fix(dart): why the tri-comparison args shape was so large — 3 residual causes (#2309) - #2342
Merged
Merged
Conversation
…al causes (#2309) Explored `dart/function/args/agree[none]_vs[gitgalaxy,tree_sitter]` (106 occurrences after #2335's named-parameter-group fix, was 205). Every one was `gg=N ts=0` or `gg=1 ts=0` -- i.e. one tool reporting a real count and the other 0. Three distinct causes, none of them a "real" arg-count disagreement: 1. **tree-sitter ground truth wrong for const/factory constructors** (`tree_sitter_accuracy_audit.py::_get_param_count`, reused by the tri-comparison gatherer). `constant_constructor_signature` (`const PointerEvent({...})`) and `factory_constructor_signature` (`factory ThemeData({...})`) don't expose a field-tagged "parameters" child -- unlike the plain `constructor_signature` -- so they fell through to `return 0`. Every `const`/`factory` constructor in the dart corpus (the dominant Flutter widget-constructor form) measured 0. Added both node types to the direct-`formal_parameter_list`-child branch that already handles `function_signature`/`setter_signature`. ~30 large mismatches. 2. **GitGalaxy over-counts a paren-less getter** (`detector.py`). A bodyless `Offset get pan;` reached `_count_top_level_args("Offset get pan")`, which has no `(` to unwrap and returns 1 (whole string as one segment). The bodyless-`;` branch now returns 0 when `not has_parens`. ~15 mismatches. 3. **GitGalaxy under-counts operator overloads** (`detector.py`). The `args` regex needs `NAME(...)`; `bool operator ==(Object other)` has `==` between the name token and the `(`, so `.search()` returns None and the count falls to 0. Added dart to the same "regex failed -> structural counter" fallback c/cpp already use (#2012), guarded on a literal `(` in the signature text so a paren-less getter body still reads 0. ~12 mismatches. Result: the args shape drops 106 -> 6, dart args accuracy 93.9% -> 99.65%, GitGalaxy and tree-sitter now agree on 1733/1739 comparable functions. The 6 survivors are all one unrelated shape -- a multi-line `@Deprecated('...' '...',)` annotation that dart's `func_start` wrongly consumes as a return-type prefix, so the annotation's own parens get counted instead of the real parameter list. Filed separately. Regenerated: both golden masters (26 diffs, all dart -- arg counts + downstream ripple, 0 non-dart content change); `tree_sitter_accuracy_baseline_ dart.json` (`args_exact_match` 1633 -> 1733). No summary-table or javascript- baseline change needed (a post-#2335 companion already refreshed the latter). Verification: test_dart.py (+ new getter/operator slicer case) + tests/ extraction (6235) green; ruff/mypy/dead-key/ast-accuracy clean; crucible_check both modes PASS; tree_sitter_accuracy_audit --all --ci + tri_comparison_chart --all --ci both green. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Contributor
squid-protocol
added a commit
that referenced
this pull request
Aug 27, 2026
#2341) (#2347) dart's `func_start` match spans any leading `@annotation(...)` prefix and the return-type prefix, so `start_idx` (the block start / args-span start) can land on an annotation. When that annotation's argument list is multi-line and carries a top-level comma -- `@Deprecated('Use x. ' 'Deprecated after v3.3.',)` is the corpus shape -- its own `(...)` is the first paren the args counter finds, and the declaration's real parameter list is never measured (`ToolbarOptions` const ctor read 0 args, several deprecated getters read 1). Fix: for dart, every args-span computation anchors at `match.start(match. lastindex)` -- the name capture -- instead of `start_idx`. Same `match.lastindex` name-anchor the typescript/javascript branch of `_slice_by_braces` already uses. `start_idx` still bounds the block itself (line numbers, body slice); only the parameter-count slice moves. This closes out the `dart/function/args/agree[none]_vs[gitgalaxy,tree_sitter]` shape entirely: **0 remaining args mismatches** across the 7-file Flutter corpus (was 205 before #2309, 106 after #2335, 6 after #2342), dart args accuracy 99.65% -> **100.0%** (`args_exact_match` 1733 -> 1739 / 1739). Regenerated: both golden masters (16 diffs, all dart -- arg counts + downstream ripple, 0 non-dart content change); `tree_sitter_accuracy_baseline_dart.json`. No summary-table or javascript-baseline change. Verification: test_dart.py (+ new annotation-prefix slicer case) + tests/ extraction (6236) green; ruff/mypy/dead-key/ast-accuracy clean; crucible_check both modes PASS; tree_sitter_accuracy_audit --all --ci green. Co-authored-by: Joe Esquibel <squid-protocol@users.noreply.github.com> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Explores and largely closes why dart is different in the tri-comparison — the
dart/function/args/agree[none]_vs[gitgalaxy,tree_sitter]shape (106 occurrences after #2335,was 205; dart is the only language with an args shape this large).
Every one of the 106 was
gg=N, ts=0orgg=1, ts=0— one tool reporting a real count, the other 0. None were a genuine arg-count disagreement. Three distinct causes:1. tree-sitter ground truth wrong for
const/factoryconstructors_get_param_countintree_sitter_accuracy_audit.py(reused by the tri-comparison gatherer):constant_constructor_signature(const PointerEvent({...})) andfactory_constructor_signature(factory ThemeData({...})) don't expose a field-taggedparameterschild — unlike plainconstructor_signature— so they hitreturn 0. Every const/factory constructor in the dart corpus measured 0. Added both node types to the direct-formal_parameter_listbranch. ~30 mismatches.2. GitGalaxy over-counts a paren-less getter
Offset get pan;→_count_top_level_args("Offset get pan")has no(to unwrap → returns 1 (whole string as one segment). The bodyless-;branch now returns 0 whennot has_parens. ~15 mismatches.3. GitGalaxy under-counts operator overloads
bool operator ==(Object other)— theargsregex needsNAME(...)but==sits between the name and the(, so.search()→ None → count falls to 0. Added dart to the "regex failed → structural counter" fallback that c/cpp already use (#2012), guarded on a literal(so paren-less getter bodies still read 0. ~12 mismatches.Result
args_exact_match(tree-sitter baseline)The 6 survivors are all one unrelated shape — a multi-line
@Deprecated('...' '...',)annotation that dart'sfunc_startwrongly consumes as a return-type prefix, so the annotation's own parens get counted. Filed as #2341.Regenerated
tree_sitter_accuracy_baseline_dart.json—args_exact_match1633 → 1733files_scanned)Verification
test_dart.py(+ new getter/operator slicer case) +tests/extraction(6235) — greencrucible_check.py— PASS both modestree_sitter_accuracy_audit --all --ci+tri_comparison_chart --all --ci— both green🤖 Generated with Claude Code