Differentiate between 'url.host' and 'url.raw_host'

[tomchristie]

Refs #1275

Throughout our URL model we're differentiating neatly between byte-wise cases and str cases.
We're always using bytes when escaping is not applied, and str when escaping is applied.

Eg...

url = httpx.URL("https://jo%40email.com:a%20secret@example.com:1234/pa th")
assert url.username == "jo@email.com"
assert url.password == "a secret"
assert url.userinfo == b"jo%40email.com:a%20secret"
assert url.path == "/pa th"
assert url.raw_path == b"/pa%20th"

This pull request is a proposal for treating IDNA domain names similarly, so...

url = httpx.URL("https://müller.de:80")
assert url.host == "müller.de"
assert url.raw_host == b"xn--mller-kva.de"

For API consistency this also necessarily results in url.netloc becoming a byte interface, which actually makes sense for the contexts in which it is used.

url = httpx.URL("https://müller.de:80")
assert url.netloc == b"xn--mller-kva.de:80"

Finally we also introduce .raw_scheme for a byte-wise representation of the scheme, for a nice consistency so that:

url = httpx.URL("https://müller.de:80/pa th")
assert url.raw == (url.raw_scheme, url.raw_host, url.port, url.raw_path)
assert url.raw == (b"https", b"xn--mller-kva.de", 80, b"/pa%20th")

[StephenBrown2]

Wouldn't raw indicate unencoded?

[StephenBrown2]

Just a couple typos, and the previous question about what raw means.

[tomchristie]

Raw, as in the "the raw bytes on the wire", or "the raw ingredients that make up the cake".
The raw representation of the host is the actual unaltered bytewise representation that's used to make the connection.

Or, in baking...

The raw ingredients: \xf0\x9f\x8e\x82
The cake: 🎂

Similar usage of "Raw" in other technical docs.

[tomchristie]

Thanks so much for the review @StephenBrown2.
(Geez, me & my typos. 😬)

[StephenBrown2]

(Geez, me & my typos. 😬)

I think it was mainly just a copy-paste issue that got propagated, but no more normlized! :-p