Stop a catch-all namespace from shadowing specific fallback mappings

[kevinschaper]

Fixes #558.

Problem

contract() consults the default prefix maps only when prefix_maps returns zero matches:

curie_list = contract_uri(uri, prefix_maps)
if len(curie_list) == 0:
    if fallback:
        curie_list = contract_uri(uri, default_curie_maps)

The JSON-LD context carries OBO -> http://purl.obolibrary.org/obo/, a catch-all that matches every OBO IRI. So curie_list is never empty for one, the fallback never runs, and obo_context's per-ontology mappings are unreachable:

>>> PrefixManager().contract("http://purl.obolibrary.org/obo/DDPHENO_0000001")
'OBO:DDPHENO_0000001'      # expected 'DDPHENO:0000001'

165 OBO idspaces are affected, including DDPHENO, FBbt, WBbt, EMAPA, ZFA, XAO, CHR, ZFS, OBA and FBdv. The MONARCH -> https://monarchinitiative.org/ catch-all hides BNODE, ISBN, ISBN-10, ISBN-13 and OMIA-breed the same way.

Latent until 2.7.0: TsvSource.set_prefix_map used to be destructive, and the transformer calls source.set_prefix_map({}) when the caller supplies no map, so the map was wiped to five entries and every OBO IRI fell through to the fallback and contracted correctly. Making it additive (#548) restored the map, and with it the dormant catch-all.

Fix

Consider both tiers and keep the most specific match — the candidate that consumed the longest IRI prefix — with prefix_maps winning ties so a caller-supplied map stays canonical. No OBO-specific special-casing; fallback=False is untouched.

prefixcommons.contract_uri already prefers the shortest CURIE when both mappings are in scope, so this only changes which tiers get compared, not how a single tier resolves.

Blast radius

Probed every namespace in all three contexts (488 URIs), patched vs unpatched PrefixManager.contract:

change	n
OBO: catch-all → specific idspace	165
MONARCH: catch-all → specific (BNODE, ISBN, ISBN-10, ISBN-13, OMIA-breed)	5
other	1 — WIKIDATA:Property:x vs WIKIDATA_PROPERTY:x, which is nondeterministic on master anyway

Everything that resolves correctly today is unchanged: HP:, MONDO:, GO:, RO:, BFO:, CHEBI:, UBERON:, HGNC: via identifiers.org, and all biolink: vocab terms. IRIs beneath the OBO namespace that no per-ontology mapping covers still contract to the catch-all — OBO:fbbt#has_function_in, OBO:go/extensions/ro_0002092 — because monarch_context supplies OBO in the fallback tier.

Test plan

• 20 new cases: tests/unit/test_kgx_utils.py covers catch-all vs fallback, prefix_maps winning a tie against an equally specific fallback mapping, and fallback=False still yielding OBO:DDPHENO_0000001. tests/unit/test_prefix_manager.py covers the eight regressed idspaces, four controls, and a relation IRI that must keep OBO:.
• 11 of the new cases fail on master, all pass with the fix.
• Full unit suite: 393 passed, 19 skipped, 1 failed. The failure is test_sink/test_jsonl_sink.py::test_write_jsonl2, which fails identically on unpatched master — pre-existing and unrelated.

Downstream

This is what untyped ~52k nodes in kg-phenio from the 2026-06-03 release on: its transform looks node categories up by CURIE prefix, and OBO isn't an ontology, so DDPHENO/FBbt/EMAPA/WBbt/ZFA/XAO/CHR/ZFS terms have been shipping as biolink:NamedThing.

https://claude.ai/code/session_01BzdpULg2jTEP8wN3jBNzrN