Repository navigation
Conversation
A parenthesis-less code span (`redis`, `data`) resolved to any single callable carrying that label, including a method owned by an unrelated class. The result was a bogus INFERRED `references` edge from the prose to a method the sentence never named: `redis` the server linked to a lone `redis()` on `Client`. A bare, qualifier-less mention now keeps a method candidate only when one of the method's owner classes is itself named in the same document (`successWithWarnings` written beside `HandlerResult` keeps its edge); `redis` with no owner cited does not. The span text in `context` tells `redis` from `redis()`: a real call site writes the parentheses and is unaffected. A bare word still resolves to a free function or a class. Closes Graphify-Labs#4609.
![graphify-labs[bot]](https://avatars.githubusercontent.com/in/4356270?s=60&v=4){kind=small}
There was a problem hiding this comment.
Graphify reviewed this change.
Worth a look — the grounded gate found no coupling regressions or blocking issues, but 1 advisory finding(s) below merit a look before merge.
Not checked: tests were not run; formal verification proved 0 of 1 changed function(s) (1 sampled, not proven).
Formal verification. PR-changed functions: 1/1 verified (0 proven, 1 may-equivalent, 0 distinguished) · 0 not verified.
Graphify review — findings
Tightens markdown mention resolution so a bare, parenthesis-less word like redis no longer links to a lone redis() method on an unrelated class. The method match survives only when its owning class is also named among the mentions (HandlerResult beside successWithWarnings), and free functions, classes and any name() call written with parentheses resolve as before. The owner check builds its set of cited names in resolve_markdown_mentions from all mentions, so the "named alongside it" test can be satisfied by a class cited in a different document rather than only the citing one.
Worth a look
- Owner-class mention set spans every document, not just the citing one —
graphify/markdown_resolution.py:200· Escalate · medium- NOT verified (no proof, no reproducing execution) — needs human review
Analysis details — impact, health, verification
Impact & health
Graphify review
Impact — 465 functions depend on the 39 functions this change touches.
Health — this change adds coupling hotspots:
- worse:
resolve_markdown_mentions()— 1 callers, 9 callees
Verification — 465 functions in the blast radius were not formally verified this run (proofs are advisory here).
Gate & verification
graphify gate
PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.
Advisory (not blocking):
- verification_scope: 211 function(s) in the blast radius were not formally verified this run
Test selection
Test selection
2 of 436 test file(s) selected (0%) via static blast radius.
tests/test_markdown_code_spans.py— impact, changed-testtests/test_watch.py— impact
Selection is safe under the controlled-regression assumption; always-run tests + a periodic full run are the backstops. Advisory — it never changes the check verdict.
Docs that may be stale (advisory)
docs/superpowers/plans/2026-05-04-incremental-updates-dedup.md§ Task 2: Create graphify/dedup.py — entropy gate + MinHash/LSH + Jaro-Winkler (lines 71-468): references changed symbolsitselfgraphify/skill-agents.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitselfgraphify/skill-aider.md§ Step 1 - Ensure graphify is installed (lines 59-106): references changed symbolsitselfgraphify/skill-amp.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitselfgraphify/skill-claw.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitselfgraphify/skill-codex.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitselfgraphify/skill-copilot.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitselfgraphify/skill-devin.md§ Step 1 - Ensure graphify is installed (lines 70-119): references changed symbolsitselfgraphify/skill-droid.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitselfgraphify/skill-kilo.md§ Step 1 - Ensure graphify is installed (lines 65-112): references changed symbolsitself
…and 10 more.
Formal verification
No difference found (not proven): No behavior difference found in resolve\_markdown\_mentions (not a proof).
The verifier ran both versions of resolve\_markdown\_mentions on many inputs and saw identical behavior every time. Strong evidence the change is safe, but evidence, not a proof.
Guarantee: Empirical: differential testing (both versions run on many generated inputs). A divergence on an untested input remains possible, so this is 'no counterexample found', not 'proven equivalent'.
Note: An input the sampler did not try could still differ.
· 1 more finding(s) on lines outside this diff (see the check run).
The set of names a document cites was built from every mention in the corpus, so a class named in a different document could satisfy the owner check for a bare method name here. Scope it per document: the owner class has to be named in the document that cites the method.
This branch has not been deployed
1 participant
What
A parenthesis-less code span in Markdown no longer resolves to a lone method on an unrelated class.
The bare-mention branch of
resolve_markdown_mentionsmatched any single callable carrying the label, and a method's_symbol_labeldrops the leading dot (Client.render→render). So prose like`redis`or`data`linked to a method the sentence never named, producing a bogusINFERRED(0.95)referencesedge:Fix
For a bare, qualifier-less mention, a method candidate is kept only when one of the method's owner classes is itself named in the same document — the primary proposal in #4609, which preserves
successWithWarningswritten besideHandlerResult. The span text already rides incontext, so the resolver distinguishesredisfromredis(); a real call site writes the parentheses and is unaffected. A bare word still resolves to a free function or a class.graphify/markdown_resolution.py: in the bare branch, filter out method candidates whose owner class is not named; add_owning_class/_owner_is_mentionedhelpers, and a per-document set of the names each document cites.tests/test_markdown_code_spans.py: three regression tests — a bare word does not resolve to a lone method (and a paren call still does), a method stays resolvable when its owner class is named alongside it, and a class named in a different document does not keep the edge.Verification
I could not collect the graphify test suite here: the environment has no
tree_sitter, soimport graphify.extractfails at collection. This is a pre-existing environment gap, not related to the change; CI is the judge.To verify the resolver rule itself, I drove
resolve_markdown_mentionsdirectly with a stubtree_sitterand hand-built nodes/edges matching whatextractemits (a_callable_classclass, a method owned via amethodedge, and a free function):redis()(parens) → edge ✅AI authorship
Prepared with MiniMax M3.1 Flash Free via the opencode agent harness (
minimax-m3.1-flash-free).Closes #4609.