Previously if function in defs had set pattern in argument,
this information would be lost.
This keeps functionArgs in set pattern,
by using functor and `__functionArgs`
that is later used by `lib.functionArgs`.
Otherwise nested types such as `attrsOf (attrsOf int)` won't have a
backwards compatible `type.nestedTypes.elemType.functor.wrapped`.
Follow-up work to https://github.com/NixOS/nixpkgs/pull/366015
This reverts commit ce8f304bb6.
The problem was simply a typo (nestedTypes.elemType ->
type.nestedTypes.elemType) ! And CI didn't run for lib in the orginal
PR which is why it didn't get caught.
This gives people some flexibility when they need a path type, and
prevents a "combinatorial explosion" of various path stops.
I've re-implemented our existing `path` and `pathInStore` types using
`pathWith`. Our existing `package` type is potentially a candidate for
similar treatment, but it's a little quirkier (there's some stuff with
`builtins.hasContext` and `toDerivation` that I don't completely
understand), and I didn't want to muddy this PR with that.
As a happy side effect of this work, we get a new feature: the ability
to create a type for paths *not* in the store. This is useful for when a
module needs a path to a file, and wants to protect people from
accidentally leaking that file into the nix store.
Passing the elaborated system defeats what pkgs/top-level/default.nix
tries to do: Pass only the original args and let defaults be inferred.
The underlying problem is that lib.systems.elaborate can not deal with
arbitrary overrides, but will often return an inconsistent system
description when partially overriding some values. This becomes most
prominent if trying to override an already elaborated system.
It creates interoperability issues that can not be reconciled by
`lib` or maintainers of projects that use the Nixpkgs library.
Occasionally, end users may be able to solve the problems they run
into, but most are not prepared to deal with this set of problems,
nor should they be.
Typical conflict:
- User wants to propagate their own lib, because it has some function
they like to use throughout their projects
- Project maintainer requires the project's lib to be used
No sane language uses a single namespace for combining all the things.
(Arguably not even C with its extensive use of prefixing)
Meanwhile, in Nix, all symbols are first class variables. We don't even
have the concept of a global top-level namespace to pour everything into.
With `lib` you can try to approximate that, I get the appeal of its
apparent simplicity, but since `lib` can't be global, we just don't even
get that apparent simplicity.
I apologize for not offering concrete solutions to this in the text.
The manuals are limited to reference documentation.
Alternatives - of which we have multiple - are best provided in
task-oriented documentation, e.g. nix.dev.
It's not the longest-open PR (May 28, 2023 -> Jan 22, 2025) but it took a while. This PR introduces a unified approach to implementing build helpers that support fixed-point arguments and bring such support to existing build helpers.
