CHANGELOG.md

Changelog

The CHANGELOG format is based on Keep a Changelog

NOTES:

• Version releases in the 0.x.y range may introduce breaking changes.
• See the auto-generated release notes for more details.

Note: version releases in the 0.x.y range may introduce breaking changes.

Unreleased

Added

Changed

Deprecated

Removed

Fixed

• crm convert now negates detect_binary_files_with_prefix when converting it to the V1 prefix_detection/ignore_binary_files field. The two flags are logical inverses, so previously carrying the value over unchanged produced recipes with inverted binary prefix-replacement behavior.

Security

[0.10.5]

Fixed

• The boolean flag for detect_binary_files_with_prefix not being inverted.
• Multiline string regression issue, #541.

0.10.4

Fixed

• Bad cast in the ArtifactFetcher recovery logic.
• Recipe/CBC file parsing issue tied to list indentation.

0.10.3

Changed

• crm convert command now defaults to use the ALLOW_DUPLICATE_KEYS flag in the parser. Our integration tests have shown that this improves crm convert compatibility more than it risks breakages in the conversion process.

Fixed

• Documentation typos.
• Minor infrastructure improvements.
• VariantsManager now more closely follows conda-build's behavior. Variants are produced per Python version supplied by the CBC file, unless a noarch recipe is used.

0.10.2

Added

• Finalizer to fetch test cases so that clean up of the temp directory doesn't cause issues.
• Conversion for skip_compile_pyc in v0 to v1.

Changed

• stack_path_to_str to always make a copy of the data being manipulated.
• pre-commit to simplify isort, black, and pylint versioning consistency.
• CbcParser to be split into CbcParser and CbcReader to be consistent with the Recipe* classes.

Removed

• Redundant README line.
• dangerous-default-value pylint ignores.

Fixed

• @psteyer fixed the parser to handle nested list yaml syntax in CBC files.
• Multi-list rendering bug to correctly render lines that are blank but are the beginning of a list.

0.10.1

Added

• ALLOW_DUPLICATE_KEYS flag to ignore duplicate key exception on parsing of recipe files with duplicate keys.
• crm convert support for python_version_independent field.
• crm convert flag for managing latest Python version.

Changed

• FLOATS_AS_STRINGS to a default flag for the VersionBumper.
• crm convert to improve pip_check conversion.

0.10.0

Added

• Automated CI testing for Python 3.13. Python 3.14 does not have all CRM dependencies available yet.

Changed

• click dependency has been pinned to 8.2.1, which is available on defaults and conda-forge channels for Python 3.11-3.14.
• Similarly,
  • pylint
  • pyfakefs
  • pytest-socket
  • sphinx dependencies have also been upgraded to use the latest versions available on conda-forge and defaults that support Python 3.11-3.14.
• Some dependencies do not yet exist for Python 3.14 on the defaults channel, so we have capped CRM support to 3.13 for now.

Deprecated

• crm graph and the associated libraries have been removed from the project. CRM v0.9.3 will be the last version containing the tool. This was done to remove graphical dependencies which has caused issues running CRM in a headless environment. crm graph has not used enough (and was never fully completed) to justify continuing to support it.

Fixed

• Fixes #476. CRM now quotes version values that could be interpreted as a float.
• Fixes Issue #220 (and others). CRM now supports all YAML multiline string formats. This should significantly improve parsing compatibility. In our testing, we believe this should allow CRM to be able to parse recipe output from conda-build artifacts.
• crm convert now deprecates the license_url field in compliance with recent changes to CEP-0014.
• crm convert now ensures run_exports contains a list if a single value is provided.

0.9.3

Changed

• @leahsch Made improvements to the README file and fixed some typos.

Fixed

• @mgorny Fixed a few edge V1 upgrade issues when a V0 recipe file contains python version information.

0.9.2

Fixed

• RecipeVariant ignores pure comment lines when evaluating selectors.
• RecipeVariant only attempts to render JINJA for node values containing JINJA expressions.

0.9.1

Added

• Canonical recipe section sorting at the root and outputs level after each patch add operation in RecipeParser.

Fixed

• RecipeParser.update_skip_statement_python() works in the event that a build section is entirely absent.

0.9.0

Added

• Recipe build variants generation.

Removed

• SelectorQuery class, replaced with BuildContext.

0.8.1

Changed

• RecipeReader.get_value() now uses render_to_object(), simplifying the logic and preserving retrieved value types better.
• RecipeReader.render_to_object() was also reworked for increased accuracy, and to better support get_value().

0.8.0

Added

• The conda_recipe_manager.ops.VersionBumper class. This takes the core logic from crm bump-recipe and exposes it as a library that can be easily leveraged by other tooling.
• Two new context-manageable functions (from_recipe_fetch() and from_recipe_fetch_corrected()) to the artifact_fetcher module. These attempt to provide an easier way of acquiring remote source resources referenced in Conda recipe files.
• HttpArtifactFetcher::get_path_to_archive() which provides access to the source archive file acquired when fetching a remote resource. This is exposed to provide compatibility with existing Conda tooling.
• A static utility function in GitArtifactFetcher that provides some limited ability to match git tags to semantic version strings. This functionality was originally created by @msentissi.

Changed

• artifact_fetcher::from_recipe() is now context-manageable.
• The crm bump-recipe command now utilizes the new VersionBumper class in a backwards-compatible way.
• The artifact fetching classes derived from BaseArtifactFetcher can now be used as context-managed instances. This provides callers with finer control over how long the temporary disk space is available.

Fixed

• A test that was accidentally initializing 2 instances of the fake filesystem.

0.7.1

Added

• --fail-on-unsupported-jinja flag to crm convert at the request of the conda-forge community. When unsupported JINJA statements are seen in a V0 recipe attempting to be upgraded to V1, this flag kills the convert process with a unique exit code.

Changed

• make test now shows verbose pytest logs.

0.7.0

Added

• Support for multi-line JINJA set statements.
• Support for JINJA list variables.
• ParsingJinjaException, which is thrown at construction if unsupported JINJA is encountered.
• force_remove_jinja flag to the parser constructors to silently remove unsupported JINJA and attempt parsing.

Changed

• JINJA expression evaluation now uses jinja2 directly, instead of custom logic.
• crm convert, if it fails to parse a recipe because of unsupported JINJA, will output a warning, remove the unsupported statements, and re-attempt V1 conversion.
• Book theme for sphinx is used for the project documentation.

0.6.4

Changed

• RecipeReaderDeps.get_all_dependencies() now has a flag to include test dependencies.

Fixed

• Various issues with crm convert around the new V1 /build/script section.

Security

0.6.3

Added

• The CRM software version to the API docs.
• Support for duplicate JINJA variable names for V0 recipes. There are some missing features around this change, but at least duplicate JINJA variables are no longer dropped. This is commonly seen when a string variable needs a conditional concatenation.
• RecipeParser::update_skip_statement_python() which provides the ability to upgrade the commonly used [py>=3*] version selector to a new python version for V0 recipe files.

Fixed

• Some formatting issues in the API docs.
• Issue #423 by making improvements to the SpdxUtils class.
• An issue with handling empty dependency sections in RecipeParserDeps::add_dependency()

0.6.2

Added

• Logging support via the standard Python logging library. By default, all modules (except commands, which contains the CLI programs) default to use the NullHandler so that the client program can determine the best way to log for their needs.
• A series of new parsing-related exceptions. Most notably, when there is a parsing failure at parser construction, a ParsingException is raised from the original exception.

Changed

• De-emphasizes/redacts the MessageTable class in favor of the standard Python library logger. It is now only used as part of the recipe conversion process in crm convert

Fixed

• Improves the V0 format correction algorithm used to increase V0 recipe compatibility.

0.6.1

Added

• crm CLI commands now support -h in addition to --help.

Changed

• CHANGELOG.md now uses the Keep a Chagelog format (from this version on)
• The retry logic in crm bump-recipe now uses a much shorter default timeout.

Fixed

• The algorithm responsible for cleaning up V0 recipes prior to the parsing stage has been greatly improved to better handle inconsistent tabs.
• Wrong interpretation of the + in JINJA variables in some cases.

0.6.0

• Contains many significant parsing bug fixes. Most of these fixes relate to quoting strings and JINJA statements. Our integration tests indicate that this increases parsing compatibility across the board. We are seeing that ~97% of all recipe files in the conda-recipe-manager-test-data set can now be parsed without throwing an exception.
• Contains a few bug fixes around parsing and rendering comments in V0 recipe files.
• Fixes several edge cases for PyPi URL upgrades with crm bump-recipe
• Adds a pytest-socket smoke test
• Includes several new quality-of-life improvements for CRM developers
• Huge shout-out to @mrbean-bremen for their help on diagnosing some pyfakefs issues on the latest point releases of Python.

0.5.1

• Fixes a JINJA rendering bug for V0 recipes in RecipeReader::get_value(..., sub_vars=True) and related functions.

0.5.0

• BREAKING CHANGE: RecipeParser::search_and_patch() has been replaced by RecipeParser::search_and_patch_replace().
• Adds some support for interpreting split and join JINJA functions.
• Adds support for negative indexing in JINJA statements.
• crm bump-recipe now attempts to correct PyPi URLs when the predicted source artifact URL cannot be found. This is to counteract changes introduced in the last year where some PyPi tarballs may have switched from using _s to -s. This comes from a PEP standard change for source artifacts.
• crm bump-recipe now replaces all older PyPi URLs with pypi.org addresses.
• Minor fixes for the crm graph command.
• Fixes risky threading behavior in crm bump-recipe.
• Fixes for converting /build/force_use_keys, /build/ignore_run_exports_from, and /build/ignore_run_exports in crm convert.

0.4.2

• Various bug fixes and improvements related to crm bump-recipe
  • Introduces the new --override-build-num flag to allow bumped recipes to start counting from a non-zero value
  • Adds support for | replace() JINJA functions to be evaluated.

0.4.1

• crm convert now prints stacktraces when the --debug flag is used.
• crm bump-recipe now includes a --save-on-failure flag that can save the contents of a bumped recipe if the bump fails to fully complete.
• Significant recipe compatibility improvements, brought on by changes made in rattler-build @0.34.1.
• Several community-reported bug fixes.

0.4.0

• Introduces MVP for the bump-recipe command. This command should be able to update the version number, build number, and SHA-256 hash for most simple recipe files.
• Starts work for scanning pyproject.toml dependencies
• Minor bug fixes and infrastructure improvements.

0.3.4

• Makes DependencyVariable type hashable.

0.3.3

• Fixes a bug discovered by user testing relating to manipulating complex dependencies.
• Renames a few newer functions from *_string to *_str for project consistency.

0.3.2

• Refactors RecipeParserDeps into RecipeReaderDeps. Creates a new RecipeParserDeps that adds the high-level add_dependency() and remove_dependency functions.
• A few bug fixes and some unit testing improvements

0.3.1

Minor bug fixes. Addresses feedback from conda-forge users.

0.3.0

With this release, Conda Recipe Manager expands past it's original abilities to parse and upgrade Conda recipe files.

Some highlights:

• Introduces the scanner, fetcher, and grapher modules.
• Adds significant tooling around our ability to parse Conda recipe dependencies.
• Adds some initial V1 recipe file format support.
• Introduces many bug fixes, parser improvements, and quality of life changes.
• Adds pyfakefs to the unit testing suite.

Full changelog available at: https://github.com/conda/conda-recipe-manager/compare/v0.2.1...v0.3.0

0.2.1

Minor bug fixes and documentation improvements. Conversion compatibility with Bioconda recipe has improved significantly.

Includes many previously missing recipe transformations.

Pull Requests

• Fixes integration tests from rattler-build 0.18.0 update (#76)
• Adds common environment settings (#75)
• Adds demo day intro slidedeck to CRM (#73)
• Upgrades basic quoted multiline strings (#72)
• Adds missing transforms for "git source" fields (#71)
• pip check improvements, more missing transforms, and some spelling enhancements (#70)
• Multiline summary fix issue 44 (#68)
• Preprocessor: Replace dot with bar functions (#67)
• Corrects using hash_type as a JINJA variable for the sha256 key (#66)
• Adds script_env support (#65)
• Adds missing build transforms (#62)
• Some minor improvements (#61)

0.2.0

Major improvements from 0.1.0, mostly dealing with compatibility with rattler-build. This marks the first actual release of the project, but still consider this work to be experimental and continually changing.

0.1.0

Migrates parser from internal project. , ,