Changelog#

Unreleased#

  • Use Rich for human-readable output, following conda-workspaces’ blue, cyan, and yellow status colors. JSON output is unchanged. Provider text is rendered literally and terminal control characters are escaped.

  • Add reproducible VHS recordings of feature management and configuration, with isolated demo setup and matching conda-workspaces terminal styling.

0.2.0 - 2026-11-13#

  • Breaking: remove conda feature export, conda feature apply, the custom intent JSON format, and Feature.portable_settings/export_intent. There is no deprecation period or replacement manifest. Use conda configuration and explicit feature commands. Existing local deployment receipts, --json output, and the provider’s Feature.apply method are unchanged. Release 0.1.0 artifacts retain the old commands.

  • Add conda feature config FEATURE --set KEY VALUE, following conda config argument syntax. Repeat --set for multiple settings. The released configure FEATURE KEY=VALUE command remains supported.

  • Add feature-scoped --get, --show, --show-sources, --describe, --validate, and confirmed --remove-key. Queries expose only declared settings. Removing an override preserves inherited configuration and checks feature requirements.

  • Rewrite the README and documentation in plainer language while retaining setup and recovery guidance.

0.1.0 - 2026-11-13#

Initial public release of conda-features, available on PyPI and Anaconda.org.

Feature management#

  • Discover installed providers through the conda_features Pluggy hook.

  • List and explain saved choices, effective configuration, readiness, and lifecycle metadata without claiming to observe provider execution.

  • Enable, configure, and disable features with previews, explicit confirmation, and JSON output.

  • Select user, installation, environment, or standalone-file configuration destinations using conda’s typed settings and precedence.

  • Plan explicit prerequisite opt-ins and protect dependents in checked environments. Shared changes inspect the current user’s registry and explicit targets, not every environment on the machine.

Deployment and recovery#

  • Declare reusable package, scalar-setting, channel, and managed-file-block steps.

  • Execute approved package transactions while protecting the running conda and Python installation from replacement.

  • Record resource ownership in versioned receipts for retry and reverse-order cleanup. Preserve pre-existing resources and later user edits.

  • Diagnose local drift, missing providers, overrides, and incomplete checkpoints through feature diagnose and conda doctor features.

  • Repair saved setup through the existing preview and confirmation flow. Interrupted disable must be resumed with disable, not repair.

Lifecycle and portability#

  • Preview explicit migration between boolean-backed features using provider-declared setting mappings.

  • Distinguish optional graduation from unconditional functionality and clean up obsolete boolean opt-ins.

  • Export versioned feature intent with explicitly portable settings. Apply validates provider identity, configuration, compatibility, dependencies, and conflicts before mutation.

  • Reject recognizable credential-bearing URLs, absolute machine paths, and non-finite values in portable settings. Generic deployment migration and deployment-setting round-tripping remain limited.

Documentation and distribution#

  • Provide Sphinx documentation organized into tutorials, how-to guides, reference, and explanation, published through GitHub Pages.

  • Add native core CI for Linux, Windows, and ARM macOS, with separate real-provider integration tests.

  • Publish the core package under BSD-3-Clause. The repository-only conda-features-examples package supplies real PyPI-support and shell-completion deployments but is not published to either registry.

  • Attach Python build provenance, conda build provenance, and CEP 27 publication evidence to the GitHub release.

Publication history#

The Python artifacts were published from commit 3098f8ae2cbb0bc15bf28c1ec5ab2573d6c2e408. An Anaconda workflow-secret fix required moving the then-mutable 0.1.0 tag to d853260e26a9543aaee769f92731f0ac5c1d7fc2. The recovery reused the published Python files byte-for-byte and built the conda package from the same published sdist.

The GitHub release was subsequently recreated with release immutability enabled. Its five assets retained their SHA-256 digests, and the recovery tag was preserved. The release is now immutable. This change did not replace packages on PyPI or Anaconda.org. The documentation’s release-verification guide explains the distinct original Python and recovery conda source commits.

On this page
Show Source