Custom URL types

[brunns]

I like to use yarl for storing and manipulating URLs. The requests library used to let me use these directly, but httpx2 does not:

import yarl, requests, httpx2
url = yarl.URL("https://brunn.ing")
r1 = requests.get(url)
r2 = httpx2.get(url)
Traceback (most recent call last):
  ...
TypeError: Invalid type for url.  Expected str or httpx2.URL, got <class 'yarl.URL'>: URL('https://brunn.ing')

I'm moving a largish codebase over from requests to httpx2, and having to wrap my URLs in str(url) in many places is painful, and I'm not sure it should be necessary.

(I used to use furl rather than yarl- the same issue exists with that.)

[Kludex]

Is your point that url=... should try to use __str__ from whatever object we get?

[brunns]

That would work, yes.

[Kludex]

We can try and see how a PR would look like, we should also check what would be the downside of doing this (I'm not sure?). I think I've faced this kind of situation in the past.

[brunns]

I think that the question is as to why httpx2's URL type explicitly forbids types other than str and httpx2.URL? That doesn't seem very Pythonic on the face of it - which makes me wonder if there's a specific reason for it? Chesterton's fence and all that.

I'd be happy to put together a PR to see what it looks like.

[Kludex]

"Explicit is better than implicit" seems a very pythonic behavior to me.

[brunns]

True, but dynamic typing is pretty core to Python. I've always considered isinstance() guards to be unpythonic, but maybe that's just me.

As we can see, requests managed to do without one.

[brunns]

#1302 created.