479 byte (minified) CSV parser and builder. Smaller when compressed. Built in ESM only.

Doesn't care about headers, keyed rows, anything but strings. Just supports the CSV spec including multi-line and quoted strings.

Does not support:

• \r\n (old DOS newlines)
• streaming

If you need those features, try tiddlycsv, which is a bit bigger. This package is intentionally tiny, mostly to prove a point.

Install via you favourite package manager and import but-csv. Has zero dependencies (obviously).

$ npm install but-csv

Parses a CSV into an array of array of strings. Supports varied line lengths. Does not convert to numbers or any other formats.

import { parse } from 'but-csv';

const out = parse('foo,bar,zing\n1,2,3\n4,5');
// out will be [['foo', 'bar', 'zing'], ['1', '2', '3'], ['4','5']]

Only supports passing a string (not a Buffer or friends). Node's operations on string are so much faster than on raw bytes (10x improvement). If you're parsing a file, do this:

const f = fs.readFileSync('source.csv', 'utf-8');
const out = parse(f);

Like parse, but you get each row at a time so you can stop early.

import { iter } from 'but-csv';

for (const row of iter('foo,bar,zing\n1,2,3\n4,5')) {
  // row will be an array of:
  // 1. ['foo', 'bar', 'zing'],
  // 2. ['1', '2', '3']
  // 3. ['4','5']
}

You can pass any value and it will be stringified before render, useful for numbers. This is unlike the parser above, which only returns strings.

import { build } from 'but-csv';

const out = build([
  ['hello', 'there"\n'],
  [1, 2],
]);

// out will be:
// hello,"there""
// "
// 1,2

Be sure to turn on your bundler's tree-shaking ability (good practice in general), but especially if you're only parsing or building, because the code is separate. Parsing is about 75% of the code, and building 25%.

It's very fast, but doesn't support streaming. To parse multiple copies of 1.csv from here, parsing all at once:

but-csv: 732.908ms
papaparse: 1.337s (1.8x)
csv-parser: 2.283s (3.1x)