Performance - target es2018

[tmcw]

So the current setup is that zod has a compile target of es5, which applies to both its ES Module and CJS distributions. Zod's internals use quite a lot of modern features. TypeScript is polyfilling all of those features in the distributed code. TypeScript's polyfills are very good and probably as fast as they can be, but they're not as fast as the native code. For example, every array or object spread and every generator gets compiled to something with a bit more indirection.

That indirection looks like it's costing a real slowdown for some Zod functions.

So, bumping the target to ES2018 would cost IE11 support if it was done uniformly. Assuming that IE11 support is still critical (that's for the maintainers to decide, not me), a possibility might be targeting es2018 in the module entrypoint and keeping es5 in the main.

Okay, so before:

~/p/zod on es2018 [$!]
130 % y benchmark
yarn run v1.22.4
$ ts-node src/benchmarks/index.ts
z.string: empty string x 2,451,532 ops/sec ±0.87% (94 runs sampled)
z.string: short string x 2,482,775 ops/sec ±0.45% (95 runs sampled)
z.string: long string x 2,404,051 ops/sec ±1.38% (91 runs sampled)
z.string: invalid: null x 89,502 ops/sec ±0.97% (89 runs sampled)
z.string: manual parser: long x 874,138,478 ops/sec ±1.56% (87 runs sampled)
z.object: empty: valid x 1,646,418 ops/sec ±5.48% (91 runs sampled)
z.object: empty: valid: extra keys x 1,559,284 ops/sec ±0.80% (91 runs sampled)
z.object: empty: invalid: null x 90,326 ops/sec ±0.92% (95 runs sampled)
z.object: short: valid x 790,794 ops/sec ±0.94% (94 runs sampled)
z.object: short: valid: extra keys x 763,664 ops/sec ±6.08% (90 runs sampled)
z.object: short: invalid: null x 90,299 ops/sec ±0.63% (93 runs sampled)
z.object: long: valid x 412,393 ops/sec ±0.45% (97 runs sampled)
z.object: long: valid: extra keys x 391,355 ops/sec ±5.61% (92 runs sampled)
z.object: long: invalid: null x 88,410 ops/sec ±0.95% (93 runs sampled)
z.union: double: valid: a x 256,918 ops/sec ±0.60% (96 runs sampled)
z.union: double: valid: b x 256,784 ops/sec ±0.39% (96 runs sampled)
z.union: double: invalid: null x 28,961 ops/sec ±6.36% (84 runs sampled)
z.union: double: invalid: wrong shape x 24,945 ops/sec ±6.41% (87 runs sampled)
z.union: many: valid: a x 126,783 ops/sec ±0.55% (93 runs sampled)
z.union: many: valid: c x 124,881 ops/sec ±1.09% (92 runs sampled)
z.union: many: invalid: null x 16,535 ops/sec ±5.17% (86 runs sampled)
z.union: many: invalid: wrong shape x 14,627 ops/sec ±6.94% (86 runs sampled)
z.discriminatedUnion: double: valid: a x 667,578 ops/sec ±0.44% (96 runs sampled)
z.discriminatedUnion: double: valid: b x 663,500 ops/sec ±0.54% (96 runs sampled)
z.discriminatedUnion: double: invalid: null x 88,595 ops/sec ±0.85% (95 runs sampled)
z.discriminatedUnion: double: invalid: wrong shape x 84,477 ops/sec ±0.62% (94 runs sampled)
z.discriminatedUnion: many: valid: a x 660,361 ops/sec ±0.39% (92 runs sampled)
z.discriminatedUnion: many: valid: c x 665,620 ops/sec ±0.84% (93 runs sampled)
z.discriminatedUnion: many: invalid: null x 91,685 ops/sec ±0.70% (92 runs sampled)
z.discriminatedUnion: many: invalid: wrong shape x 83,539 ops/sec ±0.67% (90 runs sampled)
✨  Done in 164.98s.

After:

% y benchmark
yarn run v1.22.4
$ ts-node src/benchmarks/index.ts
z.string: empty string x 7,456,311 ops/sec ±0.50% (94 runs sampled)
z.string: short string x 7,294,686 ops/sec ±0.63% (94 runs sampled)
z.string: long string x 7,267,118 ops/sec ±0.74% (92 runs sampled)
z.string: invalid: null x 68,995 ops/sec ±3.45% (90 runs sampled)
z.string: manual parser: long x 873,026,560 ops/sec ±1.81% (86 runs sampled)
z.object: empty: valid x 3,947,137 ops/sec ±0.54% (95 runs sampled)
z.object: empty: valid: extra keys x 3,381,157 ops/sec ±0.55% (93 runs sampled)
z.object: empty: invalid: null x 68,712 ops/sec ±0.78% (95 runs sampled)
z.object: short: valid x 1,699,620 ops/sec ±30.55% (93 runs sampled)
z.object: short: valid: extra keys x 1,940,140 ops/sec ±0.82% (94 runs sampled)
z.object: short: invalid: null x 69,124 ops/sec ±0.47% (94 runs sampled)
z.object: long: valid x 899,629 ops/sec ±0.53% (94 runs sampled)
z.object: long: valid: extra keys x 839,770 ops/sec ±0.77% (95 runs sampled)
z.object: long: invalid: null x 67,109 ops/sec ±3.32% (92 runs sampled)
z.union: double: valid: a x 177,326 ops/sec ±6.18% (90 runs sampled)
z.union: double: valid: b x 175,326 ops/sec ±5.59% (86 runs sampled)
z.union: double: invalid: null x 23,836 ops/sec ±3.96% (89 runs sampled)
z.union: double: invalid: wrong shape x 23,462 ops/sec ±0.86% (92 runs sampled)
z.union: many: valid: a x 72,148 ops/sec ±6.87% (86 runs sampled)
z.union: many: valid: c x 75,365 ops/sec ±4.36% (92 runs sampled)
z.union: many: invalid: null x 14,744 ops/sec ±3.83% (89 runs sampled)
z.union: many: invalid: wrong shape x 14,380 ops/sec ±0.62% (93 runs sampled)
z.discriminatedUnion: double: valid: a x 730,682 ops/sec ±0.42% (96 runs sampled)
z.discriminatedUnion: double: valid: b x 723,743 ops/sec ±0.39% (95 runs sampled)
z.discriminatedUnion: double: invalid: null x 61,431 ops/sec ±3.81% (90 runs sampled)
z.discriminatedUnion: double: invalid: wrong shape x 69,464 ops/sec ±0.82% (93 runs sampled)
z.discriminatedUnion: many: valid: a x 730,148 ops/sec ±0.74% (96 runs sampled)
z.discriminatedUnion: many: valid: c x 732,561 ops/sec ±0.55% (94 runs sampled)
z.discriminatedUnion: many: invalid: null x 65,482 ops/sec ±0.64% (94 runs sampled)
z.discriminatedUnion: many: invalid: wrong shape x 69,293 ops/sec ±0.47% (93 runs sampled)
✨  Done in 165.41s.

[colinhacks]

I ran a diff and basically the only difference is that es2018 no longer transpiles the spread operator to Object.assigns under the hood. Guess the spread can be faster in some cases...kinda surprising.

IE11 support isn't a priority and a 2x speed improvement is hard to turn down. Merging.

[syabro]

So no ie11 anymore?

[FlorianWendelborn]

@syabro you can either use an older version, or transpile it down to es5. Not a big deal