Carry contribution records through citation files
A CodeMeta document can hold people and contribution roles. Citesync compares the people across files and preserves the roles in CodeMeta. It does not translate a CRediT role into a Zenodo contributor type.
The record each tool supplies
| Tool or format | Supplies or preserves | Limit |
|---|---|---|
| CRediT Matrix (main) | Linked Person and Role records for authors and non-author contributors |
People without ORCIDs use document-local identifiers; keep native Matrix JSON for contribution scores |
| CodeMeta | Separate author and contributor people, plus roles | Each role needs its linked Person; keep the original CRediT Matrix project file for scores |
| citesync | Reads people using the pinned codemeta-drift mapping; preserves existing role objects and extra person metadata | It cannot create missing identity information or choose a Zenodo contributor type |
| CITATION.cff | Authors, name parts and available ORCIDs | No general contributor list or CRediT role slot |
| .zenodo.json | Creators, typed contributors and existing affiliations | Its contributor types are a different vocabulary from CRediT |
CodeMeta 3 defines type and id as aliases of @type and @id.
Citesync reads either spelling for people. When correcting an existing person's
ORCID, it keeps the existing key spelling. Other person identifiers, such as
profile URLs, fragments and blank-node identifiers, are not ORCIDs. Writes keep
these identifiers and refuse to remove a person if that would discard them.
It preserves unknown role URLs too.
The merged Matrix exporter returns an attribution fragment with linked people and roles. Merge that fragment into a software CodeMeta document. Local identifiers link roles within that document; they do not establish a persistent identity across documents. This producer change is on Matrix main; use a release that includes it before relying on published packages.
Matrix's integration test
runs its built CLI against citesync revision 2a95043, which includes the
codemeta-drift 0.4.0 mapping. Use citesync 0.3.0 or later for the required identity fixes.
The test verifies linked people, role preservation, author/contributor separation,
repeat-write idempotence and explicit gaps for missing Zenodo contributor types.
It does not validate a Zenodo deposit.
Run the checked example
From a development checkout, run:
uv run pytest tests/test_contribution_contract.py
The example files contain one author with an ORCID and a software role, one author without an ORCID, and a non-author contributor with an existing Zenodo type. Name parts include a particle, a hyphen and non-ASCII characters.
The tests launch the installed CLI against copies in temporary directories. They
compare the written people with hand-checked expected values and verify that a
second write changes nothing. They also check a CodeMeta target rewritten from
CFF, role preservation, correction of an aliased id, and a warning for a role
whose linked Person was removed.
A new Zenodo contributor without a chosen type remains a reported mismatch;
write leaves that contributor list intact. Set
zenodo-contributor-type to supply a default for new contributors.
An existing empty target list opts in; an absent list stays absent. A passing check
does not establish that every contribution has reached every format.
The hand-written fixture checks the citesync boundary without installing Matrix.
The existing playground CI job also runs the
Matrix integration tests using [email protected], pinned in package.json
and pnpm-lock.yaml. Run the same check locally:
pnpm install --frozen-lockfile
uv run --frozen pytest tests/test_matrix_consumer.py
These tests use Matrix's native roster and hand-checked attribution expectation, plus the existing target expectations. They check linked people and roles, local identities, idempotence and explicit Zenodo contributor types. The fixture notes record provenance. This checks citation files, not a live Zenodo deposit.