Changelog

v0.3.6-29-g1465367

Changes

• Switch conda workflow to condadev environment. [Stephen Arnold]

• Swap out flake8 for cython-lint, update setup files, remove pep8 cfg. [Stephen Arnold]

Fixes

• Cleanup tests and fix a raw string test, enable more win32. [Stephen Arnold]

  • split all runners into separate arch via matrix

  • macos does need macos-14 to get a proper arm64 build

• Apply emptygroups fix and remove conda-only patch, also. [Stephen L Arnold]

  • release workflow: restrict pypi upload to repo owner

  • tox.ini: replace deprecated pep517 module, update deploy url

Other

• Fix #42. [Andreas van Cranenburgh]

• Include current notification level in cache key. [Andreas van Cranenburgh]

  this prevents a cached regular expression being used that was created with a different notification level.

  For example, the following now generates the expected warning:

  In [1]: import re2
  In [2]: re2.compile('a*+')
  Out[2]: re.compile('a*+')
  In [3]: re2.set_fallback_notification(re2.FALLBACK_WARNING)
  In [4]: re2.compile('a*+')
  <ipython-input-5-041122e221c7>:1: UserWarning: WARNING: Using re module. Reason: bad repetition operator: *+
    re2.compile('a*+')
  Out[4]: re.compile('a*+')
  

• Support fallback to Python re for possessive quantifiers. [Andreas van Cranenburgh]

• Document lack of support for possessive quantifiers and atomic groups. [Andreas van Cranenburgh]

• Make tests pass on my system; if this behavior turns out to be inconsistent across versions/platforms, maybe the test should be disabled altogether. #27. [Andreas van Cranenburgh]

• Add NOFLAGS and RegexFlags constants; #41. [Andreas van Cranenburgh]

• Remove python versions for make valgrind. [Andreas van Cranenburgh]

• Merge pull request #33 from sarnold/conda-patch. [Andreas van Cranenburgh]

  Conda patch for None vs empty string change

• Merge pull request #32 from JustAnotherArchivist/match-getitem. [Andreas van Cranenburgh]

  Make Match objects subscriptable

• Add test for Match subscripting. [JustAnotherArchivist]

• Make Match objects subscriptable. [JustAnotherArchivist]

  Fixes #31

v0.3.6 (2021-05-05)

• Merge pull request #30 from sarnold/release-pr. [Andreas van Cranenburgh]

  workflow updates

• Add missing sdist job and artifact check to workflows, bump version. [Stephen L Arnold]

• Bump version again. [Andreas van Cranenburgh]

• Merge pull request #28 from sarnold/use-action. [Andreas van Cranenburgh]

  Workflow cleanup

• Move pypi upload to end of release.yml, use gitchangelog action. [Stephen L Arnold]

v0.3.4 (2021-04-10)

Changes

• Ci: update workflows and tox cfg (use tox for smoke test) [Stephen L Arnold]

• Rename imported test helpers to avoid any discovery issues. [Stephen L Arnold]

Fixes

• Apply test patch, cleanup tox and pytest args. [Stephen L Arnold]

• Handle invalid escape sequence warnings, revert path changes. [Stephen L Arnold]

Other

• Bump version. [Andreas van Cranenburgh]

• Improve fix for #26. [Andreas van Cranenburgh]

• Add test for #26. [Andreas van Cranenburgh]

• Fix infinite loop on substitutions of empty matches; fixes #26. [Andreas van Cranenburgh]

• Fix “make test” and “make test2” [Andreas van Cranenburgh]

• Merge pull request #25 from sarnold/last-moves. [Andreas van Cranenburgh]

  Last moves

• Another one. [Andreas van Cranenburgh]

• Fix fix of conda patch. [Andreas van Cranenburgh]

• Fix conda patch. [Andreas van Cranenburgh]

• Fix “make test”; rename doctest files for autodetection. [Andreas van Cranenburgh]

• Fix narrow unicode detection. [Andreas van Cranenburgh]

• Merge pull request #24 from sarnold/tst-cleanup. [Andreas van Cranenburgh]

  Test cleanup

• Fix Python 2 compatibility. [Andreas van Cranenburgh]

• Use pytest; fixes #23. [Andreas van Cranenburgh]

• Tweak order of badges. [Andreas van Cranenburgh]

• Makefile: default to Python 3. [Andreas van Cranenburgh]

• Update README, fix Makefile. [Andreas van Cranenburgh]

• Merge pull request #22 from sarnold/missing-tests. [Andreas van Cranenburgh]

  add missing tests to sdist package, update readme and ci worflows (#1)

• Fix pickle_test (tests.test_re.ReTests) … ERROR (run tests with nose) [Stephen L Arnold]

• Update changelog (and trigger ci rebuild) [Stephen L Arnold]

• Add missing tests to sdist package, update readme and ci worflows (#1) [Steve Arnold]

  readme: update badges, merge install sections, fix some rendering issues

v0.3.3 (2021-01-26)

Changes

• Add .gitchangelog.rc and generated CHANGELOG.rst (keep HISTORY) [Stephen L Arnold]

• Ci: update wheel builds for Linux, Macos, and Windows. [Stephen L Arnold]

Fixes

• Ci: make sure wheel path is correct for uploading. [Stephen L Arnold]

Other

• Bump version. [Andreas van Cranenburgh]

• Update README.rst. fixes #21. [Andreas van Cranenburgh]

• Merge pull request #20 from freepn/new-bld. [Andreas van Cranenburgh]

  New cmake and pybind11 build setup

v0.3.2 (2020-12-16)

• Bump version. [Andreas van Cranenburgh]

• Merge pull request #18 from freepn/github-ci. [Andreas van Cranenburgh]

  workaroud for manylinux dependency install error plus release automation

v0.3.1 (2020-10-27)

• Bump version. [Andreas van Cranenburgh]

• Change package name for pypi. [Andreas van Cranenburgh]

v0.3 (2020-10-27)

• Bump version. [Andreas van Cranenburgh]

• Merge pull request #14 from yoav-orca/master. [Andreas van Cranenburgh]

  Support building wheels automatically using github actions

• Creating github actions for building wheels. [Yoav Alon]

• Created pyproject.toml. [Yoav Alon]

  Poetry and other modern build system need to know which build-tools to install prior to calling setup.py. added a pyproject.toml to specify cython as a dependency.

• Add contains() method. [Andreas van Cranenburgh]

  • contains() works like match() but returns a bool to avoid creating a Match object. see #12

  • add wrapper for re.Pattern so that contains() and count() methods are also available when falling back to re.

• Disable dubious tests. [Andreas van Cranenburgh]

  • All tests pass.

  • Don’t test for exotic/deprecated stuff such as non-initial flags in patterns and octal escapes without leading 0 or triple digits.

  • Known corner cases no longer reported as failed tests.

  • support b inside character class to mean backspace

  • use re.error instead of defining subclass RegexError; ensures that exceptions can be caught both in re2 and in a potential fallback to re.

• Disable failing test for known corner case. [Andreas van Cranenburgh]

• Remove tests with re.LOCALE flag since it is not allowed with str in Python 3.6+ [Andreas van Cranenburgh]

• Decode named groups even with bytes patterns; fixes #6. [Andreas van Cranenburgh]

• Make -std=c++11 the default; fixes #4. [Andreas van Cranenburgh]

• Merge pull request #5 from mayk93/master. [Andreas van Cranenburgh]

  Adding c++ 11 compile flag on Ubuntu

• Adding c++ 11 compile flag on Ubuntu. [Michael]

• Merge pull request #3 from podhmo/macports. [Andreas van Cranenburgh]

  macports support

• Macports support. [podhmo]

• Use STL map for unicodeindices. [Andreas van Cranenburgh]

• Only translate unicode indices when needed. [Andreas van Cranenburgh]

• Update README. [Andreas van Cranenburgh]

• Add -std=c++11 only for clang, because gcc on CentOS 6 does not support it. [Andreas van Cranenburgh]

• Disable non-matched group tests; irrelevant after dad49cd. [Andreas van Cranenburgh]

• Merge pull request #2 from messense/master. [Andreas van Cranenburgh]

  Fix groupdict decode bug

• Fix groupdict decode bug. [messense]

• Merge pull request #1 from pvaneynd/master. [Andreas van Cranenburgh]

  Ignore non-matched groups when replacing with sub

• Ignore non-matched groups when replacing with sub. [Peter Van Eynde]

  From 3.5 onwards sub() and subn() now replace unmatched groups with empty strings. See:

  https://docs.python.org/3/whatsnew/3.5.html#re

  This change removes the ‘unmatched group’ error which occurs when using re2.

• Fix setup.py unicode error. [Andreas van Cranenburgh]

• Add C++11 param; update URL. [Andreas van Cranenburgh]

• Fix bugs; ensure memory is released; simplify C++ interfacing; [Andreas van Cranenburgh]

  • Fix bug causing zero-length matches to be returned multiple times

  • Use Latin 1 encoding with RE2 when unicode not requested

  • Ensure memory is released: - put del calls in finally blocks - add missing del call for ‘matches’ array

  • Remove Cython hacks for C++ that are no longer needed; use const keyword that has been supported for some time. Fixes Cython 0.24 compilation issue.

  • Turn _re2.pxd into includes.pxi.

  • remove some tests that are specific to internal Python modules _sre and sre

• Fix Match repr. [Andreas van Cranenburgh]

• Add tests for bug with \b. [Andreas van Cranenburgh]

• Document support syntax &c. [Andreas van Cranenburgh]

  • add reference of supported syntax to main docstring

  • add __all__ attribute defining public members

  • add re’s purge() function

  • add tests for count method

  • switch order of prepare_pattern() and _compile()

  • rename prepare_pattern() to _prepare_pattern() to signal that it is semi-private

• Add count method. [Andreas van Cranenburgh]

  • add count method, equivalent to len(findall(…))

  • use arrays in utf8indices

  • tweak docstrings

• Move functions around. [Andreas van Cranenburgh]

• Improve substitutions, Python 3 compatibility. [Andreas van Cranenburgh]

  • when running under Python 3+, reject unicode patterns on bytes data, and vice versa, in according with general Python 3 behavior.

  • improve Match.expand() implementation.

  • The substitutions by RE2 behave differently from Python (character escapes, named groups, etc.), so use Match.expand() for anything but simple literal replacement strings.

  • make groupindex of pattern objects public.

  • add Pattern.fullmatch() method.

  • use #define PY2 from setup.py instead of #ifdef hack.

  • debug option for compilation.

  • use data() instead of c_str() on C++ strings, and always supply length, so that strings with null characters are supported.

  • bump minimum cython version due to use of bytearray typing

  • adapt tests to Python 3; add b and u string prefixes where needed, &c.

  • update README

• Add flags parameter to toplevel functions. [Andreas van Cranenburgh]

• Update performance table / missing features. [Andreas van Cranenburgh]

• Workaround for sub(…) with count > 1. [Andreas van Cranenburgh]

• Handle named groups in replacement string; &c. [Andreas van Cranenburgh]

  • handle named groups in replacement string

  • store index of named groups in Pattern object instead of Match object.

  • use bytearray for result in _subn_callback

• Pickle Patterns; non-char buffers; &c. [Andreas van Cranenburgh]

  • support pickling of Pattern objects

  • support buffers from objects that do not support char buffer (e.g., integer arrays); does not make a lot of sense, but this is what re does.

  • enable benchmarks shown in readme by default; fix typo.

  • fix typo in test_re.py

• New buffer API; precompute groups/spans; &c. [Andreas van Cranenburgh]

  • use new buffer API NB: even though the old buffer interface is deprecated from Python 2.6, the new buffer interface is only supported on mmap starting from Python 3.

  • avoid creating Match objects in findall()

  • precompute groups and spans of Match objects, so that possibly encoded version of search string (bytestr / cstring) does not need to be kept.

  • in _make_spans(), keep state for converting utf8 to unicode indices; so that there is no quadratic behavior on repeated invocations for different Match objects.

  • release GIL in pattern_Replace / pattern_GlobalReplace

  • prepare_pattern: loop over pattern as char *

  • advertise Python 3 support in setup.py, remove python 2.5

• Properly translate pos, endpos indices with unicode, &c. [Andreas van Cranenburgh]

  • properly translate pos, endpos indices with unicode

  • keep original unicode string in Match objects

  • separate compile.pxi file

• Re-organize code. [Andreas van Cranenburgh]

• Minor changes. [Andreas van Cranenburgh]

• Python 2/3 compatibility, support buffer objects, &c. [Andreas van Cranenburgh]

  • Python 2/3 compatibility

  • support searching in buffer objects (e.g., mmap)

  • add module docstring

  • some refactoring

  • remove outdated Cython-generated file

  • modify setup.py to cythonize as needed.

• Implement finditer as generator. [Andreas van Cranenburgh]

• Merge pull request #31 from sunu/master. [Michael Axiak]

  Add Python 3 support.

• Add Python 3 support. [Tarashish Mishra]

• Version bump. [Michael Axiak]

release/0.2.22 (2015-05-15)

• Version bump. [Michael Axiak]

• Merge pull request #22 from socketpair/release_gil_on_compile. [Michael Axiak]

  Release GIL during regex compilation.

• Release GIL during regex compilation. [Коренберг Марк]

  (src/re2.cpp is not regenerated in this commit)

release/0.2.21 (2015-05-14)

• Release bump. [Michael Axiak]

• Merge pull request #18 from offlinehacker/master. [Michael Axiak]

  setup.py: Continue with default lib paths if not detected automatically

• Setup.py: Continue with default lib paths if not detected automatically. [Jaka Hudoklin]

• Fix issue #11. [Michael Axiak]

• Remove spurious print statement. [Michael Axiak]

• Added version check in setup.py to prevent people from shooting themselves in the foot trying to compile with an old cython version. [Michael Axiak]

release/0.2.20 (2011-11-15)

• Version bump to 0.2.20. [Michael Axiak]

• Version bump to 0.2.18 and use MANIFEST.in since python broke how sdist works in 2.7.1 (but fixes it in 2.7.3…) [Michael Axiak]

release/0.2.16 (2011-11-08)

Fixes

• Unmatched group span (-1,-1) caused exception in _convert_pos. [Israel Tsadok]

• Last item in qualified split included the item before last. [Israel Tsadok]

• Exception in callback would cause a memory leak. [Israel Tsadok]

• Group spans need to be translated to their relative decoded positions. [Israel Tsadok]

• This is not what verbose means in this context. [Israel Tsadok]

• Findall used group(0) instead of group(1) when there was a group. [Israel Tsadok]

• Dangling reference when _subn_callback breaks on limit. [Israel Tsadok]

• Infinite loop in pathological case of findall(“.*”, “foo”) [Israel Tsadok]

Other

• Version bump to 0.2.16. [Michael Axiak]

• Merged itsadok’s changes to fix treatment of D and W. Added tests to reflection issue #4. [Michael Axiak]

• Fixed issue #5, support W, S and D. [Israel Tsadok]

• Fixed issue #3, changed code to work with new re2 api. [Michael Axiak]

• Merge branch ‘itsadok-master’ [Michael Axiak]

• Merge branch ‘master’ of https://github.com/itsadok/pyre2 into itsadok-master. [Michael Axiak]

• Failing tests for pos and endpos. [Israel Tsadok]

• Set default notification to FALLBACK_QUIETLY, as per the documentation. [Israel Tsadok]

• Get rid of deprecation warning. [Israel Tsadok]

• Fix hang on findall(‘’, ‘x’) [Israel Tsadok]

• Allow weak reference to Pattern object. [Israel Tsadok]

• Allow named groups in span(), convert all unicode positions in one scan. [Israel Tsadok]

• Added failing test for named groups. [Israel Tsadok]

• Fix lastgroup and lastindex. [Israel Tsadok]

• Verify that flags do not get silently ignored with compiled patterns. [Israel Tsadok]

• Had to cheat on a test, since we can’t support arbitrary bytes. [Israel Tsadok]

• Fix lastindex. [Israel Tsadok]

• Pass some more tests - added pos, endpos, regs, re attributes to Match object. [Israel Tsadok]

• Added max_mem parameter, bumped version to 0.2.13. [Michael Axiak]

• Remove spurious get_authors() call. [Michael Axiak]

• Added Alex to the authors file. [Michael Axiak]

• Version bumped to 0.2.11. [Michael Axiak]

• Added difference in version to changelist, added AUTHORS parsing to setup.py. [Michael Axiak]

• Added check for array, added synonym ‘error’ for RegexError to help pass more python tests. [Michael Axiak]

• Made make test run a little nicer. [Michael Axiak]

• Added note about copyright assignment to authors. [Michael Axiak]

• Fix test_re_match. [Israel Tsadok]

• Fix test_bug_1140. [Israel Tsadok]

• Update readme. [Israel Tsadok]

• Preprocess pattern to match re2 quirks. Fixes several bugs. [Israel Tsadok]

• Re2 doesn’t like when you escape non-ascii chars. [Israel Tsadok]

• Merge remote branch ‘moreati/master’ [Israel Tsadok]

• Merge from axiak/HEAD. [Alex Willmer]

• Merge from axiak/master. [Alex Willmer]

• Pass ErrorBadEscape patterns to re.compile(). Have re.compile() accept SRE objects. [Alex Willmer]

• Ignore .swp files. [Alex Willmer]

• Remove superfluous differences to axiak/master. [Alex Willmer]

• Merge remote branch ‘upstream/master’ [Alex Willmer]

• Merge changes from axiak master. [Alex Willmer]

• Fix previous additions to setup.py. [Alex Willmer]

• Add url to setup.py. [Alex Willmer]

• Remove #! line from re.py module, since it isn’t a script. [Alex Willmer]

• Add classifers and long description to setup.py. [Alex Willmer]

• Add MANIFEST.in. [Alex Willmer]

• Merge branch ‘master’ of git://github.com/facebook/pyre2. [Alex Willmer]

  Conflicts:

  re2.py

• Ignore byte compiled python files. [Alex Willmer]

• Add copyright, license, and three convenience functions to re2.py. [Alex Willmer]

• Switch re2.h include to angle brackets. [Alex Willmer]

• Handle n in replace template, since re2 doesn’t. [Israel Tsadok]

• Import escape function as it from re. [Israel Tsadok]

• The unicode char classes never worked. TIL testing is important. [Israel Tsadok]

• Make unicode stickyness rules like in re module. [Israel Tsadok]

• Return an iterator from finditer. [Israel Tsadok]

• Have re.compile() accept SRE objects. (copied from moreati’s fork) [Israel Tsadok]

• Fix var name mixup. [Israel Tsadok]

• Added self to authors. [Israel Tsadok]

• Fix memory leak. [Israel Tsadok]

• Added re unit test from python2.6. [Israel Tsadok]

• Make match group allocation more RAII style. [Israel Tsadok]

• Use None for unused groups in split. [Israel Tsadok]

• Change split to match sre.c implementation, and handle empty matches. [Israel Tsadok]

• Match delete[] to new[] calls. [Israel Tsadok]

• Added group property to match re API. [Israel Tsadok]

• Use appropriate char classes in case of re.UNICODE. [Israel Tsadok]

• Fall back on re in case of back references. [Israel Tsadok]

• Use unmangled pattern in case of fallback to re. [Israel Tsadok]

• Simplistic cache, copied from re.py. [Israel Tsadok]

• Fix some memory leaks. [Israel Tsadok]

• Allow multiple arguments to group() [Israel Tsadok]

• Allow multiple arguments to group() [Israel Tsadok]

• Fixed path issues. [Michael Axiak]

• Added flags to pattern object, bumped version number. [Michael Axiak]

• Change import path to fix include dirs. [Michael Axiak]

• Updated manifest and changelog. [Michael Axiak]

• Added .expand() to group objects. [Michael Axiak]

• Removed the excessive thanks. [Michael Axiak]

• Added regex module to performance tests. [Michael Axiak]

• Pattern objects should be able to return the input pattern. [Alec Berryman]

  I’m just storing the original object passed in (found it could be str or unicode - thanks, unicode tests!). It could be calculated if important to save space.

• Further findall fix: one-match finds. [Alec Berryman]

  I read the spec more carefully this time.

• Added changelist file for future releases. [Michael Axiak]

• Added alec to AUTHORS. Bumped potential version number. Fixed findall support to work without closures. [Michael Axiak]

• Make pyre2 64-bit safe. [Alec Berryman]

  Now compiles on a 64-bit system; previously, complained that you might have a string size that couldn’t fit into an int. Py_ssize_t is required instead of plain size_t because Python’s is signed.

• Fix findall behavior to match re. [Alec Berryman]

  findall is to return a list of strings or tuples of strings, while finditer is to return an iterator of match objects; previously both were returning lists of match objects. findall was fixed, but finditer is still returning a list.

  New tests.

• Makefile: test target. [Alec Berryman]

• Note Cython 0.13 dependency. [Alec Berryman]

• Moved to my own license. I can do that since I started the code from scratch. [Michael Axiak]

• Fix formatting of table. [Michael Axiak]

• Added number of trials, fixed some wording. [Michael Axiak]

• Added missing files to MANIFEST. [Michael Axiak]

• Added performance testing, bumped version number. [Michael Axiak]

• Added readme data. [Michael Axiak]

• Updated installation language. [Michael Axiak]

• Incremented version number. [Michael Axiak]

• Fixed split regression with missing unicode translation. [Michael Axiak]

• Maybe utf8 works? [Michael Axiak]

• Got unicode to past broken test. [Michael Axiak]

• Added debug message for issue with decoding. [Michael Axiak]

• Added more utf8 stuff. [Michael Axiak]

• Delay group construction for faster match testing. [Michael Axiak]

• Got utf8 support to work. [Michael Axiak]

• Starting to add unicode support… BROKEN. [Michael Axiak]

• Added fallback and notification. [Michael Axiak]

• Added runtime library path to fix /usr/local/ bug for installation of re2 library. [Michael Axiak]

• Fixed setup and updated readme. [Michael Axiak]

• Added more setup.py antics. [Michael Axiak]

• Added ancillary stuff. [Michael Axiak]

• Added sub, subn, findall, finditer, split. Bam. [Michael Axiak]

• Added finditer along with tests. [Michael Axiak]

• Fix rst. [Michael Axiak]

• Added contact link to readme. [Michael Axiak]

• Fixed formatting for readme. [Michael Axiak]

• Added lastindex, lastgroup, and updated README file. [Michael Axiak]

• Added some simple tests. Got match() to work, named groups to work. spanning to work. pretty much at the same level as pyre2 proper. [Michael Axiak]

• Added more things to gitignore, another compile. [Michael Axiak]

• Move code to src directory for better sanity. [Michael Axiak]

• Updated re2 module to use Cython instead, got matching working almost completely with minimal flag support. [Michael Axiak]

• Suppress logging of failed pattern compilation. [David Reiss]

  This isn’t necessary in Python since errors are reported by exceptions. Use a slightly more verbose form to allow easier setting of more options later.

• Don’t define tp_new for regexps and matches. [David Reiss]

  It turns out that these are not required to use PyObject_New. They are only required to allow Python code to call the type to create a new instance, which I don’t want to allow. Remove them.

• Fix a segfault by initializing attr_dict to NULL. [David Reiss]

  I thought PyObject_New would call my tp_new, which is set to PyType_GenericNew, which NULLs out all user-defined fields (actually memset, but the docs say NULL). However, this appears to not be the case. Explicitly NULL out attr_dict in create_regexp to avoid segfaults when compiling a bad pattern. Do it for create_match too for future safety, even though it is not required with the current code.

• Add struct tags for easier debugging with gdb. [David Reiss]

  See http://sourceware.org/ml/archer/2009-q1/msg00085.html for a little more information.

• Add some convenience functions to re2.py. [Alex Willmer]

• Add a copyright and license comment to re2.py. [David Reiss]

  This is to prepare for adding some not-totally-trivial code to it.

• Use PEP-8-compliant 4-space indent in re2.py. [David Reiss]

• Use angle-brackets for the re2.h include. [Alex Willmer]

• Add build information to the README. [David Reiss]

• Add contact info to README. [David Reiss]

• Initial commit of pyre2. [David Reiss]