This project uses the CompressionStream facility — supported by all recent browsers, Node and Deno — to create .zip files.
This is not entirely trivial since, a little frustratingly, CompressionStream can natively produce .gz format data but not .zip. Thus, we pick out both the deflated data and the CRC from the .gz stream, and write them into a .zip file instead.
We don't actually have to implement any compression, so the library is fast and small.
Where CompressionStream is not available, we fall back to producing uncompressed .zip files (and calculate the CRC in heavily-optimized JavaScript). This may be OK if you are creating a .zip that is actually an .xlsx, .apk or .xpi file, for example.
The library is currently suitable for small- and medium-sized files, since it briefly requires just over 2x the total uncompressed size of your files in memory. That's because you pass it an array of files data, and for the .zip output it allocates a Uint8Array backed by a worst-case ArrayBuffer, which is the size of all the uncompressed data plus a little more for headers.
Potential future improvements could include implementing a TransformStream instead, which could enable smaller memory use and larger file sizes. However, the .zip format annoyingly puts the CRC and compressed data size before the compressed data, which limits opportunities for memory saving.
npm install littlezipThe library exposes a single function, createZip.
import { createZip } from 'littlezip';
const zip = await createZip([
{ path: 'test.txt', data: 'This is a test', lastModified: new Date('2020-01-01T00:00:00') },
{ path: 'test.bin', data: new Uint8Array([1, 2, 3]) },
]);The first argument to createZip is an array of file entries. Each entry must have path (string) and data (string, Uint8Array or ArrayBuffer) keys, and may have a lastModified (Date) key, which otherwise defaults to the current date and time.
The optional second argument defines whether we attempt to deflate the data (default: true). If false, the resulting .zip file will be as large as the input data plus a few bytes for headers.