ruff_python_spo: callee-name residual detail + the drill's proposer stage

[AdaWorldAPI]

Summary

Increment 2 of the self-adaptive drill loop (#96), scoped to what could actually be measured this round. Two pieces.

1. Callee-name detail on non_literal_assign

x = <call>() residuals used to carry the generic "call" tag. Now, when the callee resolves to a terminal name, detail = "call:<name>" (field(...) → "call:field"); an unresolvable callee (a call whose own callee is a call, a lambda, a subscript — never guessed) keeps the generic "call" tag.

Measured payoff on dismech: the 451-row call bucket from #96's histogram decomposes losslessly into three named LinkML schema factories:

355  non_literal_assign/call:PermissibleValue
 96  unresolved_annotation/binop
 91  non_literal_assign/call:EnumDefinition
  5  non_literal_assign/call:ConfigDict

355 + 91 + 5 = 451, exact — this is precisely the drill target #96's own histogram predicted without being able to name.

2. drill.rs — the proposer stage

propose(residuals, min_support) groups a ledger by (reason, detail) for the four reasons dense enough to carry one (curie_constant, non_literal_assign, unresolved_annotation, unresolved_base), thresholded on support. classify_across_corpora(labeled, min_support) runs propose over 2+ labeled corpora and splits each candidate into Generic (cleared the bar in every supplied corpus) vs CorpusScoped (some but not all) — an exact rule, not a similarity heuristic.

Scope boundary — stated in the module doc, and stated again here so it isn't overclaimed: this builds grouping + classification only. It does not build a promotion gate that re-runs extraction with a candidate row "active" and checks the coverage delta — that needs a config-consuming extractor (a trie the plain arm reads before deciding a site's classification) that doesn't exist yet. That's the next increment, not this one.

Measured — and it falsifies my own prediction

Ran plain_propose at min_support=2 over three real corpora (dismech / A2UI SDK / ruff/scripts):

18 candidate rows, 0 generic, 18 corpus_scoped

I'd predicted in the #96 handoff that some row would prove cross-corpus generic. It didn't — worth recording plainly rather than dropping quietly. The closest candidate, unresolved_annotation/binop, fires 96× in dismech and 12× in ruff/scripts but is silent on the A2UI corpus — so under the exact "fires in ALL supplied corpora" rule it stays CorpusScoped. That's the strict rule working as designed: three modest corpora is a small n for "generic," and a fuzzy magnitude-similarity threshold would have manufactured a Generic verdict this data doesn't support. curie_constant rows are, as expected, uniformly CorpusScoped to dismech.

Falsifier discipline

Every new grouping/threshold/classification site verified by a disable run — mutation asserted present, test observed red, restored, full suite green: the min_support filter, the Generic/CorpusScoped split (forced everything to Generic — both cross-corpus tests failed), and the callee-name fallback (both resolved-and-unresolved cases exercised in one fixture).

Test plan

• cargo test -p ruff_python_spo — 75/75 (70 pre-existing + 5 new)
• cargo clippy -p ruff_python_spo --all-targets -- -D warnings — clean
• cargo fmt + RUSTDOCFLAGS="-D warnings" cargo doc --no-deps -p ruff_python_spo — clean (one intra-doc link fixed, same de-link pattern as R2IL intake arm: plan, implementation spec, and ruff_r2il crate (PR 1) #94/ruff_python_spo: residual ledger for the plain arm — silent skips become named, addressed rows #96 — never widen visibility to satisfy a lint)
• uvx prek run --from-ref main --to-ref HEAD — all hooks pass, tree clean after
• plain_propose output reproduced over all three corpora, pasted above

---

Generated by Claude Code

[cursor]

Bugbot couldn't run - usage limit reached

Bugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit.

A user or team admin can review and increase usage limits in the Cursor dashboard.

(requestId: serverGenReqId_aba29efd-6479-4211-bcf0-90e524105a81)

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.