httpx.URL() does not parse url correctly with curly brackets in password.

[xlash]

I was having an error when trying to use httpx with a proxy configuration. After digging, I notice that the proxy url parsing is broken when certain characters are in the password, like a opening curly bracket in this example #2.

Known characters posing issues : [,],{,},@,/
The @ sign seems pretty obvious, but not necessarily the other ones. I can't find the specific RFC preventing the use of brackets.

httpx 0.23.0

To reproduce :

import httpx
# Test case #1
httpx.URL('HTTPS://jo%40email.com:a%20secret@müller.de:1234/pa%20th?search=ab#anchorlink')
# Test case #2
httpx.URL('HTTPS://jo%40email.com:a%20sec{ret@müller.de:1234/pa%20th?search=ab#anchorlink')
# Test case #3
httpx.URL('HTTPS://jo%40email.com:a%20sec{ret@müller.de:1234')

Outputs :

# Test case #1
URL('https://jo%40email.com:[secure]@xn--mller-kva.de:1234/pa%20th?search=ab#anchorlink')
# Test case #2
URL('HTTPS:/pa%20th?search=ab#anchorlink')
# Test case #3
URL('HTTPS:')

Expected output :

Probably raise a parsing issue ? Like if FQDN is invalid like in the Test case #2, or if it's empty like #3

Possible solution:

Encode all special characters in url with % sign before calling URL()
HTTPTransport() line 142

Workaround

You can use url encoding for username and password
username = urllib.parse.quote(username, safe='')
password = urllib.parse.quote(password, safe='')

Or build the proxy variable like this
httpx.Proxy(httpx.URL('HTTPS://müller.de:1234/pa%20th?search=ab#anchorlink', username='jo%40email.com', password='sec{ret')))