Build the `contributors` array for the register's `.zenodo.json` — build_zenodo

Every codechecker named in `codecheck.yml` across the register becomes one Zenodo contributor (register#58), crediting people whose work the register otherwise only lists on their own person page. Zenodo's contributor role vocabulary (<https://zenodo.org/api/vocabularies/contributorsroles>) has no "reviewer" or "checker" term, so every entry is typed `type` (default `"Other"`) - the record's own description is where that choice is explained to a reader.

Usage

build_zenodo_contributors(
  register,
  exclude = ZENODO_NON_PERSON_CODECHECKERS,
  type = "Other"
)

Arguments

register: The register data frame (as read from `register.csv`), with `Repository` and `Certificate` columns.
exclude: Character vector of codechecker names to omit as not naming an individual. Defaults to [ZENODO_NON_PERSON_CODECHECKERS].
type: The Zenodo contributor role to assign every entry.

Value

A list of `list(name=, orcid=, type=)` records (`orcid` omitted when not on record), sorted by name for a stable, diffable order.

Details

Reads `codecheck.yml` directly (via [get_codecheck_yml_or_null()], cached) rather than through [add_codechecker()]'s per-row identifiers, because that column collapses every codechecker with neither an ORCID nor a resolvable GitHub handle to the shared identifier `"NA"` - fine for counting checks, but it would silently merge distinct people here. Working from the raw name keeps them distinct; only [ZENODO_NON_PERSON_CODECHECKERS] is treated as not-a-person.

A codechecker is deduplicated by normalized ORCID when they have one, otherwise by their exact recorded name - so the same person named identically across certificates contributes one entry, keeping the name exactly as recorded in `codecheck.yml` (no "Family, Given" reformatting: a heuristic split gets compound and multi-word family names wrong, and the recorded spelling is the person's own).