activeloopai / activeloopai/hivemind
Perf: suffix-keyed reverse index for matchPythonSuffix (avoid O(files) scan per import)
- Dominant language
- TypeScript
- Stars
- 1.6k
- Forks
- 107
- Avg merge
- 17h 30m
- Merged PRs (30d)
- 6
Description
## Context
`matchPythonSuffix` in `src/graph/resolve/cross-file.ts` resolves a dotted-absolute Python import by suffix-matching against the known-files set. For each lookup it does up to 4 target forms × a full linear scan of `knownFiles`:
```ts
for (const t of targets) { // 4 target forms
if (knownFiles.has(t)) return t;
for (const f of knownFiles) { // O(files) scan
if (f.endsWith(`/${t}`)) { ... }
}
}
```
So cross-file resolution over a Python repo is `O(calls × files × 4)`. Fine at current scale (graphiti = 117 files, instant), but it degrades on large monorepos.
## Proposal
Build a suffix-keyed reverse index **once** (alongside `buildExportIndex`), mapping every path suffix (`mod.py`, `sub/mod.py`, `pkg/sub/mod.py`) → list of files, then make `matchPythonSuffix` an O(1) lookup + ambiguity check with no per-call scan.
## Notes
- Raised by CodeRabbit on PR #228 (Perf, Medium). Deferred from that PR since it's an optimization, not a correctness fix.
- Keep the existing high-confidence semantics: exact (root-anchored) wins; single suffix match wins; multiple → drop (ambiguous).
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.