[Nix-dev] variable scope / recursion

[Mathijs Kwik]

On Fri, Jul 6, 2012 at 1:43 PM, Kirill Elagin <kirelagin at gmail.com> wrote:
> Yeah, now I get your problem.
> It think, it's clear from the source of `applyGlobalOverrides` and from
> comments, that your `packageOverrides` is called with _original,
> non-overriden_ pkgs. This way you can to refer to _original_ attributes when
> overriding them.

Yeah I saw those comments too, but I can't see how that would work.
Because the "original" (pkgsOrig) set is built by passing pkgs (the
final set) to pkgsFun.
In other words, the set references itself, which is perfectly normal
in a lazy setting.
The notion of "original" makes sense for the attributes that get
overwritten, but for the rest, I would expect a "shine through".

>
> I think the only option is to manually simulate overriding process in
> `packageOverrides` to obtain overriden pkgs.

That would mean a modification to nixpks, and maintaining a separate
fork forever, which is what I want to avoid.

>
>
> --
> Кирилл Елагин
>
>
> 2012/7/6 Mathijs Kwik <mathijs at bluescreen303.nl>
>>
>> Maybe I better explain what I'm trying to achieve.
>>
>> I have a bunch of packages and overrides that aren't very useful for
>> others, so I can't commit them to nixpkgs.
>> I don't want to constantly maintain my own "fork" that I need to keep
>> up-to-date, but rather use the current "channels" functionality.
>>
>> So I figured it's easiest to just use packageOverrides to apply the
>> modifications and to add the additional packages.
>>
>> This works fine for almost everything, but I want to structure my
>> additional packages more like nixpkgs itself (use directories and
>> callPackage). Ideally, I don't want to distinguish between nixpkgs and
>> my own pkgs, so the scope for callPackage needs to be the composition
>> of nixpkgs with my own stuff.
>>
>> But it seems I can't get a reference to the final nixpkgs attrset
>> within the overrides function...
>>
>> Any hints?
>>
>> Thanks,
>> Mathijs
>>
>>
>>
>> On Fri, Jul 6, 2012 at 12:10 PM, Mathijs Kwik <mathijs at bluescreen303.nl>
>> wrote:
>> > On Fri, Jul 6, 2012 at 12:01 PM, Kirill Elagin <kirelagin at gmail.com>
>> > wrote:
>> >> rec is the answer (this is something like let/letrec in Lisp, if you
>> >> know
>> >> what I mean ;) ).
>> >
>> > I don't see how rec would help here.
>> > As I understand it, rec helps when an attrset has attributes that
>> > refer to each other.
>> > But in this case, bar isn't referring to foo in the same object (the
>> > overrides object I'm creating, containing just foo and bar), but to
>> > pkgs.foo.
>> >
>> > Anyway, I tried it anyway:
>> >    packageOverrides = pkgs:
>> >      rec { foo = "string";
>> >        bar = pkgs.foo + " concatenation";
>> >      };
>> >
>> > But this does not help.
>> > Still "missing attribute foo".
>> >
>> >
>> >>
>> >> --
>> >> Кирилл Елагин
>> >>
>> >>
>> >> 2012/7/6 Mathijs Kwik <mathijs at bluescreen303.nl>
>> >>>
>> >>> Hi all,
>> >>>
>> >>> I was reading this interesting piece of all-packages.nix
>> >>> (comments/whitespace removed):
>> >>>
>> >>>   applyGlobalOverrides = overrider:
>> >>>     let
>> >>>       overrides = overrider pkgsOrig //
>> >>>         (lib.optionalAttrs (pkgsOrig.stdenv ? overrides && crossSystem
>> >>> == null) (pkgsOrig.stdenv.overrides pkgsOrig));
>> >>>       pkgsOrig = pkgsFun pkgs {};
>> >>>       pkgs = pkgsFun pkgs overrides;
>> >>>     in pkgs;
>> >>>
>> >>> I'm familiar with tying-the-knot lazy trickery in haskell.
>> >>> As pkgsOrig uses pkgs (the final end result including overrides), and
>> >>> the overrider function gets passed pkgsOrig, I would expect the
>> >>> following to work:
>> >>>
>> >>>   packageOverrides = pkgs:
>> >>>     { foo = "string";
>> >>>       bar = pkgs.foo + " concatenation";
>> >>>     };
>> >>>
>> >>> However, it complains about foo missing.
>> >>> What am I missing?
>> >>>
>> >>> Thanks,
>> >>> Mathijs
>> >>> _______________________________________________
>> >>> nix-dev mailing list
>> >>> nix-dev at lists.science.uu.nl
>> >>> http://lists.science.uu.nl/mailman/listinfo/nix-dev
>> >>
>> >>
>> _______________________________________________
>> nix-dev mailing list
>> nix-dev at lists.science.uu.nl
>> http://lists.science.uu.nl/mailman/listinfo/nix-dev
>
>