Conversation
Handle .clj/.cljs/.cljc/.edn via the clojure grammar already bundled in tree-sitter-language-pack. Extracts namespaces, def* forms, protocols, records/types, :import classes, and resolves :require targets and cross-namespace calls in a post-extraction pass.
Field-tested on ring, core.async and two internal projects (400 files): - mangle operator characters in node ids so <!! / >!! and foo? / foo! no longer collapse onto one node - exclude record method implementations from cross-namespace call resolution; calls resolve to the protocol signature instead - register namespaced custom definers (helix/defnc, t/deftest) and definitions nested in top-level when/if/do/let forms - resolve alias/->Record and alias/map->Record constructors
Audited against ring (81 files, 404 top-level defs): - names bound by parameters, let-style forms, destructuring, letfn and catch shadow same-named top-level defs, so (handler req) on a handler parameter is no longer a call to the handler function - reify/proxy/extend-* forms inside a body emit implements edges and walk method bodies without treating impl heads as calls
No new packages: the extra reuses the already-locked tree-sitter-language-pack, so only the extra tables change.
![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 5 advisory finding(s) below merit a look before merge.
Graphify review — findings
Adds a Clojure/ClojureScript AST extractor covering .clj/.cljs/.cljc/.edn, wiring the new clojure extra and extract_clojure into the dispatch, extension, and language-family tables so these files classify as code and route to the grammar-backed extractor. Emits nodes for the ns form, def* forms (functions, macros, multimethods, vars, protocols, records, types), protocol/record methods, and Java :import targets, plus contains/implements and intra-file calls edges — driven by head-symbol inspection since the bundled grammar has no defn/ns concept, with a def-prefix heuristic (and a denylist) to catch custom definers like deftest. Records cross-namespace calls and :require targets as raw entries and resolves them across files via resolve_clojure_namespaces after all files are extracted.
Worth a look
- Clojure extraction now raises for files without an ns form —
graphify/extractors/clojure.py· Escalate · high- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- Arrow-prefixed Clojure functions never resolve as calls —
graphify/extractors/clojure.py· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- Qualified Clojure calls can be suppressed by unrelated refer-all namespaces —
graphify/extractors/clojure.py· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- Function-wide bound set suppresses valid calls outside local binding scope —
graphify/extractors/clojure.py· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- Namespaced fdef custom definer is never recognized —
graphify/extractors/clojure.py· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
Analysis details — impact, health, verification
Impact & health
Graphify review
Impact — 3094 functions depend on the 493 functions this change touches.
Health — this change adds coupling hotspots:
- new:
extract()— 703 callers, 45 callees - new:
_rebuild_code()— 144 callers, 55 callees - new:
detect()— 112 callers, 15 callees - new:
to_obsidian()— 38 callers, 14 callees - new:
_extract_generic()— 18 callers, 29 callees - new:
save_manifest()— 40 callers, 11 callees - new:
to_json()— 58 callers, 7 callees - new:
extract_js()— 87 callers, 4 callees - …and 80 more — each is listed as a finding
Verification — 3094 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: 2775 function(s) in the blast radius were not formally verified this run
Test selection
Test selection
299 of 299 test file(s) selected (100%) via static blast radius.
Escalated to a full run for safety — the selection is not trustworthy on its own (see below). CI should run the whole suite.
tests/test_affected_cli.py— full-run-safetytests/test_affected_member_seed.py— full-run-safetytests/test_agents_platform.py— full-run-safetytests/test_analyze.py— impact, full-run-safetytests/test_anthropic_custom_endpoint.py— full-run-safetytests/test_antigravity_install.py— full-run-safetytests/test_apm_fallback_version.py— full-run-safetytests/test_architecture_doc.py— full-run-safetytests/test_astro_extraction.py— impact, full-run-safetytests/test_astro_import_ids.py— impact, full-run-safetytests/test_atomic_canvas_export.py— impact, full-run-safetytests/test_atomic_version_stamp.py— full-run-safetytests/test_atomic_writes.py— impact, full-run-safetytests/test_backend_env_isolation.py— full-run-safetytests/test_backend_extras.py— full-run-safetytests/test_benchmark.py— full-run-safetytests/test_benchmark_raw_graph.py— full-run-safetytests/test_build.py— impact, full-run-safetytests/test_build_merge_dedup_scope.py— full-run-safetytests/test_build_merge_hyperedges_and_prune.py— full-run-safetytests/test_build_merge_shrink_guard.py— full-run-safetytests/test_builtin_global_type_refs.py— impact, full-run-safetytests/test_cache.py— full-run-safetytests/test_callflow_html.py— full-run-safetytests/test_cargo_introspect.py— impact, full-run-safetytests/test_cargo_missing_manifest.py— full-run-safetytests/test_carried_hyperedge_remap.py— impact, full-run-safetytests/test_case_sensitive_resolution.py— impact, full-run-safetytests/test_charmap_encoding.py— impact, full-run-safetytests/test_chunking.py— impact, full-run-safetytests/test_cjs_module_extension.py— impact, full-run-safetytests/test_claude_cli_backend.py— impact, full-run-safetytests/test_claude_md.py— full-run-safetytests/test_cli_broken_pipe.py— full-run-safetytests/test_cli_export.py— impact, full-run-safetytests/test_cli_help.py— full-run-safetytests/test_clojure_extractor.py— impact, changed-test, full-run-safetytests/test_cluster.py— full-run-safetytests/test_cobol_extractor.py— impact, full-run-safetytests/test_codebuddy.py— full-run-safetytests/test_community_hub_labels.py— full-run-safetytests/test_community_labels_skill.py— impact, full-run-safetytests/test_confidence.py— impact, full-run-safetytests/test_corrupt_graph_json.py— full-run-safetytests/test_cpp_nested_and_cli.py— impact, full-run-safetytests/test_cpp_objc_cross_file_calls.py— impact, full-run-safetytests/test_cpp_preprocess.py— full-run-safetytests/test_cross_extension_reexport_self_cycle.py— impact, full-run-safetytests/test_cross_language_call_resolution.py— impact, full-run-safetytests/test_cross_repo_external_call_guards.py— impact, full-run-safety- … and 249 more
non-code file(s) changed (
README.md,pyproject.toml,tests/fixtures/new_languages/sample.clj,uv.lock) → running the full suite for safety (a code graph can't see config/fixture/data deps)
changed code file(s) with no mapped test (
README.md,graphify/extractors/__init__.py,graphify/extractors/clojure.py,pyproject.toml) — a coverage gap or a missing link — running the full suite rather than only the selected tests
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.
· 2 grounded finding(s) anchored inline below; 86 more finding(s) on lines outside this diff (see the check run).
| add_edge(caller, candidates[0], "calls", "remote_call", raw) | ||
|
|
||
|
|
||
| def extract_clojure(path: Path) -> dict: |
There was a problem hiding this comment.
extract_clojure()
fans out to 18 callees (efferent coupling).
Grounded coupling-delta finding (deterministic), not an LLM guess.
There was a problem hiding this comment.
Expected for a new language extractor: extract_clojure is the per-file entry point and fans out to the same helper set (_make_id, _read_text, the libspec/import parsers, node/edge builders) that extract_erlang and extract_commonlisp do. The callee count reflects the number of Clojure forms handled, not shared coupling with the rest of the package; nothing outside extractors/clojure.py gained callers.
| add_edge(owner_id, method_id, "contains", item) | ||
| bodies.append((impl_values[1:], method_id)) | ||
|
|
||
| def process_definition(form: Node, head: tuple[str | None, str]) -> bool: |
There was a problem hiding this comment.
process_definition()
fans out to 9 callees (efferent coupling).
Grounded coupling-delta finding (deterministic), not an LLM guess.
There was a problem hiding this comment.
process_definition is the dispatch over the def* families (fn, var, type, protocol, defmethod, extend-*, custom definers); its nine callees are the per-family handlers within the same file. Splitting it further would only move the fan-out into a table. Happy to restructure if maintainers prefer a dispatch dict, but it is local to the extractor.
…ames - shadowing is now lexically scoped: a let/fn/catch binding hides a same-named def only inside that form, not the whole definition - arrow-prefixed function names (->seconds) resolve as calls; only names matching a record/type are treated as constructors, locally and in the cross-namespace resolver - a qualified call resolves only in its own namespace, never widened by :refer :all namespaces - top-level forms in files without an ns form attach to the file node - fix a misleading comment that listed s/fdef as a custom definer
|
Addressed the review findings in cf0cb87:
Tests added for each. Suite and the ring/core.async field run are green. |
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 5 advisory finding(s) below merit a look before merge.
Graphify review — findings
Adds first-class Clojure/ClojureScript/EDN support: a new extract_clojure extractor parses .clj/.cljs/.cljc/.edn files into namespace, def*, protocol, record/type, and :import nodes plus contains/implements and intra-file calls edges by inspecting each top-level form's head symbol. Registers the extension family across analyze, detect's CODE_EXTENSIONS, and the extract dispatch/resolver tables, with resolve_clojure_namespaces wiring cross-namespace calls and :require targets after all files are extracted. Documents the new clojure extra and supported-language rows in the README.
Worth a look
- process_definition uses undefined variable custom_id —
graphify/extractors/clojure.py· Escalate · high- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
definterfaceis denylisted as a non-definition —graphify/extractors/clojure.py:53· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- Clojure operator escaping still collides with valid symbol names —
graphify/extractors/clojure.py:205· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- pending_impls resolves protocol_ns but never uses it for target lookup —
graphify/extractors/clojure.py· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
- extend-protocol method implementations lose their protocol metadata —
graphify/extractors/clojure.py· Escalate · medium- agreed by 2 of 2 members but NOT verified (no proof, no reproducing execution) — consensus is not a verdict; needs human review
Analysis details — impact, health, verification
Impact & health
Graphify review
Impact — 3100 functions depend on the 499 functions this change touches.
Health — this change adds coupling hotspots:
- new:
extract()— 707 callers, 45 callees - new:
_rebuild_code()— 144 callers, 55 callees - new:
detect()— 112 callers, 15 callees - new:
to_obsidian()— 38 callers, 14 callees - new:
_extract_generic()— 18 callers, 29 callees - new:
save_manifest()— 40 callers, 11 callees - new:
to_json()— 58 callers, 7 callees - new:
extract_js()— 87 callers, 4 callees - …and 81 more — each is listed as a finding
Verification — 3100 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: 2781 function(s) in the blast radius were not formally verified this run
Test selection
Test selection
299 of 299 test file(s) selected (100%) via static blast radius.
Escalated to a full run for safety — the selection is not trustworthy on its own (see below). CI should run the whole suite.
tests/test_affected_cli.py— full-run-safetytests/test_affected_member_seed.py— full-run-safetytests/test_agents_platform.py— full-run-safetytests/test_analyze.py— impact, full-run-safetytests/test_anthropic_custom_endpoint.py— full-run-safetytests/test_antigravity_install.py— full-run-safetytests/test_apm_fallback_version.py— full-run-safetytests/test_architecture_doc.py— full-run-safetytests/test_astro_extraction.py— impact, full-run-safetytests/test_astro_import_ids.py— impact, full-run-safetytests/test_atomic_canvas_export.py— impact, full-run-safetytests/test_atomic_version_stamp.py— full-run-safetytests/test_atomic_writes.py— impact, full-run-safetytests/test_backend_env_isolation.py— full-run-safetytests/test_backend_extras.py— full-run-safetytests/test_benchmark.py— full-run-safetytests/test_benchmark_raw_graph.py— full-run-safetytests/test_build.py— impact, full-run-safetytests/test_build_merge_dedup_scope.py— full-run-safetytests/test_build_merge_hyperedges_and_prune.py— full-run-safetytests/test_build_merge_shrink_guard.py— full-run-safetytests/test_builtin_global_type_refs.py— impact, full-run-safetytests/test_cache.py— full-run-safetytests/test_callflow_html.py— full-run-safetytests/test_cargo_introspect.py— impact, full-run-safetytests/test_cargo_missing_manifest.py— full-run-safetytests/test_carried_hyperedge_remap.py— impact, full-run-safetytests/test_case_sensitive_resolution.py— impact, full-run-safetytests/test_charmap_encoding.py— impact, full-run-safetytests/test_chunking.py— impact, full-run-safetytests/test_cjs_module_extension.py— impact, full-run-safetytests/test_claude_cli_backend.py— impact, full-run-safetytests/test_claude_md.py— full-run-safetytests/test_cli_broken_pipe.py— full-run-safetytests/test_cli_export.py— impact, full-run-safetytests/test_cli_help.py— full-run-safetytests/test_clojure_extractor.py— impact, changed-test, full-run-safetytests/test_cluster.py— full-run-safetytests/test_cobol_extractor.py— impact, full-run-safetytests/test_codebuddy.py— full-run-safetytests/test_community_hub_labels.py— full-run-safetytests/test_community_labels_skill.py— impact, full-run-safetytests/test_confidence.py— impact, full-run-safetytests/test_corrupt_graph_json.py— full-run-safetytests/test_cpp_nested_and_cli.py— impact, full-run-safetytests/test_cpp_objc_cross_file_calls.py— impact, full-run-safetytests/test_cpp_preprocess.py— full-run-safetytests/test_cross_extension_reexport_self_cycle.py— impact, full-run-safetytests/test_cross_language_call_resolution.py— impact, full-run-safetytests/test_cross_repo_external_call_guards.py— impact, full-run-safety- … and 249 more
non-code file(s) changed (
README.md,pyproject.toml,tests/fixtures/new_languages/sample.clj,uv.lock) → running the full suite for safety (a code graph can't see config/fixture/data deps)
changed code file(s) with no mapped test (
README.md,graphify/extractors/__init__.py,graphify/extractors/clojure.py,pyproject.toml) — a coverage gap or a missing link — running the full suite rather than only the selected tests
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.
· 3 grounded finding(s) anchored inline below; 86 more finding(s) on lines outside this diff (see the check run).
| continue | ||
| stack.append((item, bound)) | ||
|
|
||
| def walk_body(items: list[Node], caller_id: str) -> None: |
There was a problem hiding this comment.
walk_body()
fans out to 6 callees (efferent coupling).
Grounded coupling-delta finding (deterministic), not an LLM guess.
This branch has not been deployed
1 participant
What does this PR do?
Adds a Clojure / ClojureScript extractor for
.clj,.cljs,.cljcand.edn, backed by theclojuregrammar already bundled intree-sitter-language-pack. No new dependency: the newclojureextra reuses the pin shared with theranderlangextras, anduv.lockonly gains the extra tables.Per file it extracts the namespace, every
def*form (functions, macros, multimethods anddefmethods, vars, protocols with their method signatures, records/types with their method implementations, and custom definers such asdeftestorhelix/defnc), Java:importclasses,contains/implementsedges, and intra-filecallsedges.A registered
LanguageResolverthen links:requiretargets to the namespace node(s) that declare them (or one external node per undefined namespace) and resolves cross-namespace calls through aliases,:refer,:refer :alland:useto the unique matching definition, with the same god-node guard the other resolvers use. Aliased record constructors (alias/->Record) becomereferencesedges.Precision safeguards: names bound by parameters,
let-style forms, destructuring,letfnandcatchshadow same-named top-level defs;reify/proxy/extend-*inside a body emitimplementsedges rather than treating method heads as calls; Java interop, quoted data,commentforms and#_discards are skipped. Operator characters in symbols (<!!,foo?) are mapped to readable id tokens so distinct definitions never collide.Type of change
How was this tested?
Local runs were on Python 3.14 (macOS). The code uses
from __future__ import annotationsand no 3.11+ syntax.Graphify-specific checklist
skillgen --checkpasses.)