fix(extraction): preserve symbols with colliding same-line IDs (#1349) (#2030)

Same-kind, same-name declarations on one line shared an ID and silently overwrote each other in SQLite.
Use per-extraction collision tracking to append a UTF-16 column only to later distinct declarations.
Apply the same rule across wasm, Rust language walkers, Liquid, and CFML while preserving file IDs and edge endpoints.
Add persistence, call-edge, identity-vector, and Unicode parity regressions.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

codegraph-kernel/src/ccpp/mod.rs

@@ -343,6 +343,7 @@ pub struct Walker<'t> {
     variant: Variant,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     nodes_meta: Vec<NodeMeta>,
@@ -388,6 +389,7 @@ pub fn extract(file_path: &str, source: &str, language: &str) -> Result<EmitOut,
         variant,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         nodes_meta: Vec::new(),
@@ -495,7 +497,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         // (c/cpp define no resolveBody hook, so createNode's endLine extension
         // for sibling-body grammars never fires — endLine is the node's own.)
         let end_line = node.end_position().row as u32 + 1;

codegraph-kernel/src/csharp.rs

@@ -130,6 +130,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     /// Node id string per row — ids COLLIDE for same-(kind, name, line) nodes
@@ -162,6 +163,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -317,7 +319,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1; // no resolveBody for csharp
 
         let qualified = {

codegraph-kernel/src/dart.rs

@@ -126,6 +126,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -156,6 +157,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -265,7 +267,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
 
         let qualified = {
             let mut parts: Vec<&str> = Vec::new();

codegraph-kernel/src/go.rs

@@ -88,6 +88,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     nodes_meta: Vec<NodeMeta>,
@@ -119,6 +120,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         nodes_meta: Vec::new(),
@@ -224,7 +226,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1;
 
         let qualified = extra.qualified_name.unwrap_or_else(|| {

codegraph-kernel/src/ids.rs

@@ -2,6 +2,8 @@
 //! `generateNodeId` in `src/extraction/tree-sitter-helpers.ts`:
 //!
 //!   `${kind}:${sha256(`${filePath}:${kind}:${name}:${line}`).hex[0..32]}`
+//! NodeIdAllocator appends `:<UTF-16 column>` only for later, distinct source
+//! positions sharing that legacy ID within an extraction.
 //!
 //! and the file-node special case in `TreeSitterExtractor.extract()`:
 //!
@@ -12,6 +14,33 @@
 //! in `__tests__/kernel-scaffold.test.ts`.
 
 use sha2::{Digest, Sha256};
+use std::collections::HashMap;
+
+/// Per-extraction collision handling, matching NodeIdAllocator in the wasm path.
+#[derive(Default)]
+pub struct NodeIdAllocator {
+    first_columns: HashMap<String, u32>,
+}
+
+impl NodeIdAllocator {
+    pub fn generate(
+        &mut self,
+        file_path: &str,
+        kind: &str,
+        name: &str,
+        line: u32,
+        column: u32,
+    ) -> String {
+        let id = node_id(file_path, kind, name, line);
+        // Zero-based UTF-16 column (util::col16), never tree-sitter's byte column.
+        let first_column = self.first_columns.entry(id.clone()).or_insert(column);
+        if *first_column == column {
+            id
+        } else {
+            format!("{id}:{column}")
+        }
+    }
+}
 
 pub fn node_id(file_path: &str, kind: &str, name: &str, line: u32) -> String {
     let mut hasher = Sha256::new();
@@ -41,6 +70,32 @@ pub fn file_node_id(file_path: &str) -> String {
 mod tests {
     use super::*;
 
+    #[test]
+    fn collision_only_identity_vectors() {
+        let mut ids = NodeIdAllocator::default();
+        let base = "function:bfb15544fed707794274a5c61006ea7b";
+        for (column, expected) in [
+            (2, base.to_string()),
+            (24, format!("{base}:24")),
+            (48, format!("{base}:48")),
+            (24, format!("{base}:24")),
+            (2, base.to_string()),
+        ] {
+            assert_eq!(
+                ids.generate("src/a.ts", "function", "foo", 3, column),
+                expected
+            );
+        }
+        assert_eq!(
+            ids.generate("src/b.ts", "function", "foo", 3, 24),
+            node_id("src/b.ts", "function", "foo", 3)
+        );
+        assert_eq!(
+            NodeIdAllocator::default().generate("src/a.ts", "function", "foo", 3, 24),
+            base
+        );
+    }
+
     #[test]
     fn matches_known_ts_output() {
         // Pinned vector: node -e "crypto.createHash('sha256')

codegraph-kernel/src/java.rs

@@ -124,6 +124,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     nodes_meta: Vec<NodeMeta>,
@@ -157,6 +158,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         nodes_meta: Vec::new(),
@@ -300,7 +302,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1; // no resolveBody for java
 
         let qualified = {

codegraph-kernel/src/kotlin.rs

@@ -191,6 +191,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -224,6 +225,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -383,7 +385,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         // endLine extension via resolveBody — LIVE for kotlin function/method
         // kinds (in-range for this grammar, so practically a no-op — but the
         // hook is part of the contract).

codegraph-kernel/src/lua.rs

@@ -75,6 +75,7 @@ pub struct Walker<'t> {
     is_luau: bool,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -103,6 +104,7 @@ pub fn extract(file_path: &str, source: &str, language: &str) -> Result<EmitOut,
         is_luau: language == "luau",
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -203,7 +205,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
 
         // buildQualifiedName (1447-1460) — non-file stack NAMES joined `::`;
         // the receiver override (extractMethod:1790-1792) replaces it whole.

codegraph-kernel/src/php.rs

@@ -142,6 +142,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -172,6 +173,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -316,7 +318,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1; // no resolveBody for php
 
         let qualified = {

codegraph-kernel/src/python.rs

@@ -141,6 +141,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -171,6 +172,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -276,7 +278,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1;
 
         let qualified = {

codegraph-kernel/src/rlang.rs

@@ -69,6 +69,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
 }
@@ -92,6 +93,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
     };
@@ -180,7 +182,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
 
         // buildQualifiedName (tree-sitter.ts:1447-1460): non-file stack names
         // joined `::` (namespacePrefix is always empty for R).

codegraph-kernel/src/ruby.rs

@@ -118,6 +118,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -148,6 +149,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -257,7 +259,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1; // no resolveBody for ruby
 
         let qualified = {

codegraph-kernel/src/rustlang.rs

@@ -112,6 +112,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     nodes_meta: Vec<NodeMeta>,
@@ -143,6 +144,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         nodes_meta: Vec::new(),
@@ -252,7 +254,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1;
 
         let qualified = extra.qualified_name.unwrap_or_else(|| {

codegraph-kernel/src/scala.rs

@@ -155,6 +155,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -187,6 +188,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -297,7 +299,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
 
         // buildQualifiedName (:1447-1460) — non-file stack names, `::`-joined;
         // namespacePrefix always empty (no C++ namespaces, no scala namespace).

codegraph-kernel/src/swift.rs

@@ -152,6 +152,7 @@ pub struct Walker<'t> {
     file_path: &'t str,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     node_ids: Vec<String>,
@@ -182,6 +183,7 @@ pub fn extract(file_path: &str, source: &str) -> Result<EmitOut, String> {
         file_path,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -332,7 +334,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
         let end_line = node.end_position().row as u32 + 1; // no resolveBody for swift
 
         let qualified = {

codegraph-kernel/src/tsjs/mod.rs

@@ -167,6 +167,7 @@ pub struct Walker<'t> {
     variant: Variant,
     line_starts: Vec<usize>,
     arena: Arena,
+    node_id_allocator: ids::NodeIdAllocator,
     tables: Tables,
     stack: Vec<Scope>,
     /// Node id string per row. Rows are unique but IDS COLLIDE for same
@@ -222,6 +223,7 @@ pub fn extract(file_path: &str, source: &str, language: &str) -> Result<EmitOut,
         variant,
         line_starts: util::line_starts(source),
         arena: Arena::default(),
+        node_id_allocator: ids::NodeIdAllocator::default(),
         tables: Tables::default(),
         stack: Vec::new(),
         node_ids: Vec::new(),
@@ -348,7 +350,8 @@ impl<'t> Walker<'t> {
             return None;
         }
         let start_line = self.line_of(node);
-        let id = ids::node_id(self.file_path, kind, name, start_line);
+        let column = self.col_of(node);
+        let id = self.node_id_allocator.generate(self.file_path, kind, name, start_line, column);
 
         // endLine body extension: resolveBody only (TS/JS: function-valued
         // class fields whose body nests in the arrow / HOF-wrapped arrow).