Remove user credentials in URLs when converting to a string

[zanderxyz]

Summary

As previously noted in this GitHub discussion, this library by default leaks credentials which are included in URL strings (common for basic authentication). It can also raise exceptions which contain the credentials in the error string if a request fails (see raise_for_status).

This PR updates the __str__ method on URLs to remove the user & password details. I believe this is the correct default behaviour for a library like this, as it avoids any risk of leakage. Removing the user & password entirely seems both (a) the safest option, and (b) the simplest implementation. These credentials are passed as headers in reality, and are not technically part of the URL.

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.
• I've updated the documentation accordingly. [I did not make any documentation changes as I do not think any are needed]

[zanderxyz]

This seems simpler, rather than re-implementing the logic already in _urlparse.py. The user credentials are sent in a header rather than the URL anyway.

[karpetrosyan]

Yep, hiding user credentials at the lowest layer and preventing them from being passed higher makes perfect sense.

Now it's exposed in URL.str, but not in URL.repr, which is a bit weird. str is supposed to show more user-related data, while repr is more for debugging and development. So, hiding it in repr would make more sense. However, I think preventing it entirely is the most secure way.

[zanderxyz]

Yep, hiding user credentials at the lowest layer and preventing them from being passed higher makes perfect sense.

Now it's exposed in URL.str, but not in URL.repr, which is a bit weird. str is supposed to show more user-related data, while repr is more for debugging and development. So, hiding it in repr would make more sense. However, I think preventing it entirely is the most secure way.

Just to clarify - after the change in this PR, the username & password are not exposed in either str or **repr✱. I do think this is the most secure implementation, and I don't think we really lose anything..

[grahamwhiteuk]

Anyone have a view on when this might get merged (and released)?

[grahamwhiteuk]

Seems to be taking an age to be reviewed/merge so in the meantime, if you find this and don't want to log secrets you could just alter the log level of httpx...

import logging

httpx_logger = logging.getLogger("httpx")
httpx_logger.setLevel(logging.WARNING)

[lovelydinosaur]

This PR updates the str method on URLs to remove the user & password details.

Thank you no.