Fast byteLength() #333

What problem are you trying to solve?

new TextEncoder().encode(input).byteLength is an order of magnitude slower than alternatives, including Node's Buffer.byteLength(input) and even handwritten JavaScript implementations.

Benchmarks

./benchmarks/blob.js:              202’345.0 ops/sec (± 13’993.9, p=0.001, o=0/100)
./benchmarks/buffer.js:         57’434’701.2 ops/sec (±425’763.3, p=0.001, o=9/100) severe outliers=5
./benchmarks/implementation.js: 48’441’909.6 ops/sec (±397’249.6, p=0.001, o=5/100) severe outliers=2
./benchmarks/textencoder.js:     2’667’052.4 ops/sec (±564’727.5, p=0.001, o=6/100) severe outliers=2

My benchmark repo includes a JS implementation that I believe is at least close enough to correct for benchmarking purposes, although I'm no expert in UTF-16 so there may be some mistakes.

What solutions exist today?

• new Blob([input]).size
• new TextEncoder(input).byteLength
• Buffer.byteLength(input) (Node only)
• Implementations in JS

How would you solve it?

let encoder = new TextEncoder()
let byteLength = encoder.byteLength("Hello, World!")
// >> 13

Anything else?

No response