perf(diffctx): ego_graph O(k*BFS)→O(BFS) + file_cache to universe

Fix ego_graph: hoist to_undirected outside seed loop, replace
per-node shortest_path_length with single_source_shortest_path_length.
Reduces complexity from O(seeds*ego_size*(V+E)) to O(seeds*(V+E)).

Propagate file_cache to _build_ident_index and
_expand_universe_by_rare_identifiers in universe.py.
Add _read_file helper to EdgeBuilder base class for cache-aware
file reading across all 40+ edge builders.

Author: nikolay-e

src/treemapper/diffctx/edges/base.py

@@ -185,6 +185,17 @@ def discover_related_files(
     ) -> list[Path]:
         return []
 
+    @staticmethod
+    def _read_file(path: Path, file_cache: dict[Path, str] | None = None) -> str | None:
+        if file_cache is not None:
+            content = file_cache.get(path)
+            if content is not None:
+                return content
+        try:
+            return path.read_text(encoding="utf-8")
+        except (OSError, UnicodeDecodeError):
+            return None
+
     def add_edge(
         self,
         edges: EdgeDict,

src/treemapper/diffctx/graph.py

@@ -80,13 +80,14 @@ def nx(self) -> nx.DiGraph:
 
     def ego_graph(self, seeds: set[FragmentId], radius: int = 2) -> dict[FragmentId, float]:
         scores: dict[FragmentId, float] = {}
+        if not self._g:
+            return scores
+        undirected = self._g.to_undirected(as_view=True)
         for seed in seeds:
             if seed not in self._g:
                 continue
-            undirected = self._g.to_undirected(as_view=True)
-            ego = nx.ego_graph(undirected, seed, radius=radius)
-            for node in ego.nodes:
-                dist = nx.shortest_path_length(undirected, seed, node)
+            distances = nx.single_source_shortest_path_length(undirected, seed, cutoff=radius)
+            for node, dist in distances.items():
                 hop_score = 1.0 / (1 + dist) if dist > 0 else 1.0
                 scores[node] = max(scores.get(node, 0.0), hop_score)
         return scores

src/treemapper/diffctx/scoring.py

@@ -49,6 +49,7 @@ def discover(self, ctx: DiscoveryContext) -> list[Path]:
             ctx.changed_files + edge_discovered,
             combined_spec,
             candidate_files=ctx.all_candidate_files,
+            file_cache=ctx.file_cache,
         )
 
         return list(dict.fromkeys(edge_discovered + expanded))

src/treemapper/diffctx/universe.py

@@ -132,17 +132,20 @@ def _collect_candidate_files(root_dir: Path, included_set: set[Path], combined_s
 def _build_ident_index(
     files: list[Path],
     concepts: frozenset[str],
+    file_cache: dict[Path, str] | None = None,
 ) -> dict[str, list[Path]]:
     inverted_index: dict[str, list[Path]] = defaultdict(list)
     for file_path in files:
-        try:
-            content = file_path.read_text(encoding="utf-8")
-            file_idents = extract_identifiers(content, skip_stopwords=False)
-            for ident in file_idents:
-                if ident in concepts:
-                    inverted_index[ident].append(file_path)
-        except (OSError, UnicodeDecodeError):
-            continue
+        content = file_cache.get(file_path) if file_cache else None
+        if content is None:
+            try:
+                content = file_path.read_text(encoding="utf-8")
+            except (OSError, UnicodeDecodeError):
+                continue
+        file_idents = extract_identifiers(content, skip_stopwords=False)
+        for ident in file_idents:
+            if ident in concepts:
+                inverted_index[ident].append(file_path)
     return inverted_index
 
 
@@ -219,6 +222,7 @@ def _expand_universe_by_rare_identifiers(
     already_included: list[Path],
     combined_spec: pathspec.PathSpec,
     candidate_files: list[Path] | None = None,
+    file_cache: dict[Path, str] | None = None,
 ) -> list[Path]:
     if not concepts:
         return []
@@ -228,16 +232,19 @@ def _expand_universe_by_rare_identifiers(
         files = [f for f in candidate_files if f not in included_set]
     else:
         files = _collect_candidate_files(root_dir, included_set, combined_spec)
-    inverted_index = _build_ident_index(files, concepts)
+    inverted_index = _build_ident_index(files, concepts, file_cache=file_cache)
 
     included_concept_counts: dict[str, int] = {}
     for f in already_included:
-        try:
-            content = f.read_text(encoding="utf-8")
-            for ident in extract_identifiers(content, skip_stopwords=False):
-                if ident in concepts:
-                    included_concept_counts[ident] = included_concept_counts.get(ident, 0) + 1
-        except (OSError, UnicodeDecodeError):
+        content = file_cache.get(f) if file_cache else None
+        if content is None:
+            try:
+                content = f.read_text(encoding="utf-8")
+            except (OSError, UnicodeDecodeError):
+                continue
+        for ident in extract_identifiers(content, skip_stopwords=False):
+            if ident in concepts:
+                included_concept_counts[ident] = included_concept_counts.get(ident, 0) + 1
             continue
 
     return _collect_expansion_files(inverted_index, concepts, included_set, included_concept_counts)

whitelist_vulture.py

@@ -1,4 +1,5 @@
 from treemapper.clipboard import clipboard_available
+from treemapper.diffctx.edges.base import EdgeBuilder
 from treemapper.diffctx.edges.semantic import (
     AnsibleEdgeBuilder,
     BazelEdgeBuilder,
@@ -66,4 +67,5 @@
 blast_radius = graph_analytics.blast_radius
 ScoringMode.AUTO
 PipelineConfig.low_relevance
+EdgeBuilder._read_file
 EgoGraphScoring