This new type has unsurprising merge behavior: Only attribute sets are
merged together (recursively), and only if they don't conflict.
This is in contrast to the existing types:
- types.attrs is problematic because later definitions completely
override attributes of earlier definitions, and it doesn't support
mkIf and co.
- types.unspecified is very similar to types.attrs, but it has smart
merging behavior that often doesn't make sense, and it doesn't support
all types
Previously the only way to deprecate a type was using
theType = lib.warn "deprecated" (mkOptionType ...)
This caused the warning to be emitted when the type was evaluated, but
the error didn't include which option actually used that type.
With this commit, types can specify a deprecationMessage, which when
non-null, is printed along with the option that uses the type
Submodules that have a freeform type set behave as if that was the type
of the option itself (for values that don't have an option). Since the
submodules options are shown as separate entries in the manual, it makes
sense to show the freeform type as the submodule type.
Not all modules use name attribute as the name of the submodule, for example,
environment.etc uses target. We will need to maintain a list of exceptions.
The standard attrsOf is strict in its *values*, meaning it's impossible to
access only one attribute value without evaluating all others as well.
lazyAttrsOf is a version that doesn't have that problem, at the expense
of conditional definitions not properly working anymore.
Previously when this function was called without a value coercible to a
string it would throw an error instead of returning false. Now it does.
As a result this now allows the use of a type like `either path attrs`
without it erroring out when a definition is an attribute set.
The warning about there not being a isPath primop was removed because
this is not the case anymore, there is builtins.isPath. But also there
always was `builtins.typeOf x == "path"` that could've been used
instead. However the path type now stands for more than just path types,
but absolute paths in general.
This reverts commit eec83d41e3.
This broke hydra evaluation because with this commit submodule values
are allowed to be paths, however the certmgr module uses `either
(submodule ...) path` in its type, meaning it already used paths for
something else which would now be interpreted as a submodule.
The explicit remove helped to uncover some hidden uses of `optionSet`
in NixOps. However it makes life harder for end-users of NixOps - it will
be impossible to deploy 19.03 systems with old NixOps, but there is no
new release of NixOps with `optionSet` fixes.
Also, "deprecation" process isn't well defined. Even that `optionSet` was
declared "deprecated" for many years, it was never announced. Hence, I
leave "deprecation" announce. Then, 3 releases after announce,
we can announce removal of this feature.
This type has to be removed, not `throw`-ed in runtime, because it makes
some perfectly fine code to fail. For example:
```
$ nix-instantiate --eval -E '(import <nixpkgs/lib>).types' --strict
trace: `types.list` is deprecated; use `types.listOf` instead
error: types.optionSet is deprecated; use types.submodule instead
(use '--show-trace' to show detailed location information)
```
The previous description "string" is misleading in the full options
manual pages; they are actually concatenated strings, with a specific
character.
The empty string version ("types.string") has been special-cased to
provide a better message.
Assigning a list of 10 or more elements to an option having the type
`loaOf a` produces a configuration value that is not honoring the
order of the original list. This commit fixes this and a related issue
arising when 10 or more lists are merged into this type of option.
Without this change
(coercedTo str toInt int).check "foo"
would evaluate to true, even though
(coercedTo str toInt int).merge {} [{ value = "foo"; }]
will throw an error because "foo" can't be coerced to an int.
This file doesn't evaluate in 32-bit versions of Nix because the integer
type is a signed 32-bit integer there, so 4294967296 causes an 'invalid
integer' error. I see no other way around than commenting this out :(
(s32 could be made to work by tweaking the expressions a bit, but didn't
do that for now since it'd be asymmetric to have s32 but no u32).
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
)
* lib: introduce imap0, imap1
For historical reasons, imap starts counting at 1 and it's not
consistent with the rest of the lib.
So for now we split imap into imap0 that starts counting at zero and
imap1 that starts counting at 1. And imap is marked as deprecated.
See c71e2d4235 (commitcomment-21873221)
* replace uses of lib.imap
* lib: move imap to deprecated.nix
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.
Nix style seems to have settled on not using spaces between bound
variable names and the lambda : so I also tried to make those somewhat
more consistent throughout.
If an option value is not a list, you now get
The option value `bla' in `file.nix' is not a list.
rather than
value is a string while a list was expected, at .../nixpkgs/lib/lists.nix:49:56
This reverts commit 0f0805b, because @nbp had concerns about whether
this would be a good idea and pointed out problems with this.
We currently do not have a case where "either" is used in conjunction
with submodules, but I'm reverting it anyway to prevent people from
adding options using that type in that way.
This is now being reviewed in #14053.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
So far the "either" type only handled "flat" types, so you couldn't do
something like:
type = either int (submodule {
options = ...;
});
Not only caused this the submodule's options not being checked but also
not show up in the documentation.
This was something we stumbled on with #13916.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Cc: @edolstra
- Enforce that an option declaration has a "defaultText" if and only if the
type of the option derives from "package", "packageSet" or "nixpkgsConfig"
and if a "default" attribute is defined.
- Enforce that the value of the "example" attribute is wrapped with "literalExample"
if the type of the option derives from "package", "packageSet" or "nixpkgsConfig".
- Warn if a "defaultText" is defined in an option declaration if the type of
the option does not derive from "package", "packageSet" or "nixpkgsConfig".
- Warn if no "type" is defined in an option declaration.
Nobody seems to have noticed this (except @Profpatsch) that options with
a "package" type do not get included in the manual.
So debugging this was a bit more involving because while generating the
manual there is an optionList' attribute built from the collected
attributes of all the option declarations.
Up to that point everything is fine except if it comes to
builtins.toXML, where attributes with { type = "derivation" } won't get
included, for example see here:
nix-repl> builtins.toXML { type = "derivation"; foo = "bar"; }
"<?xml version='1.0' encoding='utf-8'?>\n<expr>\n <derivation>
<repeated />\n </derivation>\n</expr>\n"
nix-repl> builtins.toXML { type = "somethingelse"; foo = "bar"; }
"<?xml version='1.0' encoding='utf-8'?>\n<expr>\n <attrs>
<attr name=\"foo\">\n <string value=\"bar\" />\n </attr>
<attr name=\"type\">\n <string value=\"somethingelse\" />
</attr>\n </attrs>\n</expr>\n"
The following function in libexpr/eval.cc (Nix) is responsible for toXML
dropping the attributes:
bool EvalState::isDerivation(Value & v)
{
if (v.type != tAttrs) return false;
Bindings::iterator i = v.attrs->find(sType);
if (i == v.attrs->end()) return false;
forceValue(*i->value);
if (i->value->type != tString) return false;
return strcmp(i->value->string.s, "derivation") == 0;
}
So I've renamed this now to "package" which is not only more consistent
with the option type but also shouldn't cause similar issues anymore.
Tested this on base of b60ceea, because building the dependencies on
recent libc/staging changes on master took too long.
Signed-off-by: aszlig <aszlig@redmoonstudios.org>
Reported-by: Profpatsch <mail@profpatsch.de>
This improves error messages when a set or a list is used where a path
was expected. For an example, if you used a package set (as opposed to a
single package) in systemPackages before this commit, the error was:
```
cannot coerce a list to a string, at "/home/nixpkgs/lib/types.nix":103:37
```
Now, the error message reads:
```
The option value `environment.systemPackages' in `/etc/nixos/configuration.nix' is not a list of paths.
```
For example, this allows writing
nix.package = /nix/store/786mlvhd17xvcp2r4jmmay6jj4wj6b7f-nix-1.10pre4206_896428c;
Also, document types.package in the manual.
Previously, conflicting definitions would merge to "true". Now they
give an error, e.g.
error: The option `hardware.enableAllFirmware' has conflicting definitions, in `/etc/nixos/configurations/misc/eelco/stuff.nix' and `/etc/nixos/configurations/misc/eelco/mandark.nix'.
This move idioms which were used in `evalOptionValue` and in the `merge`
functions of `listOf` and `attrsOf` types, such that we can use a names such
as `isDefined` and `optionalValue` instead or repeating identical
comparisons of `defsFinal == []`.
Ideally the module system could be configured pretty much completely by
the contents of the modules themselves, so add comments about avoiding
complicating it further and possibly removing now-redundant
configurability from the existing interface.
This simplifes typechecking and allows properties to be used inside of
the attribute sets.
This fixes the empty synergy-client and synergy-server services
previously generated on systems with synergy disabled.
Now we should be able to have multiple declaration of the same option as
long as all declarations have the same type. If the type has a sub module,
then it is merged with the submodules of other declarations, as done with
option sets.
In addition, the file of the option declaration is passed into the
submodule, such as the documentation can display it correctly.
It is parameterized by a function that takes a name and evaluates to the
option type for the attribute of that name. Together with
submoduleWithExtraArgs, this subsumes nixosSubmodule.
E.g.
The unique option `fileSystems./.device' is defined multiple times, in `/etc/nixos/configuration.nix' and `/etc/nixos/foo.nix'.
This requires passing file/value tuples to the merge functions.
An annoying and dangerous property of "types.string" is that it merges
multiple definitions by concatenating them, which almost never
produces a sensible result. (Those options for which it does make
sense typically should use "types.lines" instead, and things only work
because the option definitions already end in a newline.) Of course,
you can use "types.uniq types.string", but that's rather verbose, and
inconsistent with other basic types like "types.int".
Changing the behaviour of "types.string" to be unique by default is
not an option, given the large number of options that use it. So
instead, we now have "types.str", which is equivalent to "types.uniq
types.string".
For instance, if time.timeZone is defined multiple times, you now get
the error message:
error: user-thrown exception: The unique option `time.timeZone' is defined multiple times, in `/etc/nixos/configurations/misc/eelco/x11vnc.nix' and `/etc/nixos/configuration.nix'.
while previously you got:
error: user-thrown exception: Multiple definitions of string. Only one is allowed for this option.
and only an inspection of the stack trace gave a clue as to what
option caused the problem.
The major changes are:
* The evaluation is now driven by the declared options. In
particular, this fixes the long-standing problem with lack of
laziness of disabled option definitions. Thus, a configuration like
config = mkIf false {
environment.systemPackages = throw "bla";
};
will now evaluate without throwing an error. This also improves
performance since we're not evaluating unused option definitions.
* The implementation of properties is greatly simplified.
* There is a new type constructor "submodule" that replaces
"optionSet". Unlike "optionSet", "submodule" gets its option
declarations as an argument, making it more like "listOf" and other
type constructors. A typical use is:
foo = mkOption {
type = type.attrsOf (type.submodule (
{ config, ... }:
{ bar = mkOption { ... };
xyzzy = mkOption { ... };
}));
};
Existing uses of "optionSet" are automatically mapped to
"submodule".
* Modules are now checked for unsupported attributes: you get an error
if a module contains an attribute other than "config", "options" or
"imports".
* The new implementation is faster and uses much less memory.
