To be pedantic, files are objects on a file system. Their content should be a Nix expression – though file names are just convention, we cannot prevent anyone from saving e.g. a borked Nix expressions or an image with Nix logo as a file with .nix extension.
An expression can be a function. You can find all kinds of expression in the Nix manual or in this tutorial.
Yes, a NixOS module is a Nix expression. It must evaluate into either an attribute set of specific shape, or a function that returns such attribute set. Note that it only has to evaluate into those, the following expression is a valid NixOS module, even though the top-level Nix expression is neither:
let
hostname = "london";
in
{ pkgs, ... }:
{
# …
}
I would say flake is the virtual object described by the flake manifest (flake.nix file) but I do not think the terminology is settled. The Nix manual has more materialistic definition:
A flake is a filesystem tree (typically fetched from a Git repository or a tarball) that contains a file named
flake.nixin the root directory.flake.nixspecifies some metadata about the flake such as dependencies (called inputs), as well as its outputs (the Nix values such as packages or NixOS modules provided by the flake).
And then later uses the same term for flake.nix.
No. The flake.nix file uses a subset of Nix syntax, where the top-level expression can only be an attribute set. Unlike in NixOS modules, evaluating to an attribute set is not enough.
Yes. It is builtins.import (or import) for short.
Right. See this discussion for the differences.
A flake can provide various outputs. NixOS modules are one of the output kinds (example). NixOS modules can import other NixOS modules.
A flake can also contain NixOS system outputs. NixOS system configurations are actually NixOS modules themselves so they can also import NixOS modules.
Flakes can declare other flakes as inputs, which will then be available as arguments to the outputs function inside flake.nix.