openai-python, httpx, pydantic, tiktoken, openai-node — no CWE-407 defects.
All targets use proper dict/set/frozenset for membership tests in hot paths.
litellm-0001: ProxyConfig._delete_deployment() checks model_id not in
combined_id_list (a plain Python list) for each router model ID during
config refresh. O(R*C) where R=router models, C=combined list. Fix:
convert to set before loop. MEDIUM severity, 6.5x at N=500. 1/1 PASS.
Haystack: no CWE-407 defects found. Uses dicts and sets throughout
for membership checks (scheduled_components set, storage dict, etc.).
vllm-0001: punica_wrapper/utils.py convert_mapping() calls
lora_index_to_id.index(x) per token in batch — O(B*L) where
B=batch_size, L=loaded_loras. Code has "TODO index can be slow"
comment. Fix: pre-build dict for O(1) lookup. 7x at B=2000/L=64.
Ollama: all slices.Contains on bounded slices (1-8 items).
LangChain: orchestration code, all membership bounded by k param.
synapse/handlers/sync.py _get_rooms_changed: newly_joined_rooms is List[str]
checked with `in` inside for-loop over all joined rooms (line 2252).
O(J*N) where J=joined rooms, N=newly joined. Fix: use Set. 99x at J=5000.
Scanned three major emulator codebases for algorithmic complexity defects:
- MAME: proper maps/enumerators, std::find only on constant-bounded arrays
- Dolphin: exceptionally clean, HashMap/unordered_map throughout hot paths
- ScummVM: Common::find on tiny collections only, HashMap for config lookups
syncthing-0001: deviceDownloadState.blockIndexes []int with slices.Contains O(B)
called per device per block in file sync hot path. Fix: map[int]struct{}.
UNDF-2026-000000850. 2.9x measured at B=1000.
nextcloud-server: 386 in_array() calls, all on constant-size or config-level arrays.
gitea: 91 slices.Contains, all on small lists; container.Set used for hot paths.
freecad-0001: ifc_generator.py done-list O(N^2) dedup MEDIUM-HIGH 499x
freecad-0002: importDXF.py processededges list O(E^2) MEDIUM 469x
retroarch-0001: playlist_entry_exists O(R*P) linear scan in content scanner HIGH 749x
mpv: CLEAN (all data structures naturally bounded)
3/3 unit tests PASS.
clementine-0001: LibraryWatcher ScanSubdirectory FindSongByPath O(F*S) +
files_on_disk.contains O(S*F) — linear scan with TODO comment, fix with
HashMap + HashSet. HIGH severity, 250x at N=1000.
clementine-0002: SongSender indexOf(s) O(N^2) in SendAlbum/SendPlaylist/
SendUrls loops — fix with integer counter + QSet for requested_ids.
MEDIUM severity, 500x at N=1000.
MPD: CLEAN — uses std::set, bitmask arrays, std::map throughout.
Rhythmbox: CLEAN — uses g_hash_table for all membership checks.
4/4 unit tests PASS.
amarok-0001: TrackNavigator::queueIds QQueue.contains in loop MEDIUM 1000x
amarok-0002: QtGroupingProxy::mapFromSource QList.contains/indexOf in nested iteration MEDIUM 2426x
strawberry-0001: CollectionWatcher::ScanSubdirectory QStringList files_on_disk contains+removeAll HIGH 2525x
strawberry-0002: ScrobblerCache::Flush QList.contains/removeAll in loop MEDIUM 334x
tomcat-0001: WebSocket getNegotiatedSubprotocol List.contains O(R×S) MEDIUM 52x
hibernate-orm-0001: FK ordering buildRecursiveOrderedFkSecondPasses List.contains O(N²) HIGH 2.8x
netty-0001: ALPN select/selected List.contains O(S×P) MEDIUM 26x
netty-0002: DnsResolveContext dedup ArrayList.contains O(R²) MEDIUM 30x
All 4/4 unit tests PASS.
Beam: exhaustive scan of sdks/java/core + runners — all membership tests already
use proper Set types (HashSet, LinkedHashSet, ImmutableSet, TreeSet). CLEAN.
Hive hive-0001: SharedWorkOptimizer.mergeSchema() uses List.contains() in loops
for neededColumnIDs/neededColumns/virtualCols dedup. O(D*R) per list. MEDIUM, 3-4x.
Hive hive-0002: HiveRelMdSize.averageColumnSizes() uses ImmutableList.contains(i)
in column loop during Calcite metadata queries. O(C*L). MEDIUM, 3-5x.
Both patched with HashSet wrappers. 2/2 unit tests PASS.
