@codefusion-cc/base58

Ids people read, copy and paste: random ids and fixed-size byte values spelled in base58. Letters and digits only, without the look-alikes 0, O, I and l, so an id survives being read aloud, typed from a screen, double-clicked in a URL or a log, and pasted into a shell without quoting. No dependencies; runs in Workers, browsers and Node.

npm install @codefusion-cc/base58

A new id

import { randomId } from '@codefusion-cc/base58'

const deviceId = `d_${randomId(12)}` // d_5qCHTcgbQwpvYZQ9c: 96 random bits, 17 characters
const callId = randomId() // 128 random bits, 22 characters

The bytes come from crypto.getRandomValues. 12 bytes are plenty for ids that are not secrets.

Bytes an app already has

import { base58ToBytes, bytesToBase58 } from '@codefusion-cc/base58'

const segment = bytesToBase58(accountIdBytes) // always 22 characters for 16 bytes
const bytes = base58ToBytes(segment, 16) // null: not 16 bytes in base58

• The width depends only on the byte count (base58Length): 22 characters for 16 bytes, 44 for 32. A value with small leading bytes is padded with 1, the zero digit, so ids of one kind all look alike.
• The text sorts like the bytes (the alphabet is in ASCII order), so time-ordered bytes give time-ordered ids.
• A value that needs the full width spells as Bitcoin's base58 does. One with leading zero bytes does not: Bitcoin spends one 1 on each zero byte, this pads to the width.
• base58ToBytes reads only the exact width for the byte count, and never throws for what text holds (another length, a character outside the alphabet, a value too large for the bytes): it says null.
• Every function throws a RangeError for a byte count that is not one, which only a programming mistake passes.

Moving from base64url ids

Ids are opaque: rows that already hold base64url ids keep them, and new rows get base58 ones. Keep any route pattern or validation that accepted the old characters (A-Za-z0-9_- is a superset of base58), so old links keep working.