fix(groovy): tri-comparison manual verification -- 3 real engine defects fixed

[squid-protocol]

Summary

tree_sitter_language_pack's "groovy" grammar loads without error but can't parse real
Groovy declarations -- confirmed via direct inspection (30-file/13K-node sample, 3.2%
ERROR rate, no class_declaration/method_declaration node type at all). The two node
types this repo's NODE_MAPS used to map groovy to ("func", "generics_class") turned
out to be semantically wrong, not just imprecise: "func" nodes are call expressions
(getLogger(JiraService.class)), not definitions; "generics_class" nodes are generic
type usage (Map<String, X>), not class declarations. ctags has zero Groovy support at
all. Groovy is now routed through the gg_only manual-verification treatment (same category
abap/dockerfile/jcl/livecode/yaml are already in) instead of a broken tree-sitter comparison.

With no automated ground truth available, ran a manual verification per the
tri-comparison-ledger-sweep skill's fallback procedure: an independent grep-based
ground-truth scan across the full 188-file language-crucible v1.2.0 groovy corpus,
cross-checked against the real pipeline's DB output. Found and fixed three real,
independent GitGalaxy engine defects:

1. class_start had zero capture groups -- class existence was detected correctly,
   but every one of 286 named classes was literally called "Anonymous_Class" instead of
   its real name. The only language in _CLASS_START_NAMED_EXTRACTION_LANGS with this
   defect (checked all 36).
2. Prism never shielded groovy's """/''' multi-line strings -- code-shaped fixture
   text embedded in one (Gradle integration-test build-script snippets, Spock's own
   compiler-smoke-test source samples) leaked into the real code stream and corrupted
   extraction for the rest of the file via desynced brace-counting. New
   _strip_groovy_triple_quoted_strings in prism.py, same idiom as the existing
   PowerShell here-string / Lua long-bracket precedents.
3. func_start's zero-prefix branch matched the bare def keyword as a function name
   once a Spock quoted feature-method name got correctly blanked for brace-safety. Fixed two
   ways (excluded def from the exclusion lookahead; preserved a quoted name following
   def in the brace-safety shield instead of blanking it), plus a related _extract_name
   fix (it silently truncated a quoted multi-word name to its last word) -- confirmed
   zero-blast-radius elsewhere since groovy is currently the only language whose
   func_start captures a quoted name at all.

Post-fix: zero unexplained recall gaps across the whole corpus for either functions or
classes. One confirmed, unfixed false-positive class remains -- MarkupBuilder/
Jenkins-Pipeline-DSL trailing-closure calls (button(...) { }) are syntactically
indistinguishable from a real zero-prefix declaration -- filed as
#2530 rather than patched with a
fragile denylist (needs its own design pass). Verified: 661/718 (92.1%) named functions,
222/222 (100%) named classes.

Full write-up: docs/language_status/groovy.md §9.

Type of change

• Bug fix
• New feature / language support
• Parsing or engine logic (gitgalaxy/core/detector.py, language_standards.py, prism.py, or a per-language rule)
• Docs, tooling, or CI only
• Other (describe above)

CI checklist

• python tests/tools/audit_check.py passes
• python -m pytest tests/ passes -- 7145 passed
• python tests/tools/crucible_check.py passes -- both golden masters re-blessed after confirming the diff (2109 changes) traced entirely to the groovy fix plus expected small ripple (see Verification below)
• python tests/tools/tree_sitter_accuracy_audit.py --all --ci passes -- groovy no longer measured this way (see Summary)

Differential Scan target

Not applicable to a single repo -- verified against the full ~80-repo crucible corpus (see Verification).

Verification

$ python tests/tools/audit_check.py                                    -- clean
$ python -m pytest tests/                                              -- 7145 passed
$ python tests/tools/tree_sitter_accuracy_audit.py --summary-table     -- groovy removed, zero diff otherwise
$ python tests/tools/tri_comparison_chart.py --all --write             -- groovy now renders as gg_only (1-bar)

Full-corpus differential scan, done twice for control:

1. Ran a standalone golden_diff.deep_compare() (uncapped, unlike the 50-line-truncated
   pytest failure message) against the committed golden master with the 3 fixes applied:
   2109 diffs. 1587 directly groovy-attributable (directory_groups/groovy/* mass/exposure
   recomputing from corrected counts). The remaining 522 are all tiny (<1% relative, one
   5.5% outlier that's a repo-wide aggregate directly downstream of the groovy change) --
   topological-coordinate ripple from the shared force-directed graph layout recomputing
   when any node's structural mass changes, and global risk-normalization shifting slightly
   -- explicitly called out as an acceptable ripple class in CLAUDE.md's Differential Scan
   protocol. Zero off-target anomalies in any unrelated file.
2. Control run: stashed all code changes, ran the identical scan against pristine
   origin/main -- zero diffs, confirming the 2109 diffs are entirely attributable to
   this PR's fixes, not environment noise.
3. Re-blessed both golden masters (crucible_check.py --update --yes), then re-ran
   crucible_check.py plain -- both full_precision and zero_dependency now PASS clean.

Other changes in this PR

• tests/tools/tree_sitter_accuracy_audit.py: removed groovy's broken NODE_MAPS entry,
  documented the exclusion (same style as the existing cobol/dockerfile/scheme/yaml bullets),
  deleted the now-dead tests/tree_sitter_accuracy_baseline_groovy.json.
• tests/tools/tri_comparison_chart.py: added groovy to _GG_ONLY_LANGS so it renders
  correctly as a GitGalaxy-only bar instead of silently disappearing from --all; fixed
  stale language-count comments this move affected (45 total unchanged, groovy just moved
  buckets -- was mis-stated as 46 in an earlier draft of this PR, corrected).
• tests/tools/tri_comparison_reconcile.py, tests/tools/ctags_reader.py: matching
  comment-accuracy fixes for the same bucket move.
• docs/self_scan/tri_comparison_ledger.json: validated the pre-existing
  groovy/function/existence/agree[gitgalaxy]_vs[tree_sitter] shape
  (credit_tools: [gitgalaxy]) -- verified only this one entry's status/verdict changed,
  everything else is the normal chart-regen timestamp/count refresh.
• docs/self_scan/manual_verification.json: new groovy entry (function/class), same
  schema as abap/dockerfile/jcl/livecode/yaml.
• docs/language_status/groovy.md: new, sections 1-8 built via the language-status skill
  from primary sources (rules dict, test counts, closed issues, gitgalaxy-raw-output
  evidence), §9 the full tri-comparison write-up.

Co-Authored-By: Claude Sonnet 5 noreply@anthropic.com

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab