diff options
| author | Alexander_Kabui <alexanderkabua@gmail.com> | 2026-10-07 09:40:32 +0300 |
|---|---|---|
| committer | Alexander_Kabui <alexanderkabua@gmail.com> | 2026-10-07 09:40:32 +0300 |
| commit | e469eb1d496bb5ca08a8eb264bfaead2816670e3 (patch) | |
| tree | af433b43868128f2e5bfcc798b512fc5312e4c04 /src | |
| parent | 2326f8275a592e266ec8838cf694998d13f4968a (diff) | |
| download | genenetwork4-e469eb1d496bb5ca08a8eb264bfaead2816670e3.tar.gz | |
src/search/native.rs (tests): fix the C3.2/C3.3 parser-test defects the guix attempt-4 check phase exposed.
Three defects in the committed C3.2/C3.3 test code, all found by guix
attempt 4's check phase (drv jlsjsw7i, phase `check', cargo test
--offline; log: logs/guix-c1-3.log attempt-4 section) after the build
phase had already proven the vendoring mechanism and the fork link
(287 s):
1. E0596 x2: the `terms` and `prefixed` closures borrow `qp` mutably,
so both bindings need `let mut` (rustc 1.85.1 diagnostics in the
guix log; the same E0596 sat unnoticed in logs/c3-3.log above a
bogus "LIB_EXIT=0" — the dev-shell gate script swallowed cargo's
exit status; fixed by re-running the gates with explicit status
capture, logs/fix-fulltest.log).
2. The C3.2 boolean test asserted the WRONG index term for group:B6:
`XGbxd`, with a message claiming "group:B6 must lowercase to
XGbxd". That is the r1-test BXD case: search.py:204-209 lowercases
the VALUE and concatenates behind XG, so B6 -> XGb6 (xapian-delve
on the dev index: terms XGb6, XGb6btbrf2, XGb6d2, XGb6d2f2,
XGb6d2ri exist for group B6; XGbxd exists separately for the BXD
value). Corrected to XGb6 / !XGB6. The 13a7cb1 commit message
itself repeats the wrong "group:B6 -> XGbxd" claim — history is
left intact, corrected here and in LANE-REPORT/REVIEW.
3. The C3.3 probabilistic test asserted the non-stemmed field prefix
(XD/XRF/XWK), but with STEM_ALL_Z the parsed field term is
Z+<prefix>+stem — exactly what d563910's own commit message
describes ("wiki:diabetes -> ZXWKdiabet") and what the production
indexer emits (dev index: 9678 ZXWK* terms, 47868 ZXRF* terms,
ZERO XWKZ/XRFZ/XDZ terms). Corrected to ZXD/ZXRF/ZXWK, making the
assertion pin the exact indexed term form instead of a form that
never occurs.
Because of defect 1 neither test ever ran at commit time — the
C3.3 "lib tests green / FULL suite EXIT=0" evidence was a false
positive of the broken exit capture, now re-established with real
status capture.
Diffstat (limited to 'src')
| -rw-r--r-- | src/search/native.rs | 32 |
1 files changed, 20 insertions, 12 deletions
diff --git a/src/search/native.rs b/src/search/native.rs index a7b2fb5..665477a 100644 --- a/src/search/native.rs +++ b/src/search/native.rs @@ -267,7 +267,7 @@ mod tests { #[test] fn boolean_prefixes_parse_to_exact_index_terms() { let mut qp = build_query_parser(); - let terms = |query: &str| -> std::collections::BTreeSet<String> { + let mut terms = |query: &str| -> std::collections::BTreeSet<String> { qp.parse_query_checked::<&str>(query, None, None) .unwrap_or_else(|e| panic!("{query} must parse: {e:?}")) .unique_terms() @@ -288,21 +288,29 @@ mod tests { assert!(terms("chr:1").contains("XC1")); assert!(terms("peakchr:1").contains("XPC1")); // search.py:204-209, 222-223: group values lowercase (xapian - // does not stem boolean fields starting with a capital). + // does not stem boolean fields starting with a capital): + // B6 -> b6, concatenated behind XG. (The r1-test G4 smoke's + // XGbxd was the same rule for value BXD, not a stem.) let b6 = terms("group:B6"); - assert!(b6.contains("XGbxd"), "group:B6 must lowercase to XGbxd, got {b6:?}"); - assert!(!b6.contains("XGBXD"), "group must not emit the raw-case term"); + assert!( + b6.contains("XGb6"), + "group:B6 must lowercase to XGb6, got {b6:?}" + ); + assert!(!b6.contains("XGB6"), "group must not emit the raw-case term"); } /// The probabilistic-prefix registrations (C3.3): description, rif, /// wiki and the uppercase aliases must parse as weighted field /// terms, not free text (search.py:229-233). With STEM_ALL_Z the - /// field term is the prefix + Z-stemmed value (e.g. XDZinsulin), - /// so assert on the prefix, the invariant the registration owns. + /// field term is Z+<prefix>+stem — ZXDinsulin, ZXRF…, ZXWK… — the + /// exact shape the production indexer's TermGenerator emits for + /// these fields (confirmed against the dev index: term + /// ZXWKdiabet exists with 24 gene documents, ZXRF-prefixed terms + /// abound; d.v. the delve probes in LANE-REPORT). #[test] fn probabilistic_prefixes_and_aliases_parse_as_field_terms() { let mut qp = build_query_parser(); - let prefixed = |query: &str, prefix: &str| { + let mut prefixed = |query: &str, prefix: &str| { let set = qp .parse_query_checked::<&str>(query, None, None) .unwrap_or_else(|e| panic!("{query} must parse: {e:?}")) @@ -311,11 +319,11 @@ mod tests { .collect::<std::collections::BTreeSet<_>>(); set.iter().any(|t| t.starts_with(prefix)) }; - assert!(prefixed("description:insulin", "XD")); - assert!(prefixed("rif:insulin", "XRF")); - assert!(prefixed("RIF:insulin", "XRF")); - assert!(prefixed("wiki:diabetes", "XWK")); - assert!(prefixed("WIKI:diabetes", "XWK")); + assert!(prefixed("description:insulin", "ZXD")); + assert!(prefixed("rif:insulin", "ZXRF")); + assert!(prefixed("RIF:insulin", "ZXRF")); + assert!(prefixed("wiki:diabetes", "ZXWK")); + assert!(prefixed("WIKI:diabetes", "ZXWK")); } /// The index-open guards, without touching a real index. |
