How to distinguish request errors from Writable errors in undici.stream

[zeroberry]

With undici.stream, errors can come from either the request itself or the Writable returned by the factory. Right now they surface the same way, and I haven’t found a way to tell them apart, but I need this distinction to implement things like metrics, fallback handling, and selective retries.

With undici.request this feels easier, since it exposes the response readable and I can attach listeners directly (I also avoid pipeline, since it propagates errors across streams).

Question

Is there a way to distinguish between:

1. Request-level errors
2. Writable errors

Example

import { Writable } from 'node:stream'
import { stream } from 'undici'

async function main() {
  try {
    await stream('https://www.google.com/', { method: 'GET' }, () => {
      return new Writable({
        write(_chunk, _enc, cb) {
          cb(new Error('Writable failed'))
        },
      })
    })
  } catch (err) {
    // ❓ Is there a way to tell if this came from the request vs. the Writable?
    console.error('Caught error:', err)
  }
}

main()

[metcoder95]

You can just create custom errors and verify the error thrown; nonetheless, the undici.Stream has been evaluated as possibly deprecation, I'd advice to use undici.Response.body readable instead to handle the response and pipeline it or send it to another target.

[zeroberry]

Thanks for the clarification!
Is there a reason undici.stream is being considered for deprecation?

[metcoder95]

Mostly the complexity, and with undici.Response.body you can usually achieve the same as with the API (not to mention the interceptors can also provide similar capabilities with better details and ergonomics).