v0.6.0

Upgrade step: the hub commands of 0.5.0 and older refuse an export with the new formats, so upgrade
codemeta-drift with the file. A consumer that vendors the mapping schema vendors the 1.6 schema and
mapping-vectors.json with it, and runs the vectors through its resolver: a 1.6 export fails the 1.4 schema that 0.5.0 shipped.

Added

• The mapping export covers 22 more formats: DataCite, setup.py, Ruby gemspecs, Maven's pom.xml, CPAN's
  META.json, NuGet's .nuspec, Dart's pubspec.yaml, Cabal, typst.toml, Quarto's _extension.yml, the GitHub
  API, Figshare, bio.tools, Leiningen's project.clj, Octave's DESCRIPTION, publiccode.yml 0.8, swMATH, SPDX 2.3,
  CodeMeta 2.0, DCAT 3, DOAP and Wikidata, and the coverage page and SSSOM spokes cover them too. Each row was
  checked against the format's own tools and spec. See the mapping export.
• Ship mapping-vectors.json beside the mapping schema: test cases of path, document and the values the path
  resolves to, for a consumer to run through its own resolver. See the mapping export.

Changed

• The BibTeX codeRepository row reads repository | url, since plain BibTeX styles print only url.
  See the coverage page.
• Schema 1.6 extends the path grammar: a + b reads both parts, [k=v] also matches a list that holds v, and
  a key may carry a prefix. The PKG-INFO author rows use +, the Project-URL rows end in .url, and the DCAT 3
  and DOAP rows use prefixed names such as dcterms:title. See the
  path grammar.

Fixed

• A person sub-term such as author.@id is read from the alternative its term resolved through, not by first
  match, so one person's ORCID never lands on another. See the mapping export.
• The schema watcher now also opens a citesync issue when a hash in MAPPING_SOURCE differs from the file published at the pinned tag, and its run prints a notice when it skips publishing for lack of a token. The docs name the token permissions it needs.

v0.5.0

Breaking: SSSOM term:, field: and format: identifiers and the mapping_set_id have new URIs.
Regenerate stored SSSOM files with codemeta-drift hub sssom rather than joining them with 0.4.0 output.

Upgrade step: a consumer that validates the mapping export against a vendored schema vendors the 1.4 schema
with the new export, since R paths contain @. Readers that skip formats they don't use read it unchanged.
The hub commands of 0.4.0 and older refuse a 1.4 export, so upgrade codemeta-drift with the file.

Added

• See every crosswalk cell for nine citation and package formats, and why a reviewed mapping changed it, in the
  coverage matrix and on the coverage page.
  codemeta-drift hub coverage writes the matrix as Markdown or JSON. It also lists the CodeMeta crosswalks
  nobody has reviewed, each cell marked as a path, prose, empty or unmatched; --codemeta <checkout> adds them.
• The coverage page checks CodeMeta's main branch every Monday and lists the cells that changed since the pinned tag.
• The mapping export covers BibTeX, Python PKG-INFO, R DESCRIPTION and Julia Project.toml, each cell reviewed
  against CodeMeta's crosswalk. Its schema_version is 1.4, and a path key may now contain @, as in R's
  Authors@R. See the mapping export.

Changed

• SSSOM exports carry mapping-set metadata (title, description, version, curator ORCID and provider), and their
  term, field and format identifiers resolve on the coverage page. See the crosswalk hub.

v0.4.0

Breaking: Schema directories passed to export --schemas now also need zenodo.json.
Copy the verified structural subset into your schema directory.
The new watcher uses the same five-file layout; see Zenodo coverage and limits.

Added

• Run an offline schema recovery example with an explicit synthetic approval and reproducible mapping hashes.
• Check Zenodo mapping paths against a verified structural subset; importer source changes produce review drafts. This does not validate complete deposits.
• Find corresponding fields in two formats and export seven SSSOM spokes with the
  crosswalk hub. The Go spoke relates a module path to the repository field.
• Preview broken schema paths, new fields and a behind mapping pin with the
  offline schema watcher. It prints deduplicated issue drafts.
• Mapping exports include each original crosswalk cell alongside the exported path, so readers can show corrections and empty cells.
  See the mapping export contract.

Changed

• The README now leads with checking and exporting. Finding codes, the dated findings report and contributor instructions have dedicated guides.

Fixed

• Missing Zenodo importer files produce review drafts instead of aborting unrelated schema and pin checks. See Zenodo coverage.
• The recovery example reports unreadable approvals and existing output directories as CLI errors without modifying existing files.
• Hub commands reject malformed original crosswalk cells, keeping mapping validation consistent with the export contract.
• SSSOM exports use distinct field and format URI prefixes so reference readers can load them.
  See the validation and compatibility guide.
• Schema conformance checks recognize array references and tuple item schemas.
• A CELL_NO_AUTHOR finding now says the author or contributor has no path, not no row, which was
  wrong when its curated row has a null path, as CFF's contributor row does.
• export refuses a crosswalk cell that already reads as a list, such as authors[], with a CELL_PATH finding, instead of
  writing authors[][] and its sub-terms into the export.
• A check of the working tree skips a directory whose name matches an input file, such as examples/x.json, as
  --ref already did, instead of exiting 2.

v0.3.0

Added

• The export maps swhid, the Software Heritage identifier: CFF identifiers of type swh,
  Zenodo related_identifiers with scheme swh (or no scheme and a swh:1: value), and
  codemeta.json identifier, which holds the DOI too. See the term list.
• The CFF author.familyName row's comment says to join name-particle, family-names and
  name-suffix into the family name, so a consumer keeps "van" in Antonie van Leeuwenhoek.

Changed

• The export's schema_version is 1.2, since it adds a term. A reader of 1.0 or 1.1 reads it unchanged.

v0.2.0

Added

• The export maps datePublished and contributor, with the same person sub-terms as
  author. CFF 1.2.0, Cargo.toml and pyproject.toml have no contributor list, so their rows are
  null; Zenodo's need a curated row, because its crosswalk cells are empty.

Changed

• The export's schema_version is 1.1, since it adds terms. A reader of 1.0 reads it unchanged.
• A consumer now needs rows only for the terms it uses, not for every term in the export.

v0.1.0

Added

• codemeta-drift <checkout> [--ref REF] [--json] lists where CodeMeta's context,
  definitions, property table, examples, crosswalks and crosswalk.csv disagree. Each check
  has a stable code. A malformed input file exits 2, so CI doesn't mistake it for drift.
• --json prints one versioned document holding the findings, and --version prints the installed version.
• --ref reads any tag, branch or commit through one git cat-file call, about as fast as
  reading the working tree.
• The mapping export format: a JSON Schema for mapping-<codemeta-tag>.json, which says where
  each CodeMeta term sits in six metadata formats, and the export for CodeMeta 3.1.
• codemeta-drift export [checkout] --tag TAG [--out FILE] writes the export from the tag's
  commit. Rows read the crosswalk cell when it is a path, and come from a curated file with a
  reason when it is prose, empty or wrong. A cell with neither is listed and the command exits 1,
  as is a curated row whose cell has changed, and two terms whose paths overlap.
• export --schemas DIR checks every path against the format's own JSON Schema (CFF 1.2.0,
  pyproject.toml, package.json, Cargo.toml) and reports PATH_NOT_IN_SCHEMA.
• Test fixtures: the files the checks read from CodeMeta 3.1 and 4.0, and the output for each,
  and the JSON Schemas of four formats. tests/fixtures/3.1.sha256 ties the 3.1 files to the
  tag's commit.