But don't we have are sorted quasi-alphabetically to avoid merge conflicts. right above this? I do agree that alphanumeric the better word since there's attributes with numbers also.

I did a sort -f on all of them. I would like to keep it that way :)

Is there any practical reason behind this? I think that it's just painful to find the right position when adding a new attribute (in all-packages.nix, python-packages.nix or whatever) and any decent editor provides inline-search to jump to the proper position.

We should probably enforce some sort of pre commit hook to make sure it's sorted? Along with maybe a CI check.

Is there any practical reason behind this?

From the perspective of the nix-cli? no
From the perspective of a human? yes.
There's now just one location that "it should be located". Previously I would look at several locations to see which one was "most appropriate", now I can just type the package name until I find no results, and delete a character to find the package which contains the most similar prefix characters

We should probably enforce some sort of pre commit hook to make sure it's sorted?

This may be hard since we would have to "sort on paragraphs, while filtering out comments". It's likely to be do-able, but some of the expressions when I was initially sorting them were hard to handle.

I also thought of including this my nix-template cli tool jonringer/nix-template#5

There's now just one location that "it should be located". Previously I would look at several locations to see which one was "most appropriate", now I can just type the package name until I find no results, and delete a character to find the package which contains the most similar prefix characters

I asked about the sorting itself. The entire "where do I place my attribute"-thing wouldn't be necessary without it.

I asked about the sorting itself. The entire "where do I place my attribute"-thing wouldn't be necessary without it.

I'm not sure I understand, I sorted all of python-packages.nix in #96850. This PR is just making it "more official" that the expected behavior is have new additions also be sorted.

@Ma27 wonders if there is any value to having it sorted at all.

I think its mostly a matter of preference, at least for me it is. It has happened more than once a package was in the set twice, but under a slightly different name (capitalized or not). This you would spot sooner in case of a sorted list, but its rare it happens.

I think its mostly a matter of preference, at least for me it is. It has happened more than once a package was in the set twice, but under a slightly different name (capitalized or not). This you would spot sooner in case of a sorted list, but its rare it happens.

I have similar feelings. Making it easier to find packages helps with additions, modifications, and removals.

The "hard work" has already been done, I just want it to be more official.

The entire "where do I place my attribute"-thing wouldn't be necessary without it.

This was added under the "contributing guidelines" and I think it makes more sense if you look at the rendered document. https://github.com/NixOS/nixpkgs/blob/6bc12c461ff497f1c6bafeebca02033bea7b396b/doc/languages-frameworks/python.section.md#contributing-guidelines

I split the first bullet into two, and made the sorting word choice to be stronger.

We should probably enforce some sort of pre commit hook

Some hooks mess with rebasing broken commits so I personally never activated them.

Is there any practical reason behind this?

Less merge conflicts.

I believe there was a dd-agent and a datadog python package which would have been easier to spot if they were anywhere co-located.

Anyway, the ask now is, "please sort the already sorted package list". Where as before, it was, "please attempt to sort it.... it's not really sorted, but some parts are, and please just find something that is a close-enough match".

I haven't heard any strong objection to this, and it has been open for months now.