the fix to extendDerivation in #140051 unwittingly worsened eval performance by
quite a bit. set elements alone needed over 1GB extra after the change, which
seems disproportionate to how small it was. if we flip the logic used to
determine which outputs to install around and keep a "this one exactly" flag in
the specific outputs instead of a "all of them" in the root we can avoid most
of that cost.
- Remove inheritance of `lists.fold` as it isn't used anywhere.
- Inherit `foldl'` for consistency as only `cartesianProductOfSets` explicitly
reference lib.
- Inline `foldr` to generate nested attrs instead of using `listToAttrs` and `tail`.
This makes the function available without having to evaluate the
Nixpkgs fix-point, making it available in a more natural way for
code that deals with multiple Nixpkgs invocations.
Its definition is coupled to Nix rather than Nixpkgs, so it will
feel right at home in lib.
Suppose I have a Gemfile like this:
source "https://rubygems.org"
gem "actioncable"
gem "websocket-driver", group: :test
The gemset.nix generated by Bundix 2.4.1 will set ActionCable's groups
to [ "default" ], and websocket-driver's to [ "test" ]. This means that
the generated bundlerEnv wouldn't include websocket-driver unless the
test group was included, even though it's required by the default group.
This is arguably a bug in Bundix (websocket-driver's groups should
probably be [ "default" "test" ] or just [ "default" ]), but there's no
reason bundlerEnv should omit dependencies even given such an input --
it won't necessarily come from Bundix, and it would be good for
bundlerEnv to do the right thing.
To fix this, filterGemset is now a recursive function, that adds
dependencies of gems in the group to the filtered gemset until it
stabilises on the gems that match the required groups, and all of their
recursive dependencies.
$ nix repl lib
Welcome to Nix version 2.0.2. Type :? for help.
Loading 'lib'...
Added 350 variables.
-- this is the exact example from the function's documentation:
nix-repl> recursiveUpdateUntil (path: l: r: path == ["foo"]) {
# first attribute set
foo.bar = 1;
foo.baz = 2;
bar = 3;
} {
#second attribute set
foo.bar = 1;
foo.quz = 2;
baz = 4;
}
{ bar = 3; baz = 4; foo = { bar = 1; baz = 2; quz = 2; }; }
-- although the documentation says:
{
foo.bar = 1; # 'foo.*' from the second set
foo.quz = 2; #
bar = 3; # 'bar' from the first set
baz = 4; # 'baz' from the second set
}
This does break the API of being able to import any lib file and get
its libs, however I'm not sure people did this.
I made this while exploring being able to swap out docFn with a stub
in #2305, to avoid functor performance problems. I don't know if that
is going to move forward (or if it is a problem or not,) but after
doing all this work figured I'd put it up anyway :)
Two notable advantages to this approach:
1. when a lib inherits another lib's functions, it doesn't
automatically get put in to the scope of lib
2. when a lib implements a new obscure functions, it doesn't
automatically get put in to the scope of lib
Using the test script (later in this commit) I got the following diff
on the API:
+ diff master fixed-lib
11764a11765,11766
> .types.defaultFunctor
> .types.defaultTypeMerge
11774a11777,11778
> .types.isOptionType
> .types.isType
11781a11786
> .types.mkOptionType
11788a11794
> .types.setType
11795a11802
> .types.types
This means that this commit _adds_ to the API, however I can't find a
way to fix these last remaining discrepancies. At least none are
_removed_.
Test script (run with nix-repl in the PATH):
#!/bin/sh
set -eux
repl() {
suff=${1:-}
echo "(import ./lib)$suff" \
| nix-repl 2>&1
}
attrs_to_check() {
repl "${1:-}" \
| tr ';' $'\n' \
| grep "\.\.\." \
| cut -d' ' -f2 \
| sed -e "s/^/${1:-}./" \
| sort
}
summ() {
repl "${1:-}" \
| tr ' ' $'\n' \
| sort \
| uniq
}
deep_summ() {
suff="${1:-}"
depth="${2:-4}"
depth=$((depth - 1))
summ "$suff"
for attr in $(attrs_to_check "$suff" | grep -v "types.types"); do
if [ $depth -eq 0 ]; then
summ "$attr" | sed -e "s/^/$attr./"
else
deep_summ "$attr" "$depth" | sed -e "s/^/$attr./"
fi
done
}
(
cd nixpkgs
#git add .
#git commit -m "Auto-commit, sorry" || true
git checkout fixed-lib
deep_summ > ../fixed-lib
git checkout master
deep_summ > ../master
)
if diff master fixed-lib; then
echo "SHALLOW MATCH!"
fi
(
cd nixpkgs
git checkout fixed-lib
repl .types
)
The old forms presumably predates, or were made in ignorance of,
`let inherit`. This way is better style as the scoping as more lexical,
something which Nix can (or might already!) take advantage of.
The manuals are now evaluated with each derivation in `pkgs` (recursively)
replaced by a fake with path "\${pkgs.path.to.the.attribute}".
It isn't perfect, but it seems to cover a vast majority of use cases.
Caveat: even if the package is reached by a different means,
the path above will be shown and not e.g. `${config.services.foo.package}`.
As before, defaults created by `mkDefault` aren't displayed,
but documentation shouldn't (mostly) be a reason to use that anymore.
Note: t wouldn't be enough to just use `lib.mapAttrsRecursive`,
because derivations are also (special) attribute sets.
For example, this allows writing
nix.package = /nix/store/786mlvhd17xvcp2r4jmmay6jj4wj6b7f-nix-1.10pre4206_896428c;
Also, document types.package in the manual.
