py-foo ports that actually install something?

René J.V. Bertin rjvbertin at gmail.com
Thu Sep 7 16:20:22 UTC 2023


A spin-off to trac:68137 and posted to the user list to give "user users" a chance to chime in and voice support:

What if those "py-foo" ports could be something other than stub ports? I'm pretty certain that the majority of python ports that do not install compiled binaries are (largely) version agnostic.

I'm imagining a set-up where the interpreter has 2 site-packages directories, a version-specific one that has priority and a version-agnostic, central version.

This should cut down significantly on the (growing) number of duplicate installed versions of those packages even if no better alternative is found to installing an additional, newer version of the interpreter at regular intervals. Using the prioritised approach would continue to allow pyXY-foo subports to exist and take precedence, where required. pyXY-foo subports that become redundant because of this could use `replaced_by py-foo`.

I'm aware that this would mean *not* installing the byte-compiled .pyc representations; those will get generated under the user's $HOME (like they probably are already anyway at least whenever the interpreter is updated). Not installing them under $prefix thus means saving some disk space for most users.


More information about the macports-users mailing list