Sending form data values bytestring incorrectly encodes it's without first decoding it as string as requests does

[nambrosini-codes]

Hello, first of all thank you for writing this lib and writing awesome migration docs comparing with requests. It was really helpful on our migration effort.

One part of our tests broke because for requests we sent form data this way

import requests
r = requests.post('https://mock.codes/200', data={'some_key', b'<Hello>'})

Internally, requests first decodes the bytestring with the default encoding and only then quotes the value so it can be included in the url.
httpx does not implicitly decode so instead of the expected string to start with %3C (<) we get b%27%3C (b'<). That is, both b and ' are quoted, which breaks the api call.

Once we manually passed a string, it worked, so the problem might perfectly be on our side. In that case i think it deserves a place on https://www.python-httpx.org/compatibility/. It was not obvious by reading it at least for us

So the question would be:
Is it an api design choice? Or are we misusing the api?

We noticed a comment on the implementation that assumes at that point the input of the function will always be a string but it might affect others using requests and wanting to migrate to httpx

Source