
* Ran tests/generate_code.sh
CI told me to do it.
* Gracefullt handle case where source_filename == nullptr
I just learned source_filename might also be null. In that case, we
should exclude it from the hash instead of hashing a zero length
string, just like we exclude source when it is null.
Presumably nameless files will never be included (they can't, can they?)
so this doesn't really matter, but I think it's prettier anyways.
