Stripping escapes

strip(text) returns text with its escape sequences removed. It is the named export strip of linegauge and linegauge/strip, and the default export of linegauge/strip, which strip-ansi's own suite grades (Compatibility). It is a subpath rather than the root's default because the root's default is width, the string-width drop-in.

import stripAnsi from 'linegauge/strip';

console.log(stripAnsi('\u001B[1m\u001B[31mbold red\u001B[39m\u001B[22m'));
console.log(stripAnsi('\u001B[38:2::255:0:0mcolon-form truecolour\u001B[39m'));
console.log(stripAnsi('\u001B]8;;https://example.com\u001B\\docs\u001B]8;;\u001B\\'));
console.log(stripAnsi('\u001B[2K\u001B[1Gredrawn'));
console.log(JSON.stringify(stripAnsi('a trailing escape\u001B')));

bold red
colon-form truecolour
docs
redrawn
"a trailing escape\u001b"

What it removes: SGR styling with semicolon or colon parameters, every other CSI command (cursor movement, erase), OSC commands including OSC 8 hyperlinks under either terminator, DCS strings, the charset selections and the single-character escapes. An ESC that nothing parses after — a trailing one, above — is text and is kept, which is also strip-ansi's answer.

One scannerCopy Anchor Link

width measures what strip leaves, and wrap and slice find escapes with the same scanner strip is built on. A string measured, wrapped, sliced and stripped therefore never disagrees with itself about where the escapes were.

Why not Node's stripVTControlCharacters?Copy Anchor Link

node:util has a stripper, and on most input it gives the same answer. Before Node 24.21 it stops at the first : and leaves the colon form of a truecolour code behind as text — :2::255:0:0mred — so a three-column word measured after it comes out fifteen. strip removes the colon form itself before handing the rest to Node, and strip.test.ts asserts the colon form is stripped, and measured at its real width, on every Node the package supports.