Support for the Nix expression language (Nix / NixOS)

[blueyed]

Description: https://nixos.org/nix/manual/#ch-expression-language

There is some very basic config in https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/tools/misc/ctags/wrapped.nix:

--langdef=NIX
--langmap=NIX:.nix
--regex-NIX=/([^ \t*]*)[ \t]*=/\1/f/

I have (hopefully) improved this a bit locally:

--regex-NIX=/([^ \t*]*)[ \t]*=.*:/\1/f/

But it's still too basic and would need a separate parser probably.

I am new to Nix/NixOS myself, and have therefore not much knowledge
about this language myself.

[vcunat]

Nix itself parses the language - it uses standard parsing and lexing tools https://github.com/NixOS/nix/tree/master/src/libexpr (probably most useful in context of this issue). Then I know about some stuff in haskell https://github.com/peti/language-nix, and there are syntax highlighters for various editors (or highlighting libraries).

[masatake]

As an initial version, a regular expression based implementation is o.k; it is always better than nothing.
Though I have to rethink the directory structures, optlib is the first class way to implement an parser.
You can write a C based parser later. The important point is the "kinds" compatibility between regex
based parser and C based parser. If the compatibility of kinds are kept, no one takes care the way of implementation.

However, if nix implementation has real parser, why don't you utilize it? My xmcd is for you.
See data/optlib/cofee.ctags and libexec/drivers/coffeetags.

[roosemberth]

Hi, what's the status of this issue?

[masatake]

A pull request is welcome.

[lf-]

If anyone wants a Nix ctags implementation, I wrote one in my tool here: https://github.com/lf-/nix-doc

It's Rust so not possible to integrate here though.

[bsima]

fwiw this has been working for me, just drop this into ~/.ctags.d/nix.ctags

--langdef=nix
--languages=+nix
--langmap=nix:.nix
# packages are "name = <tag>" or "pname = <tag>"
--kinddef-nix=p,package,package definition
--regex-nix=/[p?]name\s*=\s*"(\w+)"/\1/p/
# functions have args, so look for a : right of the =
--kinddef-nix=f,function,function definition
--regex-nix=/(\S+)\s*=\s+\w+:/\1/f/
# attrs just have =, but only index if they have >=4 chars
--kinddef-nix=a,attr,attribute definition
--regex-nix=/\s+([a-zA-Z_0-9-]{4,20})\s*=/\1/a/

[masatake]

Why don't you make a pull request to integrate it into ctags?

[bsima]

Well I'm not very good at C, but I could give it the old college try

[masatake]

NO. You don't have to use C.
Universal Ctags has a quack translator named optlib2c that converts your .ctags to C.
So, you make your parser written in .ctags a built-in parser. See https://github.com/universal-ctags/ctags/tree/master/optlib . Many parsers are written in the way.

See 599cc35 as an example. Though it includes .c file but it is translated from .ctags from optlib2c.
A perl interpreter is needed on your platform.

Let me know if you want to integrate your .ctags into Universal Ctags.
I will explain the details.

[bsima]

Okay, I will try and follow the PkgConfig example tonight

[bsima]

I got it started here #4020

Will continue work on it tomorrow. But I think I'm moving in the right direction.