fix Rust cross-crate spurious INFERRED edges: skip scoped_identifier and trait-method blocklist from raw_calls (#908)

graphify/extract.py

@@ -3610,6 +3610,17 @@ def extract_go(path: Path) -> dict:
 
 # ── Rust extractor (custom walk) ──────────────────────────────────────────────
 
+# Common Rust trait/stdlib method names that appear in virtually every codebase.
+# Resolving these cross-file produces spurious INFERRED edges across crate
+# boundaries (issue #908) — skip them from the unresolved-call queue entirely.
+_RUST_TRAIT_METHOD_BLOCKLIST: frozenset[str] = frozenset({
+    "new", "default", "parse", "from_str", "now", "clone", "into", "from",
+    "to_string", "to_owned", "len", "is_empty", "iter", "next", "build",
+    "start", "run", "init", "app", "get", "set", "push", "pop", "insert",
+    "remove", "contains", "collect", "map", "filter", "unwrap", "expect",
+    "ok", "err", "some", "none", "send", "recv", "lock", "read", "write",
+})
+
 def extract_rust(path: Path) -> dict:
     """Extract functions, structs, enums, traits, impl methods, and use declarations from a .rs file."""
     try:
@@ -3740,6 +3751,7 @@ def extract_rust(path: Path) -> dict:
             func_node = node.child_by_field_name("function")
             callee_name: str | None = None
             is_member_call: bool = False
+            is_scoped_call: bool = False
             if func_node:
                 if func_node.type == "identifier":
                     callee_name = _read_text(func_node, source)
@@ -3749,6 +3761,10 @@ def extract_rust(path: Path) -> dict:
                     if field:
                         callee_name = _read_text(field, source)
                 elif func_node.type == "scoped_identifier":
+                    # Type::method() — still allow in-file EXTRACTED match, but
+                    # skip cross-file resolution: bare last-segment lookup ignores
+                    # crate boundaries and produces spurious INFERRED edges (#908).
+                    is_scoped_call = True
                     name = func_node.child_by_field_name("name")
                     if name:
                         callee_name = _read_text(name, source)
@@ -3769,7 +3785,7 @@ def extract_rust(path: Path) -> dict:
                             "source_location": f"L{line}",
                             "weight": 1.0,
                         })
-                else:
+                elif not is_scoped_call and callee_name.lower() not in _RUST_TRAIT_METHOD_BLOCKLIST:
                     raw_calls.append({
                         "caller_nid": caller_nid,
                         "callee": callee_name,

tests/fixtures/crate_a/Cargo.toml

@@ -0,0 +1,4 @@
+[package]
+name = "crate_a"
+version = "0.1.0"
+edition = "2021"

tests/fixtures/crate_a/src/lib.rs

@@ -0,0 +1,7 @@
+pub fn start() -> bool {
+    true
+}
+
+pub fn parse(s: &str) -> u32 {
+    s.len() as u32
+}

tests/fixtures/crate_b/Cargo.toml

@@ -0,0 +1,4 @@
+[package]
+name = "crate_b"
+version = "0.1.0"
+edition = "2021"

tests/fixtures/crate_b/src/lib.rs

@@ -0,0 +1,18 @@
+// crate_b has no dependency on crate_a.
+// These calls use Type::method() (scoped_identifier) and common names that
+// previously produced spurious INFERRED edges into crate_a (#908).
+
+pub struct Server;
+
+impl Server {
+    pub fn run(&self) {
+        // Server::start() — scoped call, should not wire to crate_a::start
+        let _ = Server::start();
+        // Url::parse() — scoped call, should not wire to crate_a::parse
+        let _ = Url::parse("http://example.com");
+    }
+
+    fn start() -> bool {
+        false
+    }
+}

tests/test_multilang.py

@@ -177,6 +177,27 @@ def test_rust_no_dangling_edges():
             assert e["source"] in node_ids
 
 
+def test_rust_no_cross_crate_spurious_edges():
+    """Scoped calls (Type::method) and blocklisted names must not produce
+    INFERRED cross-crate calls edges (#908)."""
+    from graphify.extract import extract
+    crate_a = FIXTURES / "crate_a" / "src" / "lib.rs"
+    crate_b = FIXTURES / "crate_b" / "src" / "lib.rs"
+    r = extract([crate_a, crate_b])
+    node_ids_a = {n["id"] for n in r["nodes"] if "crate_a" in (n.get("source_file") or "")}
+    node_ids_b = {n["id"] for n in r["nodes"] if "crate_b" in (n.get("source_file") or "")}
+    # No calls edge should cross from crate_b into crate_a
+    cross_crate_calls = [
+        e for e in r["edges"]
+        if e["relation"] == "calls"
+        and e["source"] in node_ids_b
+        and e["target"] in node_ids_a
+    ]
+    assert cross_crate_calls == [], (
+        f"Spurious cross-crate edges: {cross_crate_calls}"
+    )
+
+
 # ── extract() dispatch ────────────────────────────────────────────────────────
 
 def test_extract_dispatches_all_languages():