Reuse readable streams while serving range requests.
Typically range requests are made for a data directly available on the file system. However, for the cases where the data at rest is encrypted and not easily indexable, serving range request still requires decrypting file from the beginning. This can become a bottleneck for situations where a video has to be served to a browser, since for a 100MB video Chromium would typically make hundreds of requests, turning a 100MB video into tens of gigabytes of read and decrypted data.
range-finder is is an aid for these (perhaps niche) cases. Instead of creating
a brand new stream for every request it stores and tracks previously created
streams that were no fully read, and attempts to reuse them when requesting
subsequent data. For the example of Chromium video loader above, using this
module reduces the created streams from ~200 to ~3.
npm install @indutny/range-finderimport { RangeFinder, DefaultStorage } from '@indutny/range-finder';
import { createReadStream } from 'node:fs';
const storage = new DefaultStorage(() => createReadStream('/tmp/1.txt'), {
maxSize: 100,
});
const finder = new RangeFinder(storage);
const startOffset = 123;
finder.get(startOffset).pipe(process.stdout);Because streams are reused between several requests, if one of the requests gets stalled it might stall all other requests that are based on the same input stream. Depending on the use case, either more granular contexts could be used or a timeout could be added to all derived streams.
This software is licensed under the MIT License.