Choosing columns when parsing

The columns option names, selects, and renames the columns a parse emits, checked against your record type. It is the parsing side of columns on the encoder.

Name headerless rows by position

An array of keys turns each string[] row into an object, by position. It works with header: false:

parse('1,2\n3,4', { header: false, columns: ['x', 'y'] });
// [{ x: '1', y: '2' }, { x: '3', y: '4' }]

With header: true the CSV header row is dropped in favor of your keys, so you control the names whatever the file calls them:

parse('name,age\nAlex Johnson,29', { columns: ['n', 'a'] });
// [{ n: 'Alex Johnson', a: '29' }]

Rename and select with a label map

A map of CSV header label to your key renames and selects columns. Labels not in the map are dropped, so you can normalize a messy header into exactly your shape. A rename map requires header: true.

parse('Full Name,Email,extra\nAlex Johnson,alex@example.com,x', {
  columns: { 'Full Name': 'name', Email: 'email' }
});
// [{ name: 'Alex Johnson', email: 'alex@example.com' }]  ("extra" is dropped)

The keys you map to are checked against T in parse<T> and createCsvParser<T>, so a typo will not compile.