This is misleading. Because nixpkgs repackages all of npm, pip, hackage etc. packages. Stuff that on other distros are managed by their respective tools seperate from the distro packaging. Still nixpkgs is pretty vast.
Most Haskell packages are not included in the AUR and haskellPackages alone makes up over 17,000 packages on nixpkgs compared to ~1,100 on AUR - subtract just that and Arch pulls ahead. Nixpkgs also has almost twice as many Python packages because python310Packages is separate from python311Packages.
I bet there are lots of other non-language examples too: a quick search found emacsPackages which contains six thousand packages.
Showing that there are smaller scoped packages does not prove your point.
You have to prove that it has less of the software available on other distros. Even if you could find a billion small packages that represents only a single package in every distro this is not evidence or a proof that support your claim.
Simply put, you need to show where nix does not have a larger number of packages that are available on other distros.
Given that nix is mostly automated and pulls from the same root sources as every other distro you are going to be hard pressed to find a significant number of cases where nix lacks a package.
The original poster claimed nixpkgs has the largest and most uptodate package collection just because of the sheer number of nix packages. That's a wrong conclusion and thus misleading. If you want to claim that you need to do exactly what you are asking. You need to make a direct comparison what software is available and what is not. From my experience nixpkgs is better then ubuntu but worse then Arch.
The metric you are looking for is the non-unique one on repology which counts only packages that are available on multiple distros. This is in fact the category where nixpkgs has the largest lead.
I think nixpkgs is probably so much larger for two reasons:
- running unpackaged software on nixos is painful.
- it’s very easy to contribute to nixpkgs (standard github pr to a monorepo).
Those includes all those packages which aren't packaged on other distros due to them being expected to be available via their native tooling. Examples are the entirety of pythonPackages, haskellPackages, emacsPackags. vscodePackages etc. Which I mentioned before in this very thread.
It makes little sense if you compare it like this. It would be stupid for any other package repo to include those package since they already exist. The only reason nixpkgs needs to do it is for purity reasons.
The metric that matters is "how many packages can I easily install on this distro". And nixpkgs most certainly is not in the lead here.
A package is non-unique if it is available in at least 2 repositories. Guess what. Repology tracks Guix as well which means many packages will be just in those two repos, marking them as non-unique.
If you want a non-misleading comparison you have to remove all these packages, which are actually managed by a different package manager every. from nixpkgs and put that into repology.
You can remove the entirety of the guix package set (26468) from the non-unique part of nixpkgs (71875) and it would still be the largest non-unique package set (second largest non-unique is AUR at 37900).
reply