perf(lib): hoist customAlphabet factory in generateId

[kbpr21]

What & why

generateId calls customAlphabet(alphabet) on every invocation, which rebuilds
the nanoid generator (re-parsing the alphabet and recomputing its internal
mask/step) on each call. The factory only needs to be created once.

This hoists the customAlphabet(...) call to module scope so the generator is
built a single time and reused. Purely an optimization.

Behavior

No behavioral change. The alphabet and ID length (22) are identical, so generated
IDs keep the exact same format and distribution.

Before

​ts export const generateId = () => customAlphabet("...")(22) ​

After

​ts const generate = customAlphabet("...") export const generateId = () => generate(22) ​

Testing

bun run format-lint and bun run check-types pass for packages/lib
(pre-existing type errors in @supermemory/tools are unrelated to this change).

---

Session Details

• Session: View Session
• Requested by: Unknown
• Address comments on this PR. Add (aside) to your comment to have me ignore it.

[vorflux]

Summary

Reviewed the performance refactor in packages/lib/generate-id.ts, which hoists the customAlphabet generator used by generateId. The review subagent checked the diff against main for correctness, security, breaking changes, race conditions, and behavioral regressions.

Verdict

✅ Reviewed — no issues found. The change preserves the existing alphabet and 22 character length while avoiding repeated generator construction.

[vorflux]

Testing

The testing subagent classified this as an internal shared-library performance refactor with no UI, mobile, API, or integration testing required. It validated generateId runtime invariants, compared the old and new generator patterns, checked formatting/linting for the changed file, and ran a focused TypeScript check.

Commands run:

bun --eval "/* generated 10,000 IDs and checked length, alphabet membership, and duplicates */"
bun --eval "/* benchmarked old per-call customAlphabet(alphabet)(22) pattern vs new hoisted generator over 200,000 iterations */"
bunx biome check packages/lib/generate-id.ts
bunx tsc --noEmit --skipLibCheck --moduleResolution Bundler --module ESNext --target ES2022 --strict packages/lib/generate-id.ts
bun test packages/lib/generate-id.test.ts
bunx biome check packages/lib/generate-id.ts packages/lib/generate-id.test.ts

Result:

Runtime invariant check: PASS
- 10,000 generated IDs
- every ID length: 22
- alphabet: unchanged 58-character alphabet only
- duplicateCount: 0

Performance comparison: PASS
- old per-call pattern: 105.61ms
- new hoisted generator: 90.51ms
- observed speedup: ~1.17x

Changed-file Biome: PASS
- Checked 1 file in 126ms. No fixes applied.

Product-file TypeScript check: PASS

Focused local unit test validation: PASS
- 1 pass
- 0 fail
- 3 expect() calls

Biome on changed file plus local test file: PASS
- Checked 2 files in 10ms. No fixes applied.

No screenshots or recordings were produced because no UI, browser, or mobile testing was applicable.

Verdict

✅ Passed. The changed implementation preserved the expected ID shape and alphabet while showing the intended performance improvement in the testing subagent’s local benchmark.