verify_paper_claims: read the v0.2.6 paper draft (v2#61)

#114
by ppuzio - opened
Fabryka AI org

Tooling for v2#61: lets src/verify_paper_claims.py read the v0.2.6 paper draft (v2 PR 93). No data changes.

  • PAPER_NAMES: the 15 display names v0.2.6 adds.
  • tab:additions: a minus typeset as $-$ parses as a negative cell; a Net change row is the totals row of a release that also removes sources (Total added still works).
  • norm_license: a LicenseRef compares by its name, so the paper's "EU reuse policy" matches LicenseRef-EU-Reuse-Policy (the eurlex FAIL).
  • released_at: a source stops counting from its removed release. wikisource, wikiquote and european_hplt_v3_pl get removed: "0.2.6". AGENTS.md ("Cut the release", step 0) and the source-rights-review withdrawal pattern say to set it.

Left for the release commit: v0.2.6 in src/plot_version_growth.py, release: "0.2.6" on the 15 new sources (after NERGAL, v2#60), the CHANGELOG section.

Checks

  • pytest src/test_verify_paper_claims.py: 51 passed (44 before). The new removal and LicenseRef tests fail on the old script.
  • Full pytest src (LFS not smudged): the same 119 failures/errors as at ec01c58, all tests that read parquet data.
  • Against the paper draft (v2 PR 93, head 829f00b), --ref worktree: 23 FAILs, all from the two release-commit edits above. With those two simulated in memory: 0 FAILs, 22 WARNs (15 datasheets without a **License:** line, 7 expected dual-licence notes).

🤖 Generated with Claude Code

ppuzio changed pull request status to open
ppuzio changed pull request status to merged
ppuzio deleted the refs/pr/114 ref

Sign up or log in to comment