docs: log Claim 4 (haskell multi-clause function counting) in beats-AST doc - #1617
Merged
Merged
Conversation
…ST doc Investigating haskell func_start's measured 50.5% recall (issue filing pass) found the gap is almost entirely a tree_sitter_accuracy_audit.py ground-truth artifact: tree-sitter-haskell gives each pattern-match equation clause its own node, while GitGalaxy correctly merges all clauses of one function into a single reported entity (#1442). Clause-collapsed ground truth on the same corpus measures 93.9% recall using the same alignment algorithm and unchanged GitGalaxy output -- filed as #1614, with two smaller genuine residual gaps as #1615/#1616. Also removes an unrelated pre-existing duplicate section ("How long can a gap...") accidentally repeated verbatim in Claim 3. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Contributor
The initial writeup used a hand-rolled measurement script's approximate 148 real / 93.9% recall. The actual production fix (#1618) measures 146 real / 95.2% recall and leaves 7 residual misses, not 9 -- the difference was the production fix's comment-sibling handling, which my ad hoc script got wrong for two names (go, camelCaseStrToHyphenated) that turned out not to be genuine misses. Updates the claim to cite the real committed baseline and the actual issue/PR numbers now that #1614/ #1615/#1616/#1618 exist. 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
func_startrecall (measured 50.5%) surfaced that the gap is almost entirely atree_sitter_accuracy_audit.pyground-truth measurement artifact, not a real GitGalaxy defect — logged as Claim 4 indocs/why_gitgalaxy_beats_ast_here.mdper CLAUDE.md's standing instruction for cases where GitGalaxy's structural signal beats tree-sitter's raw output.Detail
Re-walking the pinned 7-file
language-crucible/data/haskell/pandoccorpus with a clause-collapse pass (merging consecutive same-name equation-clause siblings into one occurrence, the same semantics GitGalaxy's own_slice_by_indentationalready applies per #1442) moves measured recall from 50.5% (275 real / 139 found) to 95.2% (146 real / 139 found), using the audit tool's own unchanged_align_occurrences_by_line/_get_param_countand unchanged GitGalaxy output.Issues filed from this investigation, and their status:
where name args = exprlocal bindings are structurally invisible #1615 — residual real gap: single-linewhere name args = expropeners (same shape haskell func_start: single-linelet name args = exprlocal bindings are structurally invisible #1564 fixed forlet) — open=on the naming line) — openTest plan
grep -n "^## \|^### "shows a clean, non-repeating structure).