Implement PEP 753: "Well-known Project URLs in Metadata"

[ichard26]

A few months ago, PEP 753 was accepted. It is now known as the "Well-known Project URLs in Metadata" standard. We should implement it as pip does consume project URL metadata.

I haven't looked at pip's existing Project-URL or Home-page handling, so it's possible that there is very little or nothing to do on pip's end, but I wanted to file this issue to make sure we take a look at some point. The logic that mirrors PyPI's old homepage Project-URL aliasing probably needs a some tweaking to have it conform to the specification.

pip/src/pip/_internal/commands/show.py

Lines 134 to 146 in f8f0f5a

	if not homepage:
	# It's common that there is a "homepage" Project-URL, but Home-page
	# remains unset (especially as PEP 621 doesn't surface the field).
	#
	# This logic was taken from PyPI's codebase.
	for url in project_urls:
	url_label, url = url.split(",", maxsplit=1)
	normalized_label = (
	url_label.casefold().replace("-", "").replace("_", "").strip()
	)
	if normalized_label == "homepage":
	homepage = url.strip()
	break

cc @woodruffw

[woodruffw]

Thanks for pinging me @ichard26!

it's possible that there is very little or nothing to do on pip's end

Indeed 🙂 -- pip shouldn't have to change anything, since PEP 753 stipulates only presentation behavior for indices and (very soft) changes to metadata production for build backends/metadata producers.

[ichard26]

Hmm, I was under the impression that there may be some work on pip's end because pip does consume and show project URLs. I included a link to the code in the issue description. Is that logic PEP compliant?

[woodruffw]

Ah yeah, my bad -- I forgot that we relaxed the language in the PEP to be any metadata consumer, not just indices.

The current logic looks like it needs to be updated (as does PyPI's -- I haven't done that yet). The PEP's label normalization logic is here: https://peps.python.org/pep-0753/#label-normalization. That should result in a strict superset of results as the current normalization logic.

[ichard26]

Ah, that seems very easy. I'll put this on my to-do list, but of course, no promises I'll get to it. And definitely not soon... Thanks for confirming my suspicions!

[woodruffw]

No problem, sorry for the false lead!

FWIW, this spurred me into doing the first bits on Warehouse as well: pypi/warehouse#17470