fast58

Base58 encode/decode with a published JavaScript package and a separate internal native benchmark lab.

fast58-js is a Solana-friendly Base58 tool for addresses, public keys, signatures, and other byte payloads. It is the spiritual successor to bs58, built with modern JavaScript and TypeScript packaging for even better speed.

Install

bun add fast58-js

npm install fast58-js

The npm package is JS-only. It builds from packages/fast58-js/src/index.ts and publishes only the allowlisted package files from packages/fast58-js/package.json.

Monorepo Layout

The main release target is the JavaScript package published as fast58-js. The Rust/N-API experiments and cross-implementation benchmarks live in a separate private workspace package.

• packages/fast58-js: published npm package
• packages/fast58-native: private Rust/N-API benchmark lab
• packages/fast58-native/implementations/js/*: named JS algorithm candidates
• packages/fast58-native/implementations/native.ts: native binding loader and candidate registry
• packages/fast58-native/bench/run.ts: shared benchmark runner
• packages/fast58-native/test/correctness.test.ts: cross-implementation correctness coverage
• packages/fast58-native/src/algorithms/*: Rust candidate implementations

Commands

bun test
bun run build:js
bun run bench
bun run build:native

Public API

import fast58 from "fast58-js";

const encoded = fast58.encode(new Uint8Array([104, 101, 108, 108, 111]));
const decoded = fast58.decode(encoded);

• encode: encode a Uint8Array as Base58
• decode: decode a Base58 string and throw on invalid input
• decodeUnsafe: decode a Base58 string and return undefined on invalid input

Benchmarking

The benchmark runner validates every implementation against bs58, measures encode and decode separately, and reports suite winners for:

• broad mixed payloads
• 32-byte hot paths
• 64-byte hot paths
• large payloads

That lets you add a new algorithm, register it once, and compare it against every existing candidate without rewriting the harness.

Benchmark lookup table (fast58 vs npm bs58)

Measured with bun run bench on April 8, 2026 (UTC). Values are exact throughput from the benchmark output.

Broad Mix (0B–256B)

Operation	fast58 winner	fast58 gmean ops/s	npm bs58 gmean ops/s	fast58 vs bs58
Encode	js/carry-direct-copy	2,611,041	2,332,248	1.12x faster
Decode	js/carry-direct-copy	4,132,217	1,771,628	2.33x faster

32-byte Hot Path

Operation	fast58 winner	fast58 gmean ops/s	npm bs58 gmean ops/s	fast58 vs bs58
Encode	js/carry-direct-copy	583,902	585,869	0.997x (near parity)
Decode	js/carry-direct-copy	2,054,119	650,263	3.16x faster

64-byte Hot Path

Operation	fast58 winner	fast58 gmean ops/s	npm bs58 gmean ops/s	fast58 vs bs58
Encode	js/carry-direct-copy	145,854	143,991	1.01x faster
Decode	js/carry-direct-copy	536,492	147,720	3.63x faster

Large Payloads (128B–1024B)

Operation	fast58 winner	fast58 gmean ops/s	npm bs58 gmean ops/s	fast58 vs bs58
Encode	js/carry-string-copy	4,248	4,190	1.01x faster
Decode	js/carry-direct-copy	16,079	3,929	4.09x faster

Broad Mix detail by payload size (fast58 js current vs npm bs58)

Size	Encode fast58	Encode bs58	Encode ratio	Decode fast58	Decode bs58	Decode ratio
0B	351,682,466/s	155,308,749/s	2.26x	57,595,264/s	29,440,919/s	1.96x
1B	62,464,613/s	54,353,920/s	1.15x	44,391,834/s	28,028,228/s	1.58x
2B	43,614,906/s	37,017,402/s	1.18x	38,659,282/s	24,298,538/s	1.59x
4B	20,565,540/s	18,814,911/s	1.09x	29,193,032/s	15,924,983/s	1.83x
8B	10,324,255/s	10,934,733/s	0.94x	15,932,578/s	10,978,727/s	1.45x
16B	2,594,950/s	2,635,188/s	0.98x	6,539,674/s	2,845,546/s	2.30x
32B	634,977/s	650,343/s	0.98x	2,184,630/s	734,017/s	2.98x
64B	136,165/s	140,237/s	0.97x	536,357/s	149,289/s	3.59x
128B	36,012/s	34,761/s	1.04x	127,827/s	33,640/s	3.80x
256B	8,960/s	8,865/s	1.01x	32,234/s	8,283/s	3.89x

For the generic arbitrary-length implementations, the current fixed winners are:

• JS: js/carry-direct-copy
• Rust encode: native/bs58-u32
• Rust decode: native/bs58-rs