Nix pretty-printer/canonicalizer

[copumpkin]

If we can read in the AST and spit it back out again in a standard (ideally somewhat pretty) format, that'd do a reasonable job at this.

Ideally it would preserve comments and such.

[fkz]

nix-instantiate --parse is doing exactly that (not really pretty printing though and also not preserving comments and canonicalizing paths to absolute paths)

[copumpkin]

How much of a pain would it be to make it preserve comments? It seems like most ASTs other than these auto-formatters will typically discard comments fairly early on, and might also discard whitespace people intended to keep.

For example, the jsonnet canonicalizer will preserve single and double blank lines (under the assumption that the human author added them for a reason), but coalesce 3 or more blank lines into two.

[fkz]

Currently, comments are discarded already in the lexer. I'm currently thinking about changing that for an other purpose (being able to do more reflection inside nix to be able to generate documentation for functions)

[copumpkin]

That sounds awesome for several reasons :+1:

[fkz]

Ah, this is also currently only idempotent if done twice:

nix-instantiate --parse --expr '"1${"2" + "3"}4"'
>("1" + ("2" + "3") + "4")
nix-instantiate --parse --expr '("1" + ("2" + "3") + "4")'
>(("1" + ("2" + "3")) + "4")
nix-instantiate --parse --expr '(("1" + ("2" + "3")) + "4")'
>(("1" + ("2" + "3")) + "4")

[domenkozar]

https://github.com/Gabriel439/nixfmt is incomplete, but does exactly this. See https://github.com/Gabriel439/nixfmt/issues/3 for the current status

[fstamour]

Would be nice to integrate nixfmt with nix-mode (emacs mode) and nixIdea (intellij's idea plugin).

[vcunat]

Related: #1102.

[etu]

Would be nice to integrate nixfmt with nix-mode (emacs mode) and nixIdea (intellij's idea plugin).

What about aiming for the LSP method here? I think there's some works on nix LSP integrations somewhere. That would give benefits for other editors as well.

[domenkozar]

https://github.com/domenkozar/hnix-lsp works, but doesn't support comments (yet)

[ghuntley]

This now exists https://github.com/justinwoo/format-nix/ but it's performance is quite slow..

$ time ./result/bin/format-nix ../nix/packages.nix formatted ../nix/packages.nix.

real 0m6.232s user 0m6.458s sys 0m0.261s

Against https://github.com/digital-asset/daml/blob/master/nix/packages.nix

[domenkozar]

$ \time cat nix/packages.nix | ../result/bin/canonix --pipe >/dev/null
0.00user 0.00system 0:00.00elapsed 100%CPU (0avgtext+0avgdata 3540maxresident)k
0inputs+0outputs (0major+165minor)pagefaults 0swaps

using a bit modified version of https://github.com/hercules-ci/canonix/ that is not really usable yet - might be after ZuriHac.

[domenkozar]

We have https://github.com/nix-community/nixpkgs-fmt now, the main question is if it becomes part of Nix some day.

[stale[bot]]

I marked this as stale due to inactivity. → More info

[stale[bot]]

I closed this issue due to inactivity. → More info

[quatquatt]

As there's been much progress on nixfmt being fully adopted by nixpkgs, I think it's important to note an issue I'm having when it comes to simply piping nix eval into nixfmt.

nix eval prints «derivation long/nix/store/path/here» whenever it encounters a package. nixfmt has no clue what to do with this, so it errors.

To reproduce, run nix eval /etc/nixos#nixosConfigurations.MYHOSTNAMEHERE.config.environment.systemPackages | nixfmt.

While I think the eventuall pretty-printing tool should be based on nixfmt, it's not currently able to be used without issue.

[Atemu]

Perhaps nixfmt should learn to deal with such strings inside of «» and accept them as pseudo-literals for the purpose of processing Nix output.