For example, I have a feeling the 693fc94 test case isn't exactly right. You're calling response.pause in the user's code, but that's not necessary. As a user you can do the following and get the expected output:

var Http = require("http")

Http.get("http://example.com", function(res) {
	setTimeout(function() {
		res.on("data", (data) => console.log("Data:", data.toString()))
	}, 500)
})

The hook into response's data event in the recording code is the thing that changes behavior — it immediately unpauses the stream. I suppose you could call res.pause in the recording code, but I have a suspicion that may then break automatic unpausing on data the user code does. —https://github.com/gr2m/http-recorder/issues/3#issuecomment-999501190

Follow up to https://github.com/moll/node-mitm/issues/76#issuecomment-998713363 thanks @moll

• [x] #4
• [x] #5
• [x] Does piping one stream into OutgoingMessage work, too? 👉 #2
• [x] #11
• [x] #9

See example at https://github.com/sindresorhus/got/blob/main/documentation/3-streams.md

import fs from "node:fs";
import fs from "node:fs";
import { promisify } from "node:util";

import HttpRecorder from "@gr2m/http-recorder";
import got from "got";

const pipeline = promisify(stream.pipeline);

HttpRecorder.enable()

// For POST, PUT, PATCH, and DELETE methods, `got.stream` returns a `stream.Writable`.
pipeline(
  fs.createReadStream("test.txt"),
  got.stream.post("http:/httpbin.org/post"),
  process.stdout
);

Breaking changes

• rename .enable() to .start() and .disable() to .stop() as it works better with a "recorder"
• prefer .addListener/.removeListener over .on/off for distinction from .start/.stop methods and web compat.

Did you test when calling write, end et al. with a chunk and a callback

—https://github.com/gr2m/http-recorder/issues/3#issuecomment-999501190

That's a minor thing, but given readable.setEncoding, the recorder is converting strings back to buffers presuming UTF-8(?).

My thinking here was to normalize the requestBodyChunks passed to the the record event listeners. I'm not worried about the performance overhead as this is a dev tool. Do you see any other problem? I will default encoding to utf-8 as part of #17

Related to missing output, I have a sense hooking into data immediately will cause some applications to miss parts of the content body. Those that don't start reading the response right after response and instead expect the stream to be paused for a while. Perhaps they're passing the response first to an async function, but by that time, the recording logic will have already unpaused the stream. —https://github.com/moll/node-mitm/issues/76#issuecomment-998713363

(the parent of ClientRequest) that according to https://github.com/nodejs/node/blob/113133badbe3c2f40f61eac733006207e06b9cc2/lib/_http_outgoing.js#L833 seems to call an internal method directly (not going through write). Wouldn't that miss output? —https://github.com/moll/node-mitm/issues/76#issuecomment-998713363

This PR contains the following updates:

| Package | Change | Age | Adoption | Passing | Confidence | |---|---|---|---|---|---| | @types/node (source) | ^17.0.2 -> ^18.0.0 | | | | |

Configuration

📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

• [ ] If you want to rebase/retry this PR, click this checkbox.

This PR has been generated by Mend Renovate. View repository job log here.

This PR contains the following updates:

| Package | Change | Age | Adoption | Passing | Confidence | |---|---|---|---|---|---| | tsd | ^0.24.0 -> ^0.25.0 | | | | |

Release Notes

Configuration

📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

• [ ] If you want to rebase/retry this PR, check this box

This PR has been generated by Mend Renovate. View repository job log here.

Certain http errors are not causing the "record" event to be emitted, making my http recorder logic not run. For example, if I call a url that doesn't exist (and I would imagine timeouts). I would like to be able to record events like this for testing purposes

I believe this is because you are hooking into the node "response" event, which is probably not fired for errors in which there is no response, like timeouts. What do you think?

I'm working on a library which records the timing of HTTP requests in a test suite to generate metrics around slow HTTP requests, your library works great for intercepting these HTTP requests, but I can't see any easy way to record how long each request is taking.

Here's a very short change which adds an extra attribute to the record event with the time taken for the request. I'm not particularlly tied to this implemention and so another option might be to add a "start" event, which would allow the timing of the requests outside of the library.

If you're happy with the change then i'm ok if you add changes directly to the PR or let me know any cahnges you'd like.