aio-libs / yarl / 36971974712

98%

push

github

Drop localhost from file URLs in WHATWG mode (#1961)

<!-- Thank you for your contribution! -->

## What do these changes do?

In WHATWG mode a `file` URL now follows the WHATWG file host state: the
`localhost` host (in any case, also once percent-decoded or IDNA-mapped)
becomes the empty host, and a `file` URL with userinfo or a port is
rejected. `file://localhost/test` is `file:///test`, `file://localhost`
is `file:///`, and `file://example:1/` fails. RFC 3986 mode is
unchanged.

Parsing, `URL.build()`, `join()`, `with_host()`, `with_scheme()`,
`with_port()`, `with_user()`, `with_password()` and a change of mode all
apply the same rule, so a built or joined URL matches the parsed one. A
Windows drive letter authority such as `file://C:/` is not treated as a
port; drive letters are left for a follow-up.

All 10 rows of the "file host that WHATWG rewrites" category move from
"WHATWG mode differs from WHATWG, RFC mode follows RFC 3986+3987" to
"yarl follows each standard in its mode" (264 -> 274):
`file://example:1/`, `file://localhost`, `file://localhost/`,
`file://localhost/test`, `file:\\localhost//`,
`file://localhost//a//../..//`, `\/localhost//pig` and
`//localhost//pig` against `file://lion/`,
`file://localhost//a//../..//foo` and `file://localhost////foo`. In
addition, `file://loC𝐀𝐋𝐇𝐨𝐬𝐭/usr/bin` ("non-ASCII or percent-encoded
host") moves from "Both modes differ from their standard" to "RFC mode
differs from RFC 3986+3987, WHATWG mode follows WHATWG", since its
WHATWG leg is now right. "yarl differs, RFC 3986+3987 and WHATWG agree"
stays at 0 rows.

## Are there changes in behavior for the user?

Yes. In the default WHATWG mode `URL("file://localhost/p")` is now
`URL("file:///p")`, and a `file` URL with userinfo or a port raises
`ValueError`. RFC mode and `encoded=True` parsing keep both as before.

## Is it a substantial burden for the maintainers to support this?

No. The rule lives in one check helper plus the existing host moving
helper (`_... (continued)

97 of 194 branches covered (50.0%)

Branch coverage included in aggregate %.

295 of 296 new or added lines in 2 files covered. (99.66%)

10095 of 10227 relevant lines covered (98.71%)

0.99 hits per line