Issue regarding usage of namespace aspects with ._

[amanako]

I recently discovered and started implementing new ._ synthetic attribute across my config,
but when trying to use it with namespace aspects(freeform children) it fails. Is this to
be expected, and if so is it possible to implement ._ for namespace aspects as well? AFAIK
aspects pass through same pipeline transformations and something is off.

Using ._ synthetic attribute with den's direct aspects evaluates successfuly:

{denTest, ...}: {
  flake.tests.bogus = {
    test-something = denTest (
      {
        inputs,
        den,
        lib,
        igloo,
        tuxHm,
        ...
      }: {
        den.hosts.x86_64-linux.igloo.users.tux = {};

        # do something for testing
        den.aspects.foo.bar = {
          nixos.programs.localsend.enable = true;
        };

        den.aspects.foo.barbar = {
          nixos.system.stateVersion = "25.11";
        };

        den.aspects.igloo.includes = [den.aspects.foo._];

        expr = {
          localsend = igloo.programs.localsend.enable;
          stateVersion = igloo.system.stateVersion;
        };

        expected = {
          localsend = true;
          stateVersion = "25.11";
        };
      }
    );
  };
}

nix-unit-check> ✅ system-agnostic.bogus.test-something
nix-unit-check> 🎉 1/1 successful
nix-unit-check> Writing "check derived from mv7zl7xb3q9clawrw0lrpclmhci6lq93-nix-unit-check.drv is ok
nix-unit-check> " to /nix/store/3rwfm6b6a9njlifraxfly3b4jl6z5jj4-nix-unit-check

Hoewer when trying with direct namespaces (either direct or provides):

{denTest, ...}: {
  flake.tests.bogus = {
    test-something = denTest (
      {
        inputs,
        # Add namespace to scope
        foo,
        den,
        lib,
        igloo,
        tuxHm,
        ...
      }: {
        # Add namespace
        imports = [(inputs.den.namespace "foo" false)];

        den.hosts.x86_64-linux.igloo.users.tux = {};

        # do something for testing - direct children of foo namespace
        foo.bar = {
          nixos.programs.localsend.enable = true;
        };

        foo.barbar = {
          nixos.system.stateVersion = "25.11";
        };

        # Try to include all direct children of "foo"
        den.aspects.igloo.includes = [foo._];

        expr = {
          localsend = igloo.programs.localsend.enable;
          stateVersion = igloo.system.stateVersion;
        };

        expected = {
          localsend = true;
          stateVersion = "25.11";
        };
      }
    );
  };
}

produces:

error: builder for '/nix/store/vr04z5w6gs3abv6yjbjhj52pan2rs22n-nix-unit-check.drv' failed with exit code 1;
       last 25 log lines:
       >
       >        … while evaluating the attribute 'value'
       >          at /build/storedata/nix/store/m9k87k3yi7z8mmrnvnyajqc28i0mqgg2-source/lib/types.nix:763:25:
       >           762|                         inherit (def) file;
       >           763|                         value = def';
       >              |                         ^
       >           764|                       }
       >
       >        … while calling the 'elemAt' builtin
       >          at /build/storedata/nix/store/m9k87k3yi7z8mmrnvnyajqc28i0mqgg2-source/lib/lists.nix:347:43:
       >           346|   */
       >           347|   imap1 = f: list: genList (n: f (n + 1) (elemAt list n)) (length list);
       >              |                                           ^
       >           348|
       >
       >        error: attribute '_' missing
       >        at /build/storedata/nix/store/4zr3ffw6w2px7f29ip80bpzfpkb88vd0-source/modules/bug.nix:27:39:
       >            26|
       >            27|         den.aspects.igloo.includes = [foo._];
       >              |                                       ^
       >            28|
       >
       >
       > 😢 0/1 successful
       > error: Tests failed
       For full logs, run:
          nix log /nix/store/vr04z5w6gs3abv6yjbjhj52pan2rs22n-nix-unit-check.drv

[sini]

It's possible this is already fixed in #589 -- any chance you could test against the branch and let me know? If not I'll try and reproduce when I have a moment.

[amanako]

I tested it with the branch and error is identical.

[sini]

Thanks for checking. I'll try and get a fix for this stacked into 589

[sini]

merged

[amanako]

Tested it again and it passes. Keep up the good work!

[amanako]

Upon further inspection, seems that subaspects with the same name are problematic.
This example:

{denTest, ...}: {
  flake.tests.bogus = {
    test-something = denTest (
      {
        inputs,
        # Two namespaces
        foo,
        bar,
        igloo,
        ...
      }: {
        imports = [
          (inputs.den.namespace "foo" false)
          (inputs.den.namespace "bar" false)
        ];

        den.hosts.x86_64-linux.igloo.users.tux = {};

        # Use same subaspect
        foo.barbar = {
          zig.nixos.programs.localsend.enable = true;
          zig2.nixos.system.stateVersion = "26.05";
        };

        bar.barbar = {
          fish.nixos.programs.fish.enable = true;
          zsh.nixos.programs.zsh.enable = true;
        };

        # Include order matters
        den.aspects.igloo.includes = [
          bar.barbar._
          foo.barbar._
        ];

        expr = {
          localsend = igloo.programs.localsend.enable;
          stateVersion = igloo.system.stateVersion;
          fish = igloo.programs.fish.enable;
          zsh = igloo.programs.zsh.enable;
        };

        expected = {
          localsend = true;
          stateVersion = "26.05";
          fish = true;
          zsh = true;
        };
      }
    );
  };
}

causes:

       > ❌ system-agnostic.bogus.test-something
       > /build/nix-8-479361931/expected.nix --- Nix
       > 1 { fish = true; localsend = _false_; sta  1 { fish = true; localsend = true; stat
       > . teVersion = "25.11"; zsh = true; }     . eVersion = "26.05"; zsh = true; }

and switching order of includes emits:

       > ❌ system-agnostic.bogus.test-something
       > /build/nix-8-1451852217/expected.nix --- Nix
       > 1 { fish = false; localsend = true; sta  1 { fish = true; localsend = true; stat
       > . teVersion = "26.05"; zsh = false; }    . eVersion = "26.05"; zsh = true; }

So appears that only first ._ includes are evaluated and all others are rejected(tried with more includes as well). The only workaround was naming one of them something different from the other. I don't know if this is expected though. Sorry if I am burdening you but thought I'd point out, and as always thanks for great work you do here.

[sini]

I'll take a deeper look at how multiple namespaces compose under den.denful. This is only for the root level, nested should work, right?

[amanako]

Now that you mention it I noticed something strange and was able to reproduce it. If nesting 3 or more levels deep (den.aspects.aspect1.aspect2.aspect3._ or foo.aspect1.aspect2.aspect3._ for namespaces) same error about attribute "_" missing occurs. Less than that is fine.

{denTest, ...}: {
  flake.tests.bogus = {
    test-something = denTest (
      {
        inputs,
        den,
        foo,
        igloo,
        ...
      }: {
        imports = [(inputs.den.namespace "foo" false)];
        foo.shell.interpreter.fish = {
          entry.nixos.programs.fish.enable = true;
          additional.nixos.programs.fish.useBabelfish = true;
        };

        den.hosts.x86_64-linux.igloo.users.tux = {};

        den.aspects.igloo.includes = [
          foo.shell.interpreter.fish._
        ];

        expr = {
          fishEnable = igloo.programs.fish.enable;
          babelFish = igloo.programs.fish.useBabelfish;
        };

        expected = {
          fishEnable = true;
          babelFish = true;
        };
      }
    );
  };
}