Skip to content

Commit 70b8ba0

Browse files
committed
docs: freshness — README test count matched to the actual suite (~75, was ~90), bench ✦ marks framed as a one-machine snapshot (close lanes flip: Node has measured ahead on the primitive-set row elsewhere), nodejs/node#64433 upgraded from filed to found-and-fixed upstream in STORY + README; PR-CAMPAIGN.md untracked — a public targeting playbook can only sour the PRs it plans, while the compat-audit evidence stays
1 parent 9d43c21 commit 70b8ba0

2 files changed

Lines changed: 11 additions & 5 deletions

File tree

‎README.md‎

Lines changed: 7 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -117,6 +117,10 @@ Five-way numbers (Node 24, Apple Silicon; run `npm run bench` yourself):
117117
| POJO tree, ~1k nodes | 249µs | 200µs | 689µs | 355µs | 539µs |
118118
| tiny object, ×500k | 0.16µs | — | — | 0.19µs | 0.71µs |
119119

120+
The ✦ marks are a snapshot of one machine and Node build, and the close lanes
121+
genuinely flip — other machines have measured Node ahead on the primitive-set
122+
row — so rerun the bench where the margin matters.
123+
120124
Fastest on flat data and primitive sets; within ~1.5× of Node on object sets
121125
and POJO trees — and the gap **is** the feature: full sharing registration is
122126
what the incumbents skip, and skipping it is exactly why they answer `true`
@@ -169,7 +173,7 @@ UNORDERED // === Symbol.for('unordered-collection')
169173

170174
## Verification
171175

172-
Beyond ~90 unit cases (including adversarial batteries for forged
176+
Beyond ~75 unit cases (including adversarial batteries for forged
173177
constructors, lying subclasses, detached ArrayBuffers, and prototype-only
174178
fakes), the suite cross-validates against a **brute-force isomorphism
175179
oracle** — an exhaustive bijection search transcribed directly from the
@@ -182,7 +186,8 @@ with proofs of the soundness architecture and the complexity story, is in
182186
## The story
183187

184188
How this library came to be — the GI-completeness realization, the algorithm,
185-
and the bugs the fuzzer found across the ecosystem (including Node core):
189+
and the bugs the fuzzer found across the ecosystem (including a since-fixed
190+
crash in Node core):
186191
[STORY.md](./STORY.md).
187192

188193
## License

‎STORY.md‎

Lines changed: 4 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -154,9 +154,10 @@ random data, ask every library, and flag whoever disagrees with the
154154
brute-force answer. This got out of hand in the best way.
155155

156156
The fuzzer found a **crash in Node core**: `assert.deepStrictEqual`,
157-
`assert.notDeepStrictEqual`, and `util.isDeepStrictEqual` all throw a raw
158-
`TypeError` on two-entry Maps with `null` keys — filed as
159-
[nodejs/node#64433](https://github.com/nodejs/node/issues/64433). It found
157+
`assert.notDeepStrictEqual`, and `util.isDeepStrictEqual` all threw a raw
158+
`TypeError` on two-entry Maps with `null` keys — reported as
159+
[nodejs/node#64433](https://github.com/nodejs/node/issues/64433) and since
160+
fixed upstream. It found
160161
a previously unreported **false-negative class in dequal**
161162
([#41](https://github.com/lukeed/dequal/issues/41)): equal Maps whose keys
162163
are structurally identical get rejected by greedy pairing —

0 commit comments

Comments
 (0)