Set up reproducible Python environments with conda-lock #2968

* Add a conda-lock setup for discussion. * Move python-snappy into project.dependencies in pyproject.toml * Remove sphinx-autoapi from pypi deps, and no longer required snappy-python * Switch to using conda-forge version of recordlinkage v0.16 * Update conda-lock.yml now that all dependencies are available on conda-forge * Consolidate conda env files under environments/ dir * Add a GitHub action to relock dependencies * Quote the pip install command * Remove pip install of pudl from environment.yml * Rename workflow * Only build lockfile from pyproject.toml, don't install extras. * Just install conda-lock, not pudl, before running conda-lock. * install conda-lock with pip * Move all remaining dev-environment.yml deps to pyproject.toml * Add other platforms; make draft PR against dev. * Comment out dev base branch for now. * Remove pandas extras and recordlinkage deps from pyproject.toml * Use conda-lock --micromamba rather than --mamba * Don't specify grpcio, or specific recordlinkage version * Render platform-specific environment files in github action * Fix paths relative to environments directory * Add some comment notes to workflow * Render environment for Read The Docs. * Use environment not explicit rendered lockfile * Add readthedocs specific sphinx extension * Don't render explicit conda env for RTD since it can't read it. * Build linux-aarch64 lockfile. Use conda-lock.yml in workflows. * Comment out non-working linux-aarch64 platform for now. * Switch to using rendered lockfiles. * Remove deprecated environment files * Switch to using a micromamba docker image * Install git into the docker image. * Use micromamba and unrendered multi-platform lockfile. * Add main category to micromamba environment creation. * Use conda-lock not base as env name * Add a conda-lock setup for discussion. * Move python-snappy into project.dependencies in pyproject.toml * Remove sphinx-autoapi from pypi deps, and no longer required snappy-python * Add linux-aarch64 platform back into conda-lock settings.

- Remove deprecated API_KEY_EIA envvar - Add .gitignore in new environments directory - Create Makefile and migrat tox.ini logic into it - Replace spaces in Makefile with tabs - Install pudl via pip install --no-deps to avoid contaminating the locked environment - Move pip install and coverage XML logic from GHA workflow into the Makefile - Increase the minimum versions of some dependencies. - Move update-lockfile GHA workflow logic into Makefile - Attempt to run slow tests in parallel using "wait" to prevent premature exit of the shell

* Use micromamba not conda in Dockerfile CMD, also use pip install --no-deps * Use micromamba not conda in command passed to build container * Use default mambauser rather than catalyst in docker container * Remove --no-capture-output which isn't supported by micromamba. Is this a problem? * Remove uninterpolated vars in .env and more --no-capture-output * Separate ETL and pytest commands. * Stop trying to run tests in parallel. Sigh. * Add google cloud sdk to conda environment. * Install Google Cloud SDK from conda-forge. * Add back in the making of required directories. Oops. * Attempt to have micromamba run pass through output * Use prettier to standardize formatting of lockfiles. * Add dagster (server startup) target to Makefile * Update conda lockfile and rerender environment files * Attempt to trigger update-lockfile when pyproject.toml is changed. * Remove non-required upper bounds on dependency versions. * Use correct branch name for update-lockfile trigger. * Fix incorrect nbformat minimum version * Update Makefile in response to PR feedback

* Remove dagster-postgres version to avoid PyPI conda-forge conflict. * Break up and distribute the nuke target. * Resolve issues with pandera dependency. Reorganize makefile. * Add triggers and commit/PR for workflow_dispatch, pull_request, schedule and set appropriate GITHUB_REF values for each case. * Use push instead of pull_request to trigger on path. This avoids re-locking the dependencies every single time you push to a PR that had a change to pyproject.toml *somewhere* in it. * Also trigger based on path if .github/workflows/update-lockfile.yml changes. * Update conda-lock.yml and rendered conda environment files.

* Move previous dev, test, and datasette optional dependencies into the required dependencies to simplify application installation. * Test make nuke; parallelize --live-dbs tests * Move prettier into conda-only dependencies * Update conda-lock.yml and rendered conda environment files. * Remove action test file hashlog * Remove merge markers. * Remove transitive astroid dependency that's now correctly included in solve. * Use the real immature library version of dagster-postgres (0.21.6) rather than the accidentally packaged 1.5.6 version found in conda. We'll need to keep an eye out for when dagster-postgres graduates to the stable versioning and update it. This is a bit of a mess because of some broken automation in the conda packaging for dagster which has now been fixed. * Update "make pudl" to remove the old PUDL DB and reinitialize with alembic, rather than writing to the DB that already exists. * Fixed some groupby.agg() deprecation warnings. * Fix dagster-postgres version (again). * Update username in path to settings file * Avoid bugs in ferc_to_sqlite --clobber; don't use cache_dir for pip install. * Make FERC extraction output removal more specific. * Bump numpy and numba minimum versions. * Bump black version in pre-commit * Bump ruff pre-commit hook version

* Rename tox-pytest and update-lockfile workflows. * Make scheduled PRs against dev rather than conda-lockfile branch. * Update to pyarrow 14 and grpcio 1.59 b/c security * Update release notes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Set up reproducible Python environments with conda-lock #2968

Set up reproducible Python environments with conda-lock #2968

Commits on Nov 14, 2023