🌳 perf: make the library actually tree-shakable

Importing one decorator pulled in the message and validator of all 68 —
4,909 bytes instead of 1,837 through esbuild, 4,823 instead of 1,818
through webpack. Nothing failed and nothing warned. The library was simply
about three times heavier than it needed to be in every consumer's bundle.

Every rule is a top-level call: `export const IsString = rule(...)`. rollup
proves such a call side-effect-free by reading the factory, which is why it
was already emitting 1,823 bytes — and why a single-bundler measurement
would have shown no problem at all. esbuild and webpack do not do that
analysis and keep the call. Thirty declarations now carry /*#__PURE__*/,
which tsc preserves into the ESM emit, and all three bundlers now land
within 20 bytes of each other.

Measured, minified, esbuild / rollup / webpack:

  flattenErrors      287 /   292 /   291
  one decorator     1837 /  1823 /  1818
  validateSync      3722 /  3554 /  3823
  toPlainSync       7744 /  7769 /  7832
  toInstanceSync    7900 /  7942 /  7956
  a typical DTO    10395 / 10402 / 10372
  everything       26266 / 25671 / 26879

The serializer and deserializer drop independently — read JSON and you do
not pay for writing it. Both mapping entry points keep the validator,
because `validate` defaults to true and that is a real reference rather
than a missed optimisation.

The annotations are a promise to the bundler, so I checked the three
factories they cover: rule, pattern and affix each return a closure and
touch nothing outside themselves. A false promise here would mean silent
deletion in someone else's production build.

src/treeshake.test.ts pins the property. It asserts content rather than
only bytes — it names the rules that must not appear — and one case asserts
everything IS present when everything is used, so a "shaken" result cannot
come from a bundle that failed to build. That mattered: two earlier passes
at this measurement reported a clean sweep of shaken symbols because rollup
had failed to resolve its entry and grep was reading missing files as
absence. Strip the annotations and the test fails with
`"must be a latitude" should have been shaken out`.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SAcqrz3FcadkYr3xG32CjK
This commit is contained in:
Claude
2026-08-05 16:20:19 +00:00
parent b4f0657d09
commit 2a2f7345ad
5 changed files with 237 additions and 34 deletions
+28
View File
@@ -494,6 +494,34 @@ costs a few percent on `toPlain`, which is the price of never emitting `{}` wher
used to be; primitives are handled inline and the check is skipped for arrays and dates, so
it is one `Symbol.toStringTag` read per object.
## Bundle size
Cereale tree-shakes. Every rule is declared so that a bundler can drop the ones you did not
import, which matters for a library with 68 decorators — you pay for what you name and nothing
else. Minified bytes, measured through three bundlers, and
[pinned by a test](src/treeshake.test.ts) so it cannot quietly regress:
| What you import | esbuild | rollup | webpack |
| --- | ---: | ---: | ---: |
| `flattenErrors` | 287 | 292 | 291 |
| one decorator | 1,837 | 1,823 | 1,818 |
| `validateSync` | 3,722 | 3,554 | 3,823 |
| `toPlainSync` | 7,744 | 7,769 | 7,832 |
| `toInstanceSync` | 7,900 | 7,942 | 7,956 |
| a typical DTO — 5 decorators, map, validate, format errors | 10,395 | 10,402 | 10,372 |
| the whole library | 26,266 | 25,671 | 26,879 |
The serializer and the deserializer drop independently: read JSON and you do not pay for
writing it. The validator is kept by both, because `validate` defaults to `true` and the entry
points reference it whatever a given call site passes.
This did not come for free. Every rule is a top-level call — `export const IsString = rule(…)` —
and rollup can prove such a call side-effect-free by reading the factory, but esbuild and
webpack will not. Without a `/*#__PURE__*/` annotation on each of them, importing one decorator
pulled in the message and validator of all 68: **4,909 bytes instead of 1,837**. Nothing failed;
the library was simply three times heavier in every consumer's bundle, and the only way to find
out was to measure.
## Notes and Limitations
- **Rules are checked, types are not inferred.** You write both the field type and the rule;