citesync

Software citation metadata.

On this page

0002. Field locations come from a vendored mapping, in a separate repo

  • Status: accepted
  • Date: 2026-10-28

Context

Each file format keeps a field in a different place (CFF title is CodeMeta name). CodeMeta publishes crosswalks for this. Reading them live, or copying them by hand, would drift from the spec.

Decision

codemeta-drift is a separate repo. It owns the crosswalk data and exports a mapping file with a JSON Schema. citesync vendors a copy in src/citesync/, pinned to a codemeta-drift release in MAPPING_SOURCE with the sha256 of each file. A test checks the hashes.

The mapping says where a field sits. How a value is written (SPDX id or URL, ORCID URL or bare iD, Family, Given) stays in citesync as one codec per format and field, each with an inverse for writing.

An unknown major schema_version, or no row for a compared field in a format citesync reads, is exit 2. (Since 0.5.0 that is exit 3: 2 means only a usage error. See check.)

Two repos, not one, so the hub can move to the CodeMeta organisation whole if they ask for it.

Consequences

  • Adopting a new mapping needs a codemeta-drift release and a reviewed repin.
  • A new or changed mapping row can make existing CI fail, so it is a minor release with a changelog note.
  • The mapping holds paths only. Shapes a path cannot express (Cargo .workspace = true, npm author as string or object) stay in the format adapters.