RequestError: unexpected end of file

[mvolfik]

Describe the bug

• Node.js version: 18.15.0
• OS & version: Linux Mint 21.1 Cinnamon 64-bit

Actual behavior

node:internal/process/esm_loader:97
    internalBinding('errors').triggerUncaughtException(
                              ^

RequestError: unexpected end of file
    at PassThrough.<anonymous> (file:///home/volfmatej/temp/got-repro/node_modules/got/dist/source/core/index.js:599:31)
    at Object.onceWrapper (node:events:628:26)
    at PassThrough.emit (node:events:525:35)
    at emitErrorNT (node:internal/streams/destroy:151:8)
    at emitErrorCloseNT (node:internal/streams/destroy:116:3)
    at process.processTicksAndRejections (node:internal/process/task_queues:82:21)
    at BrotliDecoder.zlibOnError [as onerror] (node:zlib:189:17)

Expected behavior

Request succeeds and body is printed.

Code to reproduce

import {got} from "got"; // got 12.6.0

const resp = await got("https://www.yelp.com/search/snippet?find_desc=Hotels&request_origin=user&l=g%3A-2.2323167%2C53.476563899999995%2C-2.2373167%2C53.4715639");
console.log(resp.body);

Checklist

• [ ] I have read the documentation. I'll be honest, I didn't, but I'm strongly convinced this should work, given it works for www.example.com
• [x] I have tried my code with the latest version of Node.js and Got.

[mvolfik]

Hm, after more investigation it seems like the root of the issue is a corrupted (or some other issue with it) brotli compression. Both Firefox and Chromium on my machine handle it correctly in full, curl --compressed returns only part of the response, and got failed as I report above. Attached is a zip archive (because of github file limitations) which contains the damaged brotli response, you can run a test server for example with deno (but you can spin up basically any simple http server and add headers as below)

deno run -A server.js. server.js contents:

const data = await Deno.readFile("issue.brotli");
console.log(data);
async function handleHttp(conn) {
  for await (const e of Deno.serveHttp(conn)) {
    e.respondWith(
      new Response(data, {
        headers: {
          "content-encoding": "br",
          "content-type": "application/json",
        },
      })
    );
  }
}

for await (const conn of Deno.listen({ port: 8000 })) {
  handleHttp(conn);
}

Attached: issue.zip

I'll leave it up to you whether this is something you want to fix/improve (I suppose it would be in the decompress-stream package, is it possible to easily move this issue there?), or close as wontfix

[mvolfik]

I got an email notification about an issue comment from @WillianAgostini who noted that this works with axios. The comment is now deleted, not sure why, because it really does work with axios.

For convenience, I host the broken file at ftp.mvolfik.com/broken.brotli.json. It serves the broken brotli version with content-encoding: br regardless of request headers.

[WillianAgostini]

Yes, I've deleted it because I was looking for why it works on axios.

Here is the implementation where is created a stream to download all data, so on the end, all data is concatenated and converted to json. https://github.com/axios/axios/blob/6f360a2531d8d70363fd9becef6a45a323f170e2/lib/adapters/http.js#L506-L558

[WillianAgostini]

Based on implementation in axios and the docs https://github.com/sindresorhus/got/blob/main/documentation/3-streams.md, here is a minimal example to resolve this issue.

import got from "got";

const url = "https://www.yelp.com/search/snippet?find_desc=Hotels&request_origin=user&l=g%3A-2.2323167%2C53.476563899999995%2C-2.2373167%2C53.4715639";

const stream = got.stream(url, {
    decompress: false,
});

const chunks = [];
stream.on('data', (chunk) => {
    chunks.push(chunk);
});

stream.once('end', () => {
    const buffer = Buffer.concat(chunks);
    console.log(JSON.parse(buffer.toString()));
});

[mvolfik]

Well the decompress just disables sending of the compression headers, then you don't need to do anything fancy with streams:

import got from "got";

const url = "https://www.yelp.com/search/snippet?find_desc=Hotels&request_origin=user&l=g%3A-2.2323167%2C53.476563899999995%2C-2.2373167%2C53.4715639";

console.log(await got(url, {
    decompress: false,
}));

I am aware of this workaround, I just wanted to point out the issue with the broken brotli stream, but as I said in my second comment, I'll leave it up to you whether this is something you want to fix, or close as wontfix because site serves invalid data.