Follow gg2rdf to https:// subject IRIs and document the store migration

[retog]

Closes #12.

Companion to plazi/gg2rdf#34 (the fix — gg2rdf mints https:// subjects) and plazi/turtle-hook-nq#8 (the other loader).

What changes

gg2rdf now writes every Plazi resource IRI as https://, the scheme this loader has always named its graphs with. The subject namespaces that have to match gg2rdf's output follow:

• config/config.ts: treatmentUriPrefix and materialCitationUriPrefix, which the single-graph delete is scoped by.
• src/sparql.ts: TAXON_NAME_NAMESPACE and the taxon-concept namespace in SWEEPABLE_NAMESPACES, which the sweep and the taxomplete index are limited to.
• graphUriPrefix is untouched, as the comment on it demands; the doc comments that explained the two prefixes by their scheme now explain them by their role.
• test-data/000040332F2853C295734E7BD4190F05.ttl is regenerated gg2rdf output from the branch in Mint Plazi resource IRIs with https:// as the canonical scheme gg2rdf#34 (byte-identical to flipping the scheme in the old fixture); the inline test data, col_test.ts's CoL→taxon-name link and the README's sweep query follow suit.

Store migration

A store loaded before the switch holds http:// subjects that this loader can no longer find — in single-graph mode the delete follows links from the stored treatment IRI, so an update to an old-style treatment would insert the new one next to the old rather than replace it. The README gains a section "Migrating a store to https:// IRIs": the DELETE/INSERT WHERE that renames every affected IRI in subject and object position, scoped to our graph and nothing else; the graph-per-file variant; why predicates need no renaming; and the order of operations (stop loader → rename → deploy → start). This is the text from the discussion in plazi/gg2rdf#33, placed where an operator will look for it.

Test engine

src/test_deps.ts moves oxigraph from 0.4.11 to 0.5.11 (test-only dependency, not in the image). Reason, and worth knowing: with 0.4.11, the owned-subject DELETE leaves the material citations behind once the filter prefixes are https:// — while a SELECT with the identical WHERE clause returns them. Narrowed down to the STRSTARTS(STR(?owned), "https://treatment.plazi.org/id/<id>/") clause; replacing its literal, or dropping it, makes 0.4.11 delete correctly, so this is an optimiser fault in the update path of the old version, not a property of the statement. 0.5.11 evaluates it correctly. The production endpoints (GraphDB on LINDAS, Blazegraph) are not oxigraph.

Verification

deno test --allow-read --allow-env --allow-net src/: 27 passed, 0 failed (same count as main).

🤖 Generated with Claude Code

[Copilot]

Copilot review overview

🔵 Needs a closer look

The graph-per-file migration omits COL_GRAPH, and the documented issue references need correction.

Review effort: Lite
Findings: None

What changed in this PR

Updates Turtle-Hook to consume gg2rdf’s canonical https:// resource IRIs and documents store migration.

Changes:

• Updates namespaces, fixtures, tests, and Catalogue of Life links.
• Adds migration guidance for both store graph modes.
• Upgrades the test-only Oxigraph dependency.

File	Summary
test-data/​000040332F2853C295734E7BD4190F05.ttl	Regenerated fixture with HTTPS IRIs.
src/​test_deps.ts	Upgrades test-only Oxigraph.
src/​sparql.ts	Updates taxon namespaces and documentation; nit (1 vote) to reference #34 for the subject-IRI change.
src/​sparql_test.ts	Adjusts SPARQL fixtures and assertions for HTTPS IRIs.
src/​col_test.ts	Updates Catalogue of Life test links.
README.md	Documents HTTPS migration; moderate issue (1 vote): graph-per-file migration omits COL_GRAPH; nit (1 vote) to correct issue references.
config/​config.ts	Updates subject prefixes; nit (1 vote) to reference #34 for the subject-IRI change.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[retog]

Addressed the Copilot review in eafcdcc:

• References: README.md, config/config.ts and src/sparql.ts now cite Mint Plazi resource IRIs with https:// as the canonical scheme gg2rdf#34 (merged) with #33 as the issue.
• COL_GRAPH in the graph-per-file migration: left out on purpose, now stated in the README. The Catalogue of Life data built by plazi/catologueoflife-to-rdf contains only catalogueoflife.org IRIs and no Plazi resource, so the rename has nothing to do there. The owl:sameAs to a taxon-name.plazi.org IRI in src/col_test.ts is a synthetic fixture line, not something the real CoL graph holds.