⚡ Bolt: 검색 성능 최적화

.jules/bolt.md

@@ -67,3 +67,6 @@
 ## 2025-02-12 - [Fast Path Execution in Directory Traversal and Log Parsing]
 **Learning:** Checking for string existence (`if "silence_" not in stderr`) before invoking regex matchers provides significant speed improvements when parsing large blocks of text. Similarly, moving expensive I/O operations like `os.path.realpath` inside conditional blocks prevents redundant disk access when configuration (like path exclusions) isn't utilized.
 **Action:** When working on large text processing or disk operations, verify if early exit conditions or conditional execution can bypass the expensive system or library calls.
+## 2024-10-08 - [Optimize Search Set Intersection and Loops]
+**Learning:** In tight loops involving `set`s (like when computing the intersection of matching documents for multi-word search terms), creating a new `set` with the `&` operator allocates memory and iterates needlessly. Modifying a target `set` in place using `.intersection_update()` avoids this allocation. In the same vein, using `sum()` with a generator expression introduces significant overhead due to object allocation of the generator. Replacing it with a simple `for` loop that accumulates in a local variable provides a substantial speedup.
+**Action:** Always prefer `intersection_update` over `&` when working with sets you can modify safely. For performance-sensitive loops, replace `sum(a for b in c)` with a basic inline `for` loop to eliminate the generator overhead.

transcript_search.py

@@ -241,16 +241,19 @@ def search(self, query: str) -> list[Match]:
             postings = self._postings.get(term)
             if not postings:
                 return []
-            candidates = (
-                set(postings) if candidates is None else candidates & postings
-            )
+            if candidates is None:
+                candidates = set(postings)
+            else:
+                candidates.intersection_update(postings)
             if not candidates:
                 return []
 
         matches = []
         for position in candidates or ():
             entry = self._entries[position]
-            score = sum(entry.counts[term] for term in unique_terms)
+            score = 0
+            for term in unique_terms:
+                score += entry.counts[term]
             matches.append(
                 Match(
                     recording_id=entry.recording_id,