replaceVarsWith: init

[wolfgangwalther]

Working my way towards removing substituteAll in #237216.

This takes the extended features of nix substituteAll to a replaceVars variant to get rid of those cases that use substituteAll to build a full package with meta information etc.

There are at least 15 more packages which will make use of replaceVarsWith, but I didn't include the changes here to keep the number of notifications lower while we iterate on the function itself. All those other cases that I currently know of use isExecutable = true.

Part of my overall effort for structuredAttrs in #205690.

@philiptaron @emilazy

Things done

• Built on platform(s)
  • x86_64-linux
  • aarch64-linux
  • x86_64-darwin
  • aarch64-darwin
• For non-Linux: Is sandboxing enabled in nix.conf? (See Nix manual)
  • sandbox = relaxed
  • sandbox = true
• Tested, as applicable:
  • NixOS test(s) (look inside nixos/tests)
  • and/or package tests
  • or, for functions and "core" functionality, tests in lib/tests or pkgs/test
  • made sure NixOS tests are linked to the relevant packages
• Tested compilation of all packages that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review rev HEAD". Note: all changes have to be committed, also see nixpkgs-review usage
• Tested basic functionality of all binary files (usually in ./result/bin/)
• 25.05 Release Notes (or backporting 24.11 and 25.05 Release notes)
  • (Package updates) Added a release notes entry if the change is major or breaking
  • (Module updates) Added a release notes entry if the change is significant
  • (Module addition) Added a release notes entry if adding a new NixOS module
• Fits CONTRIBUTING.md.

---

Add a 👍 reaction to pull requests you find important.

[philiptaron]

Welcome to the committers, @wolfgangwalther! Glad to have you here, and thanks as always for your high quality and well-thought-out PRs and collaboration.

[wolfgangwalther]

Looking through the current merge conflict I see #360576 - which would be an alternative way of dealing with those cases, without introducing replaceVarsWith.

Opinions, anyone?

[philiptaron]

I should be able to give this a thorough review this week. Thank you for your patience with the silence so far, @wolfgangwalther.

The interface I expected for replaceVarsWith was an adapted stdenv.mkDerivation, rather than a function of three arguments. I'm open to a function of three arguments, but I need to wrap my head around it.

The short version I had in mind was:

{ replaceVarsWith }:

path: attrs:

replaceVarsWith {
  src = path;
  replacements = attrs;
}

It's sort of fundamentally odd to me for a With variant to have anything other than one attrset as its input.

Regarding #360576, it's sort of a brute-force approach, and I don't like it as much as I like a general-purpose replaceVarsWith.

[wolfgangwalther]

The interface I expected for replaceVarsWith was an adapted stdenv.mkDerivation, rather than a function of three arguments.

That works for me, too.

I'm not bound to the current interface at all - I'm all ears for alternative suggestions.

[wolfgangwalther]

It's sort of fundamentally odd to me for a With variant to have anything other than one attrset as its input.

Hm. For me it's the opposite. For example, compare zipAttrsWith and zipAttrs. The latter is a specialized / partially applied variant of the former. We do the same here.

Now the question whether partial applications is any useful in this case... probably not.

I also found another example: cleanSource and cleanSourceWith. This seems to be much more like what you have in mind.

[wolfgangwalther]

The interface I expected for replaceVarsWith was an adapted stdenv.mkDerivation, rather than a function of three arguments. I'm open to a function of three arguments, but I need to wrap my head around it.

I tried the approach with one attrset as argument in a fixup commit. I like it more as well. It also makes the diff nicer to read.

Regarding #360576, it's sort of a brute-force approach, and I don't like it as much as I like a general-purpose replaceAttrsWith.

I put lsb-release back in and moved it to the new replaceVarsWith as well.

[philiptaron]

I like it a lot. One request for the copy-and-paste machine to come out, then I'd like to merge. This looks great in all the users.

[philiptaron]

If the PR is just the replaceVarsWith, is it able to target master?

[wolfgangwalther]

If the PR is just the replaceVarsWith, is it able to target master?

Nope, see #357395, which caused a lot of darwin rebuilds, too. Since replaceVars is also changed here, this will cause the same. Also #357395 has not made its way to master, yet, so there would be a few annoying merge conflicts coming up.

[philiptaron]

Looks great, tastes great. Want to use your new committer powers to push the button?