Prefer urlsplit() over urlparse()

[Adam Johnson]

Many places in Django use [urlparse()](​https://docs.python.org/3/library/urllib.parse.html#urllib.parse.urlparse), which supports the rarely-used “path parameter” syntax (not to be confused with query parameters). The urlsplit() function is similar but does not parse such path parameters, which makes it a bit faster.

I think most or all calls to urlparse() can be replaced with urlsplit(), and similarly urlunparse() with urlunsplit(). This may make a small but measurable performance difference in common paths, such as in CsrfViewMiddleware or the test Client.

See more in this Anthony Sottile video: ​https://www.youtube.com/watch?v=ABJvdsIANds , where he reports a 3% import time improvement on the Stripe project.

[Sarah Boyce]

Accepting for someone to make updates and confirm with benchmarks. 👍

[Jake Howard]

From some basic testing, it looks like this should have a nice improvement:

In [1]: import urllib.parse

In [2]: %timeit urllib.parse.urlparse("https://example.com")
1.52 µs ± 37.9 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

In [3]: %timeit urllib.parse.urlsplit("https://example.com")
258 ns ± 1.14 ns per loop (mean ± std. dev. of 7 runs, 1,000,000 loops each)

The difference stays about constant, even with longer URLs. The specifics will obviously vary by hardware and Python version (the above is 3.12.3), but ~6x improvement is definitely worthwhile.

[Jake Howard]

​PR.

[Jake Howard]

Comments addressed / replied-to in PR.

[Mariusz Felisiak]

Fixed by ff308a06047cd60806d604a7cf612e5656ee2ac9