pyenchant

Add Python 3.10 support

• Fix Dict.__del__ sometimes raising TypeError upon exit (#98). Patch by @rr-
• Default development branch is now called main
• Bump black to 21.6b0

• Add support for Python 3.9
• Add trove classifiers for all supported Python versions
• Run pyupgrade across the code base
• Update documentation about MacPorts

• On Windows, set PATH instead of calling SetDllDirectory before loading the
  Enchant C library. This allows PyEnchant to co-exist with other libraries in the same program. Fix #207.

• Add enchant.get_user_config_dir()
• Fix: enchant.get_enchant_version() now returns a str, not some bytes

• Add missing LICENSE.txt in source distribution

Highlights

• Uncouple PyEnchant version from the Enchant version. This release should be compatible with Enchant 1.6 to 2.2

• Fix using PyEnchant with Enchant >= 2.0

• Add support for pypy3, Python 3.7 and Python 3.8

• New website, hosted on https://pyenchant.github.io/pyenchant/

• Add enchant.set_prefix_dir()

Breaking changes

• Drop support for Python2

• macOS: The C enchant library is no longer embedded inside the wheel - you should install the C enchant library with brew or ports.

Clean ups

• Port test suite to pytest.

• Add tbump configuration to simplify the release process

• Format code with black.

• Remove compatibility layers with Python2 from enchant.utils

• Use flake8 to catch some errors during CI

• Fix some PEP8 naming violations

• Switch to GitHub Actions for CI

• Removed deprecated is_in_session method, for compatibility with enchant 2.

• New website and documentation, generated with Hyde and Sphinx.

• Fix SpellChecker.replace() when the replacement is shorter than the erroneous word; previously this would corrupt the internal state of the tokenizer. Thanks Steckelfisch.

• Make Dict class pickle-safe. Among other things, this should help with strange deadlocks when used with the multiprocessing module.

• Ability to import the module even when the enchant C library isn't installed, by setting PYENCHANT_IGNORE_MISSING_LIB env var.

• New utility function "trim_suggestions", useful for trimming the list of suggestions to a fixed maximum length.

• Change the way DeprecationWarnings are issued, to point to the line line in user code rather than inside pyenchant. Thanks eriolv.

• Add GetSpellChecker() method to wxSpellCheckerDialog. Thanks bjosey.

• restore compatibility with Python 3 (including 3.2 beta1).

• fix unittest DeprecationWarnings on Python 3.

• statically compile libstdc++ into pre-built windows binaries.

• DictWithPWL: use pwl and pel to adjust the words returned by suggest().

• Fix tokenization of utf8 bytes in a mutable character array.

• get_tokenizer(): pass None as language tag to get default tokenizer.

• prevent build-related files from being included in the source tarball.

• Bundle pre-compiled libraries for Mac OSX 10.4 and later.

• Improved handling of unicode paths on win32.

• Changed DLL loading logic for win32, to ensure that we don't accidentally load older versions of e.g. glib that may be on the DLL search path.

• Added function get_enchant_version() to retrieve the version string for the underlying enchant library.

• Upgraded bundled enchant to v1.6.0.

• Fixed bug in printf() utility function; all input args are now converted to strings before printing.

• Fixed loading of enchant DLL on win32 without pkg_resources installed.

• Fixed HTMLChunker to handle unescaped < and > characters that are clearly not part of a tag.

• Upgraded to enchant v1.5.0:

  • new Broker methods get_param() and set_param() allow runtime customisation of provider data

• Added the concept of 'chunkers' to enchant.tokenize.get_tokenizer(). These serve split split the text into large chunks of checkable tokens.

• implemented a simple HTMLChunker class

• Moved error classes into 'enchant.errors' for easier importing

• Moved testcases into separate files so they're not loaded by default

• Allowed SpellChecker to use default language if none is specified

• Improved compatibility with Python 3

• Fixed termination conditions in English tokenization loop.

• Improved unicode detection in English tokenizer.

• Made enchant spellcheck all of its docstrings as part of the unittest suite.

• Modify utils.get_resource_filename and utils.win32_data_files for compatibility with py2exe (which was broken in the move to ctypes). Thanks to Stephen George for the fix.

• SpellChecker.add_to_personal renamed to SpellChecker.add and fixed to use the corresponding Dict method.

• Migrated from SWIG to ctypes

  • now runs under PyPy!

  • also opens possibilities for Jython, IronPython, ...

• Compatibility updates for Python 3.0, mostly around unicode strings

• Dropped compatibility with Python 2.2

• upgrade to enchant v1.4.2

• windows version can now be installed at a path containing unicode characters

• treat combining unicode marks as letters during tokenization

• cleanup of wxSpellCheckerDialog, thanks to Phil Mayes

• upgrades of bundled components in Windows version

  • upgraded glib DLLs

  • latest dictionaries from OpenOffice.org

  • latest version of Hunspell

• Re-worked the tokenization API to allow filters but still remove non-alpha-numeric characters from words by default. This introduces some minor backward-incompatibilities to the API, hence the full minor version bump.

  • 'fallback' argument to get_tokenizer() was removed, just catch the Error and re-try with whatever is appropriate for your application.

  • filters should be passed into get_tokenizer() as the second argument, rather than applied as separate functions.

  • Basic whitespace-and-punctuation tokenization separated from the language-specific parts.

  • Internal details of Filter classes expanded and generalized

  • English tokenization rules reverted to 1.1.5 version

• Implemented "filters" that allow tokenization to skip common word forms such as URLs, WikiWords, email addresses etc.

• Now ships with enchant-1.3.0, meaning:

  • PWLs can return a useful list of suggestions rather than the empty list

  • Hunspell replaces MySpell as the default Windows backend

• Tokenization doesn't split words at non-alpha characters by default

• GtkSpellCheckerDialog contributed by Fredrik Corneliusson

• Removed deprecated functionality:

  • Dict.add_to_personal

  • All registry handling functionality from enchant.utils

  • enchant.utils.SpellChecker (use enchant.checker.SpellChecker)

• Removed PyPWL, as native enchant PWLs can now suggest corrections

• Fix hang in included MySpell (Windows distribution)

• Workaround for some MySpell/unicode problems

• Update to latest setuptools ez_setup.py

• No longer need to use the registry under Windows

• Moved to setuptools for managing distribution

• Implemented unittest TestCases, works with python setup.py test

• Plugins on Windows moved to "enchant" subdirectory

• SpellChecker now coerces to/from unicode automatically

• Use python default encoding rather than UTF-8 where appropriate

• Various documentation cleanups

• bug fixes:

  • (1230151): count of live instances done by normalized key

  • Accept unicode strings as broker orderings