CHANGELOG.md

textX changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Everything that is documented in the official docs is considered the part of the public API.

Backward incompatible changes are marked with (BIC). These changes are the reason for the major version increase so when upgrading between major versions please take a look at related PRs and issues and see if the change affects you.

Unreleased

Added

• nchar attribute to TextXError class, which represents the substring length for the model object where the error was found. Also, this value is now returned by get_location(), so no changes are required in user code. See 420. Thanks @davidchall.
• linetype parameter to PlantUML generator which controls line style. See the tip in the visualization docs.

Fixed

• Rendering of multiplicities of association in PlantUML generator.

Changed

• Meta-model diagram export now raises attributes/references to abstract meta-classes along the inheritance chain. See 423.
• Documentation migrated to mdbook and GitHub Actions.

4.0.1 (released: 2023-11-12)

Fixed

• Use flit-core for test projects. See 418. Thanks @mgorny for reporting.

4.0.0 (released: 2023-11-10)

Fixed

• Removed word boundary from INT rule regex. Thanks @NevenaAl for reporting the issue and providing the test (401).

Changed

• (BIC) Removed Python support for 3.6 and 3.7. The minimal supported version is 3.8.
• Migrated to pyproject.toml for project configuration. See (416).
• Migrated to importlib for language/generator registration/discovery as setuptools pkg_resources is deprecated and unsupported from Python 3.12. See 411. Thanks @EmilyBourne and @nchammas for reporting. See (416).
• (BIC) Added dependency to importlib-metadata backward compatibility package for Python versions < 3.10. If Python is >=3.10 this library is not used. See (416).
• Use ruff instead of flake8 for linting. ruff can also be used for code formating. See (416).
• Relaxed dependency for click to accept 8.x. See 414. Thanks @RootMeanSqr for reporting the issue. See (416).
• Use flit for package building and publishing.

3.1.1 (released: 2023-02-10)

Fixed

• Source distribution packaging issue ([392]). Thanks yurivict@GitHub.

3.1.0 (released: 2023-02-08)

Fixed

• Fixed RREL lookup in case of multi-meta models (some special cases were not handled correctly; #379).
• Fixed test suite invocation to use pytest over py.test that stopped working in pytest-7.2.0. (#389). Thanks mgorny@GitHub.

Changed

• Changed separator in obj. rule refs from | to :. Old separator will still be allowed until version 4.0. (#385, #384)
• Removed the dependency on future package (#388). Thanks mgorny@GitHub.
• Removed vendored six library. We don't need 2.x support anymore. Thanks davidchall@GitHub for reporting the issue (#390).

3.0.0 (released: 2022-03-20)

Added

• Added RREL-'fixed name'-extension, allowing to follow model elements given a fixed name (e.g. an object defined in a builtin model). Details described in rrel.md (#364).
• Added ability to access the full path of named objects traversed while resolving a RREL expression (#304).
• Added decorator textx.textxerror_wrap for object processors to automatically transform non-TextXErrors to TextXErrors in order to indicate the filename and position of the element being processed (#306).

Fixed

• model_param_defs on TextXMetaMetaModel (#360).
• Interpreting of backslash special chars (e.g. \n, \t) in grammar string matches (#323). Possible (BIC) - backslash chars were not interpreted in grammar files and raw Python strings prior to this fix.
• Exception/error messages (#320)
• Relaxed assert in model creation enabling some model changes in user classes (#311)
• Model export to dot in cases where textX object is replaced in the processor(#301)
• Do not allow "empty" RREL expressions (compare unittests in test_rrel.py; #355)

Changed

• Inheritance chain calculation. Possible (BIC) (#369).
• Added def_file_name attribute to RefRulePosition for storing the definition's model file name in case of cross-references between models. (#313,#277)
• Migrated from Travis CI to GitHub Actions (#307)
• Dropped support for deprecated Python versions. The lowest supported version is 3.6. (BIC)

2.3.0 (released: 2020-11-01)

Added

• textx generate. Documented passing in arbitrary parameters which can be used in the generator function. Also, implemented passing of model parameters defined in the meta-model (model_param_defs and _tx_model_params) (#299)
• get_location function for producing line/col/filename from any textX object. (#294)
• builtin_models of type ModelRepository to meta-model constructor. Used to supply pre-loaded models for ImportURI based scoping providers as a fallback to search into. (#284)
• Initial implementation of TEP-001 (#111) allowing to specify scope provider behavior within the grammar itself. #274 and #287 introduce the RREL (reference resolving expression language) to define how references are resolved. Details see rrel.md.
• Parameter should_follow of callable type to get_children and get_children_of_type model API calls to decide if the element should be traversed. (#281)

Fixed

• Fixed bug with Falsy user classes in get_children (#288)
• Fixed bug with unhashable objects during dot export (#283)
• Fixed bug where (Ext)RelativeName scope providers accepted any referenced object that contained the lookup name in its name. Thanks ipa-mdl@GitHub (#267)
• Fixed bug in flow_dsl test project causing static files not being included in package build/installation. Thanks sebix@GitHub (#272).
• Fixed bug, where user classes not used in the grammar caused exceptions (#270): now, when passing a list of user classes, you need to use them in your grammar. You can alternatively also pass a callable (see metamodel.md; #273). Also, using base classes for rules from imported grammars in conjunction with user classes is not allowed and results in an exception.
• Fixed bug in export.py concerning html escaping in the dot export of a textx meta-model (#276).

Changed

• _tx_model_param_definitions deprecated in favor of model_param_defs (#298).
• click is now an optional dependency, only when CLI is needed (#292)
• Make warning about not overwriting generated file more visible (01341ec3)
• Truncate long strings during dot export for better diagram readability (#282).
• Changed unhashable type exception when a list is used for name attributes by raising a more informative exception and extending docs to document the issue and a proper way to solve it (#40, #266).

v2.2.0 (released: 2020-08-03)

Added

• Initial docs for Jinja code generator support (from textX-jinja) (#264).
• Analyzing grammars programmatically as plain textX models (grammar_model_from_str/file) (#235)
• Initial startproject scaffolding (from textX-dev) docs (#234)
• Generator helper functions get_output_filename and gen_file (#233)
• textx version command (#219)
• Versions for languages/packages in list-languages and list-generators commands (#228)
• Added the ability to specify extra parameters during model_from_file or model_from_str and to define which extra parameters exist in the meta-model (#243).

Fixed

• Fixed several instances of invalid truthiness checking. Thanks markusschmaus@GitHub (#250)
• Fixed applying multiple grammar rule modifiers (#246)
• Fixed exception on calling check CLI command with relative path name.
• Fixed return value of textx generate and check commands: we return a failure on error now (#222)
• Fixed type checking for references to builtin elements (#218)

Changed

• User classes can now be immutable (e.g. attr.frozen) or can use __slots__. Thanks markusschmaus@GitHub ([#256, #260, #261])
• Cleanup of setup configuration and install scripts #231
• Dot/PlantUML rendering of meta-models: remove rendering of base types, improve rendering of required/optional, render match rules as a single table. (#225)
• Allow passing kwargs in metamodel_for_file/language registration API calls. (#224)

v2.1.0 (released: 2019-10-12)

Added

• Added new function textx.scoping.is_file_included (#197).

Changed

• Allow passing kwargs (specially - file_name) argument when loading metamodel from string (needed for textX-LS v0.1.0) (#211).
• Changed the parser rule for regex matches. Spaces are not stripped any more from the beginning and the end of the regexp-pattern. This could be possible BIC for some special cases #208.
• Changed function name textx.scoping.get_all_models_including_attached_models to textx.scoping.get_included_models (marked old function as deprecated) (#197).
• Delete all models touched while loading a model, when an error occurs while loading in all repositories (strong exception safety guarantee) (#200).

v2.0.1 (released: 2019-05-20)

Added

• Registration and discovery of languages and generators (#187)
• New textx CLI commands for listing generators and languages (list-generators, list-languages) and calling a generator (generate) (#187)
• Meta-models may now reference other registered meta-models using the reference statement (#187)
• Adding examples and documentation related to scope providers (related to model modification through scope providers) (#168)
• metamodel export feature for PlantUML (#165)
• textx_isinstance from textx.scoping.tools made available in textx (#164, #157)
• CLI extensibility (#162, #161)
• An initial version of FAQ page (#138). Thanks Aluriak@GitHub
• A version of calc.py exercises usage of text.scoping.tools.textx_isinstance() to inspect model objects types during traversal. (#136, #123). Thanks dkrikun@GitHub
• A version of calc.py in expression example that exercises dynamically adding properties to object classes (#126). Thanks dkrikun@GitHub
• python like imports (named import rules, scope providers affected) (#114)
• Added STRICTFLOAT as buildin type to allow to distinguish ints from floats in NUMBER. Fixed docu link (#98). Possible (BIC)
• Added flake8 and coverage checking (#92)

Changed

• All textX commands implemented using textX CLI extensibility. check command reworked to support the new registration feature (#187) (BIC)
• (Meta-)model visualization reworked as a set of textX generators (#187). (BIC)
• Made scope provider implementation of RelativeName and ExtRelativeName more readable (#186). Minor functional changes, not very probable to have any impact (only affects model-paths containing a list not at the end of the path; see #186). Possible (BIC).
• Improved handling of abstract rules references. Improved the definition of rules for various cases. Docs + tests (#185, #166) (BIC)
• Changed the time of call of match rule object processors to be during the model construction. This enable proper override of base types processors and calls on nested match rules (#183, #182, #96). Possible (BIC)
• CLI migrated to the click library (#162)
• Docs improvements (#146, #153, #151). Thanks simkimsia@GitHub.
• FQN constuctor param follow_loaded_models removed and introduced callback scope_rediction_logic to implement arbitrary logic on FQN resolution (#133, #114, #103) (BIC)
• Changed attribute name for the meta-model object (from metamodel._parser to metamodel._parser_blueprint). (#93) (BIC)
• Started using Keep a Changelog (#174)
• Started using Semantic Versioning (#174)
• Dropped support for Python 3.3

Fixed

• White-spaces in string matches were erroneously stripped (#188)
• Calling of match rule object processors (#183, #182, #96)
• Circular rule references in grammars (#173, #159, #155)
• Assertion error while calling object processors with multi meta models (extended grammars) and custom types (#141, #140)
• Using a sequence of matches and rule reference in an abstract rule choice alternative, and explicitly disallowing referencing more than one rule in an abstract rule choice alternative (#134)
• Unicode requirement for (meta)-model strings API parameters made strict. This should prevent common errors with Python 2.x. (#120) (related: #99, #105, #117). Possible (BIC)
• OS incompatibility fixes (path separator). (#114)
• Object processors called twice for imported models (#108, #118)
• Documentation and examples regarding NUMBER base type (#97, #100). Thanks approxit@GitHub

v1.8.0 (released: 2018-10-06)

• #71
  • Regular expression with group support
  • See the docs for usage.
• #69
  • Added search path feature (issue #66) - search path support for model files (importURI scope providers; see docs/scoping.md).
• #77
  • New multi meta model support for references-only for better meta model modularity (referencing models without having access to the grammar, see docs/multimetamodel.md).
• #79
  • Fixing obj_processors calling.
• #84
  • New contribution guide.
• #81
  • Bugfix: lost encoding for multi meta-model.
• #68
  • changed parser access in metamodel (private attribute "_parser")
• mkdocs documentation now uses mike for multiversion support.

v1.7.1 (released: 2018-07-02)

• Fixed bug with obj. processor call. #72
• Fixed bug with pos_to_linecol in the context of multi-file models. #67
• More tests.
• Thanks goto40@GitHub for the above fixes/tests.

v1.7.0 (released: 2018-05-31)

• A major feature of this release is multi-(meta-)model support with configurable resolving techniques. Thanks Pierre Bayerl (goto40@GitHub)!

  The docs sections are here and here.

  Details follow:

  • added new function textx.get_model.children to search arbritrary children using a lambda predicate.

  • remapped textx.model.get_children_of_type to the new children function (changed the logic, such that the root node is also checked to be model object).

  • added new metamodel function to register scope providers. Scope providers are callables, which return the referenced object.

  • added optional attribute "_tx_model_repository", see metamodel.py documentation

  • added attribute "scope_provider" like "obj_processors" to organize scope providers

  • added an optional argument to model_from_str and model_from_file: "pre_ref_resolution_callback": this is required internally to prepare the loading of other model files.

  • changed reference resolution in model.py

    • moved default resolution to textx.scoping.py
    • select the scope provider based on rule and rule-attribute (see scoping.py documentation)
    • added a Postponed type to postpone the resolution
    • introduced a multi-pass resolution (implemented at the end of parse_tree_to_objgraph; introduced new helper argument, e.g., a new optional argument "is_this_the_main_model" and "pre_ref_resolution_callback" (see metamodel.py above) to support reference resolution in the presence of different model files.

  • added a new module textx.scoping, to provide some scope providers (e.g. a fully qualified name provider) - see scoping.py:

    • full qualified names for reference names (e.g. package.package.class)
    • global scope (model distributed over different files - loaded globally)
    • import scope (model distributed over different files - loaded when imported)
    • relative scopes (e.g. instance.method where method is defined for the class of the instance in a model of classes, methods and instances)
    • selecting the metamodel based on a file pattern when loading models

  • added tests (mostly scope related - some of them also test other stuff, like buildins)

  • exceptions where adapted to include a file name (makes errors more visible)

  • The metamodel now allows to specify a global model repository. With this you can now share models across metamodels (before you could only do this within one metamodel or language).

  • The metamodel clones the parser when parsing a model file. The meta model holds one parser, which is clone for every model to be parsed.

    Backward incompatible change: The metamodel.parser is only a blueprint and cannot be used to, e.g., determine model element positions in the file. Use your_model._tx_parser instead, e.g., textx.get_model(obj)._tx_parser).

  • TextXModelParser now has a clone method. (TBC: is the clone ok: see responsibility of the method)

  • model.py: the resolution loop logic now mostly moved to a separate object ReferenceResolver, which holds the parser.

    • The reference resolver are build from all model files affected (loaded). This may involve multiple meta models.
    • Then all references are resolved in one loop.
    • Finally the helper objects (ReferenceResolver) are purged.

  • The MetaModelProvider has a clear-method now (useful for tests).

  • Added tests.

• Backward incompatible change: match filters and object processors unified. Now there are only object processors but they are called for any type of rule and the returned value if exists is used instead of the original object. See the docs. In order to migrate your match filters just register them as object processors.

• Backward incompatible change: all methods of textx.model module get a get_ prefix. See this commit and the docs

• Fixing FLOAT regex. Thanks Boris Marin (borismarin@GitHub)!

• Fixing position info on obj cross ref. Thanks Daniel Elero (danixeee@GitHub)!

v1.6.1 (released: 2017-11-22)

• Fixing build for PyPI.

v1.6 (released: 2017-11-18)

• Fix class masking for split grammar and rule overriden. Thanks aranega@GitHub!
• Simplifying metaclass handling. Using Python type system.
• Fixing base namespace construction.
• Rework of textX metaclass. Added a proper repr for both instances and classes.
• Introduced _tx_fqn class attribute.
• Introduced six library to handle Py2/Py3 metaclass usage.
• Simplification. Changed recursive instance resolving for an iterative. Thanks aranega@GitHub.
• Cleanup. Refactorings and code reorganization.
• textX api functions and classes are now available directly from textx module.

v1.5.2 (released:2017-11-17)

• Fixing bug with assignments in repetition. Commit e918868.
• Fix in resolving of attributes. Commit 8d18073.
• More robust obj. processor call.

v1.5.1 (released: 2017-06-14)

• Fixed issue #34 triggered with multiple assignments with the different types leading to errors in reference resolving.
• Fixed inheritance of base types.
• Some fixes in the docs (thanks borismarin@github).

v1.5 (released: 2017-05-15)

• For a more detailed feature highlights overview for this release please see what's new section in the docs.
• fixed issue #27 Operator precedence. Note: backward incompatibility! Be sure to read what's new section in the docs before upgrade.
• Issue #30 - match filters. See the docs.
• Added support for unordered groups. See the docs.
• Added _tx_metamodel attribute on the model object for easy access to the language metamodel. See the docs.
• Fixed issue #33. Much more robust support for multiple assignments to the same attribute.
• Fixes in namespace support. Import now works even for grammar defined as Python strings.
• Added special attribute _tx_position_end on model objects with the information where the object ends in the input stream.
• Added various handy functions for querying a model. See the docs on model API.
• New examples and additions to the docs. Added comparison to Xtext for users with previous Xtext exposure. Added references to ppci tutorial by Windel Bouwman.
• textX now have an Emacs mode!

v1.4 (released: 2016-05-31)

• Significant performance improvements done in Arpeggio (see issue #22).
• Fixed wrong rule type detection and infinite recursion in dot export for recursive match rules (issues #23 and #25).
• Added new performance tests to keep track of both speed and memory consumption.
• Memoization is disabled by default. Added memoization parameter to meta-model instantiation.
• New example - IBM Rational Rhapsody format.

v1.3.1 (released: 2016-05-17)

• Fixing several bugs in special cases regarding abstract rule with a single rule reference and match suppression.
• Improvements in docs, tutorials and examples.

v1.3 (released: 2016-05-05)

• Added support for syntactic predicates (! and &). A.K.A. negative and positive lookahead. Tests. Docs.
• Added support for match suppression (operator -). Tests. Docs.
• Added Entity tutorial + video.
• Added StateMachine example from Martin Fowler's DSL book.
• Various improvements in documentation and examples.

v1.2 (released: 2016-04-28)

• Reworked rule types detection and handling. Simplified textX grammar.
• Performance improvements.
• Updated docs to reflect changes.
• Additional examples. Fixing and tidying up.
• Various improvements in the dot rendering. Added * prefix for abstract classes.
• Extended tests. Added test for all examples.
• Improvements in textx command.

v1.1.1 (released: 2016-04-01)

• Added missing commands package.

v1.1 (released: 2016-04-01)

• Metamodel API tiding. Some undocumented internally used attrs/methods made private.
• textx command is console entry point now.
• Some fixes and improvements in the docs and examples.

v1.0 (released: 2016-03-03)

• Migrated docs to MkDocs.
• Docs improvement.
• Fixed escaping of backslashes in the dot exporter.
• Added textx command/script for (meta)model checking/visualization.

v0.4.2 (released: 2015-08-27)

• Added debug parameter to model_from_str.
• All prints converted to DebugPrinter.
• Fixed issue when object processor is applied to the list of BASETYPE elements.
• Fix in model export to dot to render proper base python type values inside lists.

v0.4.1 (released: 2015-08-08)

• Fixed bug with multiple rule params.
• Fixed ws rule loaded from file.
• Fixed loading of README in setup.py to use utf-8 encoding always.
• Added more tests.

v0.4 (released: 2015-05-08)

• Reworked meta-class/objects initialization.
• Added _tx_ prefix to textX class and object special attributes to prevent name clashing with the user attributes.
• Entity example.
• More tests.
• Documentation finished (except the advanced tutorial).

v0.3.1 (released: 2015-04-04)

• Boolean type matching for BASETYPE defined attributes.
• More tests.

v0.3 (released: 2015-02-10)

• Refactoring textX grammar. Support for ordered choice of sequences. Tests.

• Introduced new parameters to meta-model construction:

  • ignore-case - for case sensitive/insensitive matching.
  • autokwd - for matching on word boundaries for keyword-like matches.
  • skipws - for enabling/disabling automatic whitespace skipping.
  • ws - for redefinition of white-spaces.
  • auto_init_attributes - for auto initializing of model attributes based on their types.

• Object processor callbacks.

• Support for user meta-classes.

• Rule modifiers (ws, skipws, noskipws)

• Grammar import.

• Meta-model redesigned.

• Support for unicode grammars.

• Python 2/3 compatibility.

• Documentation. Work in progress.

• Travis integration.

v0.2 (released: 2014-09-20)

• Tests.
• Started documentation writing.
• A keyword-like string matches will be matched with the respect of word boundaries.
• Grammar redesigned.
• Introduced repetition operator modifiers (eolterm, separator).
• Support for multiple attribute assignment in the same rule.
• Position is now stored on model objects.
• Support for model processor callbacks.
• Filename is now accessible on model if the model is created from file.
• Various bugfixes.

v0.1.2 (released: 2014-08-17)

• Bugfix in conversion of basetypes to python.

v0.1.1 (released: 2014-08-17)

• Using relative imports.

v0.1 (released: 2014-08-17)

• Initial release. Most planed stuff in place.
• XText-like language fully functional. Rules, match rules, abstract rule, list matches, string matches, regex matches...
• Metamodel and model construction.
• Export to dot.