kibertoad / toad-cache / 28660607611

100%

push

github

Optimize cache implementations and improve hit statistics (#101)

* perf: eliminate redundant work on cache hot paths

- Replace has-then-get double hash lookups with a single lookup and
  undefined check in get/set/delete/expiresAt across all cache classes
  (safe: backing stores are prototype-less objects / Maps and stored
  item wrappers are never undefined)
- Preallocate the result array in getMany
- Reuse the existing Map in clear() instead of allocating a new one
- Precompute the archive deadline in HitStatistics so archiveIfNeeded,
  which runs on every tracked operation, is a single Date.now()
  comparison instead of Date arithmetic with three divisions
- Hoist the chained record lookup in the currentRecord getter
- Nest the empty-hit check inside the falsy-hit branch (empty values
  are a strict subset of falsy values)
- Drop the this.last guard before bumpLru, which already early-returns
  on the same condition

No public API or behavior changes; all 94 tests pass unchanged with
100% coverage. Local 200k set+get micro-benchmark: LruObjectHitStatistics
~2x faster, FifoMap ~25%, LruMap ~14%, Object variants ~10-15%.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013aDZMcXSvZghQbYwGKTaYH

* @
ci: drop Node 20 from matrix and bump pnpm to 11.9.0

pnpm 11.1.x+ requires Node >=22.13 (its store now requires the
node:sqlite builtin, absent in Node 20), so the 20.x leg crashed
during install with ERR_UNKNOWN_BUILTIN_MODULE. Run CI on 22/24/26,
update the coverage workflow off Node 20, bump the pnpm pin to the
latest 11.9.0, and align devEngines with the new toolchain floor.
@

* ci: run coverage workflow on Node 24

---------

Co-authored-by: Claude <noreply@anthropic.com>

216 of 216 branches covered (100.0%)

Branch coverage included in aggregate %.

65 of 65 new or added lines in 6 files covered. (100.0%)

350 of 350 relevant lines covered (100.0%)

22.96 hits per line