Polish test case for PR: replace into-stream to Readable.from

[stanleyxu2005]

Checklist

• run npm run test and npm run benchmark
• tests and/or benchmarks are included
• documentation is changed or added
• commit message and code follows the Developer's Certification of Origin
  and the Code of conduct

[stanleyxu2005]

I'd like to polish the test case @climba03003 and @mcollina create. But it's strage that your previous PR and this PR are failing at my local env.

The length of input and output strings are completly mismatched

[stanleyxu2005]

If I replace unidici.request with unidici.fetch, it works at my local env (but still partially failed with GitHub pipelines. Maybe the api of unidici is different for legacy Nodejs versions)

  const response1 = await fetch(`${address}/compress`)
  const response2 = await fetch(`${address}/no-compress`)
  const body1 = await response1.text()
  const body2 = await response2.text()
  t.equal(body1, body2)
  t.equal(body1.length, twoByteUnicodeContent.length)
  t.equal(body2.length, twoByteUnicodeContent.length)

Am I discovered an encoding issue for unidici? 😆

[gurgunday]

That's interesting

node 14 doesn't support undici fetch I think... you might want to try node-fetch@2 if you'd like to use it here

[climba03003]

I don't understand why it is strange on your environment. The test case works fine between different OS and Node versions.

I have also tested on my local environment.
What's the problem of undici.request ?

[gurgunday]

I don't understand why it is strange on your environment. The test case works fine between different OS and Node versions.

I have also tested on my local environment. What's the problem of undici.request ?

@climba03003 I was going to ask if you had an idea on why undici.request might be returning a different response text length, maybe there is some post processing based on the charset of the response?

So request didn't work: ffe60f6

And fetch did: c1882b9

[stanleyxu2005]

I don't understand why it is strange on your environment. The test case works fine between different OS and Node versions.

I have also tested on my local environment. What's the problem of undici.request ?

Here are my thoughts of failing or not failing @climba03003

In my attached img, the input length (334k) and output length (284k) are not identical.

(1) The previous PR you created, only checks the responses of compressed and uncompressed. They are identical on supported platforms (including my dev env1 - Win11), but for some reasons, not identical on my dev env2 - a legacy Windows.

(2) In my PR, I check the input and output length in additional. That's why it is failed.

(3) unidici seems to handle encoding differently, I cannot tell, if the output string is still the okay, from the length, I dont think so. After I have switched to node-fetch, the input and output length identical (which is more easier to understand). I'm not familiar with unidici, so I cannot say if unidici currput the data or not.

(4) The core implementation of your previous PR is proven functional

[gurgunday]

lgtm

[mcollina]

Please use use my test. undici.fetch() correctly decode this. undici.request() for some reason does not. I'll open an issue.

Use undici@5.

[gurgunday]

@mcollina undici fetch doesn’t work on node 14 I think, do you know if that’s the case?

[mcollina]

lgtm

[mcollina]

You are right, there is no web streams in node v14.