Check that the minimum dependency versions follow xarray's policy rules.
Note
Be aware that at the moment there is no public python API, so there are no stability guarantees.
Before we can validate environments we need to create a policy file (named e.g. policy.yaml). This allows us to specify the conda channels, the platforms we want to check on, the actual support windows, any overrides, packages that are not supposed to be checked, and allowed violations.
These windows are checked according to xarray's policy, that is, a version can be required as soon as it is at least N months old. As opposed to a fixed support window this means we can never end up requiring a one-month old dependency (or being able to drop all existing versions) for packages with infrequent releases.
For example:
channels:
- conda-forge
platforms:
- noarch
- linux-64
policy:
# policy in months
# Example is xarray's values
packages:
python: 30
numpy: 18
default: 12
overrides:
# override the policy for specific packages
package3: 0.3.1
# these packages are completely ignored
exclude:
- package1
- package2
- ...
# these packages don't fail the CI, but will be printed in the report as a warning
ignored_violations:
- package4If there are no packages with overrides, exclude, or ignored_violations,
you can set them to an empty mapping or sequence, respectively:
...
overrides: {}
exclude: []
ignored_violations: []The names of the conda channels to check. Usually, conda-forge.
The names of the platforms to check. Usually, noarch and linux-64 are sufficient.
The main policy definition. packages is a mapping of package names to the number of months, with default specifying the default for packages that are not in packages.
Any package listed in ignored_violations will show a warning if the policy is violated, but will not count as an error, and it is possible to force a specific version using overrides.
With the policy file, we can check environment files. There are currently two kinds supported: conda environment definitions as yaml files, and pixi environments.
Check also minimum-versions --help and minimum-versions validate --help.
To validate a conda environment file, run:
minimum-versions validate --policy ./policy.yaml path/to/env1.yamlor with an explicit prefix
minimum-versions validate --policy ./policy.yaml conda:path/to/env1.yamlWe can also validate multiple files at the same time:
minimum-versions validate --policy .../policy.yaml .../env1.yaml .../env2.yamlTo validate pixi environments, we need to specify the manifest-path (usually either pyproject.toml or pixi.toml):
minimum-versions validate --policy ./policy.yaml --manifest-path pixi.toml pixi:test-envwhere name in pixi:<name> is the name of the environment.
Again, we can validate multiple environments at once:
minimum-versions validate --policy ./policy.yaml --manifest-path pixi.toml pixi:test-env1 pixi:test-env2To check how validation would look at a certain point in time, use the --today option:
minimum-versions validate --policy ./policy.yaml ./env1.yaml --today 2025-10-01