Skip to content

#310 §6: farce-v1 exemplar manifest (Emu War, Football War → war; Dreyfus affair → justice), draft until the seeder takes events - #313

Draft
holden wants to merge 1 commit into
mainfrom
claude/310-farce-exemplars
Draft

holden wants to merge 1 commit into
mainfrom
claude/310-farce-exemplars

Conversation

@holden

@holden holden commented Nov 4, 2026

Copy link
Copy Markdown
Member

#310 §6: the farce-v1 exemplar manifest, and what blocks seeding it

Draft until the seeder can take event subjects (see below). On the owner's words "Yes, rows 1–3" (5 November 2026).

What is in the branch

priv/exemplars/farce-v1.json, three rows in first-v1's shape, checksum 3b8ff07e… stamped (verified against Examples.Manifest.checksum/1 on first-v1):

row word sense (WordNet) subject evidence
1 war "waging of armed conflict" Emu War, event, Q14665 The Argus 3 and 19 Nov 1932 (Trove ids as Wikipedia cites them); Wikipedia
2 war same Football War, event, Q223085 (Wikidata label "Hundred Hour War", enwiki "Football War") BBC News 27 June 2019; Wikipedia
3 justice "quality of being just or fair" Dreyfus affair, event, Q173457 Zola, J'accuse…!, L'Aurore 13 Jan 1898 (Wikisource); NYT 7 Mar 2013; Wikipedia

Every Wikidata id and URL was checked on 2026-11-05.

Dry run against v2 (read-only, --dry-run, from this worktree, 5 Nov 00:5x UTC; nothing written, max assertion id 3902392 unchanged)

row 1: Emu War → war — refused (subject_held_as_concept)
row 2: Football War → war — would_seed (would mint the subject from Wikidata)
row 3: Dreyfus affair → justice — would_seed (would mint the subject from Wikidata)

Rows 2 and 3 are optimistic: a dry run makes no request, and the real mint goes through SourceIdentity.Creators.classify/2, which returns {:permanent, :not_a_creator_kind} for anything that is not a human or an organisation, so both would be refused subject_not_mintable at seeding.

Why, and the bounded fix this PR needs before it leaves draft

  • Examples.Manifest accepts event subjects and predicate_endpoint_rules allow entity:event → sense for illustrates, but Examples.Seeder mints only through Creators (persons, organisations).
  • The Emu War is held as a concept because the Wikidata walk recognises events only for P31 in Q1656682 Q1190554 Q495307 Q752783 (absorb/sources/wikidata.ex:721); its class is Q645883 military operation. Seeder.held_subject/2 refuses any held concept (subject_held_as_concept).

Fix, in this branch: (1) a general Wikidata mint for event, work and artifact subjects in the seeder, typed by the walk's own entity_kind/2 and required to equal the manifest's kind; (2) the event class list extended with the military-conflict classes (at least Q645883 military operation, Q198 war, Q178561 battle), with the M2 consequence stated; (3) a held concept whose stored wikidata_instance_of now maps to the manifest's kind is re-typed in the seeding transaction and the change recorded, instead of refused; (4) tests: classify fixtures for events, seeder dry-run and seed on these three rows, the existing first-v1 behaviour unchanged. No migration, no config/*.exs or mix.lock change.

Live steps after merge, each on the owner's typed yes, one at a time

Fast-forward deploy (no restart); mix dd.exemplars.seed priv/exemplars/farce-v1.json --as internal-contributor@example.test from the serving checkout with DD_NO_OBAN=1 (three Wikidata requests at Wikidata's own pace); the owner accepts each nomination on /connections/:id as the reviewer account; then the #310 §7 comparison batch.

🤖 Generated with Claude Code

…war, the Dreyfus affair on justice)

Three event rows nominated on the owner's words "Yes, rows 1–3" (5 November
2026), in first-v1's shape: WordNet sense matches, Wikidata ids verified
(Q14665, Q223085, Q173457), dated evidence checked on 2026-11-05, checksum
stamped. Nothing is public until a reviewer accepts on /connections/:id.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@coderabbitai

coderabbitai Bot commented Nov 4, 2026

Copy link
Copy Markdown

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

  • Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant