Sidecar-enrich OC material/object-type categories onto the published wide parquet (fixes #260)

[rdhyee]

Problem

The Explorer serves material/object-type categories from the frozen iSamples Zenodo export, which carries stale or incorrect concept values for a class of OpenContext samples. This surfaced in two reports:

• interactive explorer bug: material category incorrect on sample #260 — ark:/28722/k2p55x96j (a ceramic) shows "Anthropogenic metal material" in the popup. Verified against live data: the export's list for this PID is [anthropogenicmetal, biogenicnonorganicmaterial, rock] — the correct value "Other anthropogenic material" (otheranthropogenicmaterial) is not present at all. It exists only in Eric's OpenContext PQG, which is more current than the frozen export.
• Related: Issues with material types #265 — the broad SKOS root material leaking into the facet list (mitigated build-side in build: stop SKOS root 'Material' leaking into the material facet (#265) #271, but that can't recover values absent from the export).

No index/selection change on our side can fix #260, because the correct concept isn't in the data we publish. The fix has to bring OpenContext's corrected concept values into the published wide parquet.

Proposed fix: a material/object-type sidecar

Reuse the existing enrichment pattern we already run for OC thumbnails (scripts/enrich_wide_with_oc_thumbnails.py):

1. Source of truth = Eric's OpenContext PQG (narrow/wide), which has the corrected has_material_category / has_sample_object_type edges.
2. Build a pid → {material_uri, object_type_uri} sidecar for OC PIDs.
3. Overlay it onto the published wide parquet (and downstream sample_facets_v2 / facet_summaries via build_frontend_derived.py) before publish, OC rows taking precedence over the frozen export.

This fixes the popup and the underlying facet values in one pass.

Open questions

• Confirm with @ekansa that the OC PQG is the intended source of truth to overlay onto the export.
• Which OC PQG snapshot/URL is canonical for this overlay?
• Scope: material only, or material + object_type + context?
• Sequencing vs. build: stop SKOS root 'Material' leaking into the material facet (#265) #271 (build-side root drop) — sidecar should run before/with build_frontend_derived.py.

Related

• interactive explorer bug: material category incorrect on sample #260 (popup wrong material), Issues with material types #265 / build: stop SKOS root 'Material' leaking into the material facet (#265) #271 (root concept leak), thumbnail sidecar precedent in scripts/enrich_wide_with_oc_thumbnails.py, DATA_PROVENANCE.md (docs+scripts: data provenance map + build scripts for the 6 unscripted derived parquet #264).

---

— 🤖 rbotyee (RY's bot). Spun up from the #260 triage; RY skimmed. Ping @rdhyee.