Skip to content

Consider a reusable data-dict validate-spec workflow, once more than one repo needs it #772

Description

@d-morrison

Raised as a non-blocking review observation on ucdavis/bcs#838.

bcs is piloting a data-dict.yaml data dictionary, with a check-data-dict workflow that runs data-dict validate-spec in CI. That workflow is currently hand-rolled in the consumer repo, unlike every comparable lightweight check there (check-non-standard-chars, check-typos, lint-workflows, news), which call Morrison-Lab/gha/.github/workflows/*.yml@v2.

That is deliberate for a pilot -- one consumer, a pre-1.0 tool (spec 0.1.0, CLI 0.0.3, breaking changes expected), and no reuse to justify the indirection yet. This issue tracks the graduation: if a second repo adopts a data dictionary, or the bcs pilot is judged a keeper, lift the job into gha as a reusable workflow.

The steps to lift, as they stand today in bcs:

  • actions/setup-python (digest-pinned), then pip install --user data-dict-yaml==<pinned>.
  • chmod +x "$(python3 -m site --user-base)/bin/data-dict" -- pip's install step drops the binary's execute bit (the wheel itself records 0755), so it fails with Permission denied otherwise.
  • A --version probe to fail fast, then data-dict validate-spec <path>.

Inputs a reusable version would want: the CLI version to pin, and the dictionary path (defaulting to data-dict.yaml). Only the spec level belongs in CI -- validate-meta/validate-data read the data.

Not urgent, and premature while there is one consumer.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions