diff options
| author | Pjotr <wrk@bosa.lan> | 2026-11-06 13:36:09 +0000 |
|---|---|---|
| committer | Pjotr <wrk@bosa.lan> | 2026-11-06 13:36:09 +0000 |
| commit | 3cfacfc429451d39db96fe57904db9bb04ed18d1 (patch) | |
| tree | e7e8e8fc1ed6b95c5a68eacebd8ff492dc498c0e /test/python | |
| parent | 4ea099490114ec845285642280d4def715fa0c20 (diff) | |
| download | genenetwork4-3cfacfc429451d39db96fe57904db9bb04ed18d1.tar.gz | |
search: cover the native engine with a live database
The suite covered each half of /search separately: tests/search.rs pins the MariaDB side (locus, abbreviation, complete/partial) but through the gn3 stub engine, and the search_native* tests pin the native engine against a dev index and never need a database. Native engine + live db_webqtl -- the combination that actually runs in production -- was only ever checked by hand. tests/search_native_db.rs closes that: route level, dataset:bxdpublish must come back 200 with meta.state complete and annotated hits; at the engine seam, dataset:BXDPublish must match the same documents as dataset:bxdpublish; and the match set's ids must be the database's PublishXRef ids for the BXD group. The strict id comparison runs only against a full index, selected with GN4_FULL_INDEX (skipped by default, like GN4_PARITY_FIXTURES). Against the dev index the ids are printed as reference data instead: it is a different snapshot, missing 1 772 of the database's traits and holding 349 the database no longer has, which is exactly what made the strict form worth having.
Diffstat (limited to 'test/python')
0 files changed, 0 insertions, 0 deletions
