nixpkgs/lib/modules.nix

1504 lines
58 KiB
Nix
Raw Normal View History

Convert libs to a fixed-point 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 )
2017-07-29 00:05:35 +00:00
{ lib }:
let
2020-10-20 14:10:21 +00:00
inherit (lib)
addErrorContext
2020-10-20 14:10:21 +00:00
all
any
attrByPath
attrNames
catAttrs
2020-10-20 14:10:21 +00:00
concatLists
concatMap
concatStringsSep
2020-10-20 14:10:21 +00:00
elem
filter
foldl'
functionArgs
getAttrFromPath
genericClosure
2020-10-20 14:10:21 +00:00
head
id
imap1
isAttrs
isBool
isFunction
isInOldestRelease
isList
2020-10-20 14:10:21 +00:00
isString
length
mapAttrs
mapAttrsToList
2020-10-20 14:10:21 +00:00
mapAttrsRecursiveCond
min
optional
optionalAttrs
2020-10-20 14:10:21 +00:00
optionalString
recursiveUpdate
2020-10-20 14:10:21 +00:00
reverseList sort
seq
setAttrByPath
substring
throwIfNot
trace
typeOf
types
unsafeGetAttrPos
warn
warnIf
zipAttrs
zipAttrsWith
;
2020-10-20 14:10:21 +00:00
inherit (lib.options)
isOption
mkOption
showDefs
showFiles
showOption
unknownModule
;
inherit (lib.strings)
isConvertibleWithToString
;
showDeclPrefix = loc: decl: prefix:
" - option(s) with prefix `${showOption (loc ++ [prefix])}' in module `${decl._file}'";
showRawDecls = loc: decls:
concatStringsSep "\n"
(sort (a: b: a < b)
(concatMap
(decl: map
(showDeclPrefix loc decl)
(attrNames decl.options)
)
decls
));
/* See https://nixos.org/manual/nixpkgs/unstable/#module-system-lib-evalModules
or file://./../doc/module-system/module-system.chapter.md
2021-10-29 12:58:28 +00:00
!!! Please think twice before adding to this argument list! The more
that is specified here instead of in the modules themselves the harder
it is to transparently move a set of modules to be a submodule of another
config (as the proper arguments need to be replicated at each call to
evalModules) and the less declarative the module set is. */
evalModules = evalModulesArgs@
{ modules
, prefix ? []
, # This should only be used for special arguments that need to be evaluated
# when resolving module structure (like in imports). For everything else,
# there's _module.args. If specialArgs.modulesPath is defined it will be
# used as the base path for disabledModules.
specialArgs ? {}
, # `class`:
# A nominal type for modules. When set and non-null, this adds a check to
# make sure that only compatible modules are imported.
class ? null
, # This would be remove in the future, Prefer _module.args option instead.
args ? {}
, # This would be remove in the future, Prefer _module.check option instead.
check ? true
}:
let
2021-12-02 17:47:17 +00:00
withWarnings = x:
warnIf (evalModulesArgs?args) "The args argument to evalModules is deprecated. Please set config._module.args instead."
warnIf (evalModulesArgs?check) "The check argument to evalModules is deprecated. Please set config._module.check instead."
2021-12-02 17:47:17 +00:00
x;
legacyModules =
optional (evalModulesArgs?args) {
config = {
_module.args = args;
};
}
++ optional (evalModulesArgs?check) {
config = {
_module.check = mkDefault check;
};
};
regularModules = modules ++ legacyModules;
# This internal module declare internal options under the `_module'
# attribute. These options are fragile, as they are used by the
# module system to change the interpretation of modules.
#
# When extended with extendModules or moduleType, a fresh instance of
# this module is used, to avoid conflicts and allow chaining of
# extendModules.
internalModule = rec {
_file = "lib/modules.nix";
key = _file;
options = {
_module.args = mkOption {
# Because things like `mkIf` are entirely useless for
# `_module.args` (because there's no way modules can check which
# arguments were passed), we'll use `lazyAttrsOf` which drops
# support for that, in turn it's lazy in its values. This means e.g.
# a `_module.args.pkgs = import (fetchTarball { ... }) {}` won't
# start a download when `pkgs` wasn't evaluated.
type = types.lazyAttrsOf types.raw;
# Only render documentation once at the root of the option tree,
# not for all individual submodules.
# Allow merging option decls to make this internal regardless.
${if prefix == []
then null # unset => visible
else "internal"} = true;
# TODO: Change the type of this option to a submodule with a
# freeformType, so that individual arguments can be documented
# separately
description = ''
Additional arguments passed to each module in addition to ones
like `lib`, `config`,
and `pkgs`, `modulesPath`.
This option is also available to all submodules. Submodules do not
inherit args from their parent module, nor do they provide args to
their parent module or sibling submodules. The sole exception to
this is the argument `name` which is provided by
parent modules to a submodule and contains the attribute name
the submodule is bound to, or a unique generated name if it is
not bound to an attribute.
Some arguments are already passed by default, of which the
following *cannot* be changed with this option:
- {var}`lib`: The nixpkgs library.
- {var}`config`: The results of all options after merging the values from all modules together.
- {var}`options`: The options declared in all modules.
- {var}`specialArgs`: The `specialArgs` argument passed to `evalModules`.
- All attributes of {var}`specialArgs`
Whereas option values can generally depend on other option values
thanks to laziness, this does not apply to `imports`, which
must be computed statically before anything else.
For this reason, callers of the module system can provide `specialArgs`
which are available during import resolution.
For NixOS, `specialArgs` includes
{var}`modulesPath`, which allows you to import
extra modules from the nixpkgs package tree without having to
somehow make the module aware of the location of the
`nixpkgs` or NixOS directories.
```
{ modulesPath, ... }: {
imports = [
(modulesPath + "/profiles/minimal.nix")
];
}
```
For NixOS, the default value for this option includes at least this argument:
- {var}`pkgs`: The nixpkgs package set according to
the {option}`nixpkgs.pkgs` option.
'';
};
_module.check = mkOption {
2015-06-15 16:10:26 +00:00
type = types.bool;
internal = true;
default = true;
description = "Whether to check whether all option definitions have matching declarations.";
};
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
_module.freeformType = mkOption {
type = types.nullOr types.optionType;
internal = true;
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
default = null;
description = ''
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
If set, merge all definitions that don't have an associated option
together using this type. The result then gets combined with the
values of all declared options to produce the final `
config` value.
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
If this is `null`, definitions without an option
will throw an error unless {option}`_module.check` is
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
turned off.
'';
};
2022-07-12 12:33:10 +00:00
_module.specialArgs = mkOption {
readOnly = true;
internal = true;
description = ''
2022-07-12 12:33:10 +00:00
Externally provided module arguments that can't be modified from
within a configuration, but can be used in module imports.
'';
};
};
config = {
_module.args = {
inherit extendModules;
moduleType = type;
};
2022-07-12 12:33:10 +00:00
_module.specialArgs = specialArgs;
};
};
merged =
let collected = collectModules
class
(specialArgs.modulesPath or "")
(regularModules ++ [ internalModule ])
({ inherit lib options config specialArgs; } // specialArgs);
in mergeModules prefix (reverseList collected);
options = merged.matchedOptions;
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
config =
let
# For definitions that have an associated option
declaredConfig = mapAttrsRecursiveCond (v: ! isOption v) (_: v: v.value) options;
# If freeformType is set, this is for definitions that don't have an associated option
freeformConfig =
let
defs = map (def: {
file = def.file;
value = setAttrByPath def.prefix def.value;
}) merged.unmatchedDefns;
in if defs == [] then {}
else declaredConfig._module.freeformType.merge prefix defs;
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
in if declaredConfig._module.freeformType == null then declaredConfig
# Because all definitions that had an associated option ended in
# declaredConfig, freeformConfig can only contain the non-option
# paths, meaning recursiveUpdate will never override any value
else recursiveUpdate freeformConfig declaredConfig;
checkUnmatched =
lib/modules: Implement freeform modules For programs that have a lot of (Nix-representable) configuration options, a simple way to represent this in a NixOS module is to declare an option of a type like `attrsOf str`, representing a key-value mapping which then gets generated into a config file. However with such a type, there's no way to add type checking for only some key values. On the other end of the spectrum, one can declare a single separate option for every key value that the program supports, ending up with a module with potentially 100s of options. This has the benefit that every value gets type checked, catching mistakes at evaluation time already. However the disadvantage is that the module becomes big, becomes coupled to the program version and takes a lot of effort to write and maintain. Previously there was a middle ground between these two extremes: Declare an option of a type like `attrsOf str`, but declare additional separate options for the values you wish to have type checked, and assign their values to the `attrsOf str` option. While this works decently, it has the problem of duplicated options, since now both the additional options and the `attrsOf str` option can be used to set a key value. This leads to confusion about what should happen if both are set, which defaults should apply, and more. Now with this change, a middle ground becomes available that solves above problems: The module system now supports setting a freeform type, which gets used for all definitions that don't have an associated option. This means that you can now declare all options you wish to have type checked, while for the rest a freeform type like `attrsOf str` can be used.
2020-03-19 23:17:41 +00:00
if config._module.check && config._module.freeformType == null && merged.unmatchedDefns != [] then
lib/modules: improve error-message for undeclared options if prefix contains no options An easy-to-make mistake when declaring e.g. a submodule is the accidental confusion of `options` and `config`: types.submodule { config = { foo = mkOption { /* ... */ }; }; } However the error-message The option `[definition 1-entry 1].foo' defined in `<expr.nix>' does not exist. is fairly unhelpful because it seems as the options are declared at the first sight. In fact, it took a colleague and me a while to track down such a mistake a few days ago and we both agreed that this should be somehow caught to save the time we spent debugging the module in question. At first I decided to catch this error in the `submodules`-type directly by checking whether `options` is undeclared, however this becomes fairly complicated as soon as a submodule-declaration e.g. depends on existing `config`-values which would've lead to some ugly `builtins.tryExec`-heuristic. This patch now simply checks if the option's prefix has any options defined if a point in evaluation is reached where it's clear that the option in question doesn't exist. This means that this patch doesn't change the logic of the module system, it only provides a more detailed error in certain cases: The option `[definition 1-entry 1].foo' defined in `<expr.nix>' does not exist. However it seems as there are no options defined in [definition 1-entry 1]. Are you sure you've declared your options properly? This happens if you e.g. declared your options in `types.submodule' under `config' rather than `options'.
2020-08-14 17:26:18 +00:00
let
firstDef = head merged.unmatchedDefns;
baseMsg =
let
optText = showOption (prefix ++ firstDef.prefix);
defText =
addErrorContext
"while evaluating the error message for definitions for `${optText}', which is an option that does not exist"
(addErrorContext
"while evaluating a definition from `${firstDef.file}'"
( showDefs [ firstDef ])
);
in
"The option `${optText}' does not exist. Definition values:${defText}";
lib/modules: improve error-message for undeclared options if prefix contains no options An easy-to-make mistake when declaring e.g. a submodule is the accidental confusion of `options` and `config`: types.submodule { config = { foo = mkOption { /* ... */ }; }; } However the error-message The option `[definition 1-entry 1].foo' defined in `<expr.nix>' does not exist. is fairly unhelpful because it seems as the options are declared at the first sight. In fact, it took a colleague and me a while to track down such a mistake a few days ago and we both agreed that this should be somehow caught to save the time we spent debugging the module in question. At first I decided to catch this error in the `submodules`-type directly by checking whether `options` is undeclared, however this becomes fairly complicated as soon as a submodule-declaration e.g. depends on existing `config`-values which would've lead to some ugly `builtins.tryExec`-heuristic. This patch now simply checks if the option's prefix has any options defined if a point in evaluation is reached where it's clear that the option in question doesn't exist. This means that this patch doesn't change the logic of the module system, it only provides a more detailed error in certain cases: The option `[definition 1-entry 1].foo' defined in `<expr.nix>' does not exist. However it seems as there are no options defined in [definition 1-entry 1]. Are you sure you've declared your options properly? This happens if you e.g. declared your options in `types.submodule' under `config' rather than `options'.
2020-08-14 17:26:18 +00:00
in
if attrNames options == [ "_module" ]
# No options were declared at all (`_module` is built in)
# but we do have unmatched definitions, and no freeformType (earlier conditions)
then
let
optionName = showOption prefix;
in
if optionName == ""
then throw ''
${baseMsg}
2021-08-25 22:37:33 +00:00
It seems as if you're trying to declare an option by placing it into `config' rather than `options'!
''
else
throw ''
${baseMsg}
However there are no options defined in `${showOption prefix}'. Are you sure you've
declared your options properly? This can happen if you e.g. declared your options in `types.submodule'
under `config' rather than `options'.
''
lib/modules: improve error-message for undeclared options if prefix contains no options An easy-to-make mistake when declaring e.g. a submodule is the accidental confusion of `options` and `config`: types.submodule { config = { foo = mkOption { /* ... */ }; }; } However the error-message The option `[definition 1-entry 1].foo' defined in `<expr.nix>' does not exist. is fairly unhelpful because it seems as the options are declared at the first sight. In fact, it took a colleague and me a while to track down such a mistake a few days ago and we both agreed that this should be somehow caught to save the time we spent debugging the module in question. At first I decided to catch this error in the `submodules`-type directly by checking whether `options` is undeclared, however this becomes fairly complicated as soon as a submodule-declaration e.g. depends on existing `config`-values which would've lead to some ugly `builtins.tryExec`-heuristic. This patch now simply checks if the option's prefix has any options defined if a point in evaluation is reached where it's clear that the option in question doesn't exist. This means that this patch doesn't change the logic of the module system, it only provides a more detailed error in certain cases: The option `[definition 1-entry 1].foo' defined in `<expr.nix>' does not exist. However it seems as there are no options defined in [definition 1-entry 1]. Are you sure you've declared your options properly? This happens if you e.g. declared your options in `types.submodule' under `config' rather than `options'.
2020-08-14 17:26:18 +00:00
else throw baseMsg
else null;
checked = seq checkUnmatched;
extendModules = extendArgs@{
modules ? [],
specialArgs ? {},
prefix ? [],
}:
evalModules (evalModulesArgs // {
inherit class;
modules = regularModules ++ modules;
specialArgs = evalModulesArgs.specialArgs or {} // specialArgs;
prefix = extendArgs.prefix or evalModulesArgs.prefix or [];
});
type = types.submoduleWith {
inherit modules specialArgs class;
};
2021-12-02 17:47:17 +00:00
result = withWarnings {
_type = "configuration";
options = checked options;
config = checked (removeAttrs config [ "_module" ]);
_module = checked (config._module);
inherit extendModules type;
class = class;
};
in result;
# collectModules :: (class: String) -> (modulesPath: String) -> (modules: [ Module ]) -> (args: Attrs) -> [ Module ]
#
# Collects all modules recursively through `import` statements, filtering out
# all modules in disabledModules.
collectModules = class: let
# Like unifyModuleSyntax, but also imports paths and calls functions if necessary
loadModule = args: fallbackFile: fallbackKey: m:
if isFunction m then
unifyModuleSyntax fallbackFile fallbackKey (applyModuleArgs fallbackKey m args)
else if isAttrs m then
if m._type or "module" == "module" then
unifyModuleSyntax fallbackFile fallbackKey m
else if m._type == "if" || m._type == "override" then
loadModule args fallbackFile fallbackKey { config = m; }
else
throw (
"Could not load a value as a module, because it is of type ${lib.strings.escapeNixString m._type}"
+ optionalString (fallbackFile != unknownModule) ", in file ${toString fallbackFile}."
+ optionalString (m._type == "configuration") " If you do intend to import this configuration, please only import the modules that make up the configuration. You may have to create a `let` binding, file or attribute to give yourself access to the relevant modules.\nWhile loading a configuration into the module system is a very sensible idea, it can not be done cleanly in practice."
# Extended explanation: That's because a finalized configuration is more than just a set of modules. For instance, it has its own `specialArgs` that, by the nature of `specialArgs` can't be loaded through `imports` or the the `modules` argument. So instead, we have to ask you to extract the relevant modules and use those instead. This way, we keep the module system comparatively simple, and hopefully avoid a bad surprise down the line.
)
else if isList m then
let defs = [{ file = fallbackFile; value = m; }]; in
throw "Module imports can't be nested lists. Perhaps you meant to remove one level of lists? Definitions: ${showDefs defs}"
else unifyModuleSyntax (toString m) (toString m) (applyModuleArgsIfFunction (toString m) (import m) args);
checkModule =
if class != null
then
m:
if m._class != null -> m._class == class
then m
else
throw "The module ${m._file or m.key} was imported into ${class} instead of ${m._class}."
else
m: m;
/*
Collects all modules recursively into the form
{
disabled = [ <list of disabled modules> ];
# All modules of the main module list
modules = [
{
key = <key1>;
module = <module for key1>;
# All modules imported by the module for key1
modules = [
{
key = <key1-1>;
module = <module for key1-1>;
# All modules imported by the module for key1-1
modules = [ ... ];
}
...
];
}
...
];
}
*/
collectStructuredModules =
let
collectResults = modules: {
disabled = concatLists (catAttrs "disabled" modules);
inherit modules;
};
in parentFile: parentKey: initialModules: args: collectResults (imap1 (n: x:
let
module = checkModule (loadModule args parentFile "${parentKey}:anon-${toString n}" x);
collectedImports = collectStructuredModules module._file module.key module.imports args;
in {
key = module.key;
module = module;
modules = collectedImports.modules;
disabled = (if module.disabledModules != [] then [{ file = module._file; disabled = module.disabledModules; }] else []) ++ collectedImports.disabled;
}) initialModules);
# filterModules :: String -> { disabled, modules } -> [ Module ]
#
# Filters a structure as emitted by collectStructuredModules by removing all disabled
# modules recursively. It returns the final list of unique-by-key modules
filterModules = modulesPath: { disabled, modules }:
let
moduleKey = file: m:
if isString m
then
if substring 0 1 m == "/"
then m
else toString modulesPath + "/" + m
else if isConvertibleWithToString m
then
if m?key && m.key != toString m
then
throw "Module `${file}` contains a disabledModules item that is an attribute set that can be converted to a string (${toString m}) but also has a `.key` attribute (${m.key}) with a different value. This makes it ambiguous which module should be disabled."
else
toString m
else if m?key
then
m.key
else if isAttrs m
then throw "Module `${file}` contains a disabledModules item that is an attribute set, presumably a module, that does not have a `key` attribute. This means that the module system doesn't have any means to identify the module that should be disabled. Make sure that you've put the correct value in disabledModules: a string path relative to modulesPath, a path value, or an attribute set with a `key` attribute."
else throw "Each disabledModules item must be a path, string, or a attribute set with a key attribute, or a value supported by toString. However, one of the disabledModules items in `${toString file}` is none of that, but is of type ${typeOf m}.";
disabledKeys = concatMap ({ file, disabled }: map (moduleKey file) disabled) disabled;
keyFilter = filter (attrs: ! elem attrs.key disabledKeys);
in map (attrs: attrs.module) (genericClosure {
startSet = keyFilter modules;
operator = attrs: keyFilter attrs.modules;
});
in modulesPath: initialModules: args:
filterModules modulesPath (collectStructuredModules unknownModule "" initialModules args);
/* Wrap a module with a default location for reporting errors. */
setDefaultModuleLocation = file: m:
{ _file = file; imports = [ m ]; };
/* Massage a module into canonical form, that is, a set consisting
of options, config and imports attributes. */
unifyModuleSyntax = file: key: m:
let
addMeta = config: if m ? meta
then mkMerge [ config { meta = m.meta; } ]
else config;
addFreeformType = config: if m ? freeformType
then mkMerge [ config { _module.freeformType = m.freeformType; } ]
else config;
2016-05-09 05:53:27 +00:00
in
2013-10-28 16:40:36 +00:00
if m ? config || m ? options then
let badAttrs = removeAttrs m ["_class" "_file" "key" "disabledModules" "imports" "options" "config" "meta" "freeformType"]; in
if badAttrs != {} then
throw "Module `${key}' has an unsupported attribute `${head (attrNames badAttrs)}'. This is caused by introducing a top-level `config' or `options' attribute. Add configuration attributes immediately on the top level instead, or move all of them (namely: ${toString (attrNames badAttrs)}) into the explicit `config' attribute."
else
{ _file = toString m._file or file;
_class = m._class or null;
2013-10-29 13:15:33 +00:00
key = toString m.key or key;
disabledModules = m.disabledModules or [];
imports = m.imports or [];
options = m.options or {};
config = addFreeformType (addMeta (m.config or {}));
}
else
# shorthand syntax
throwIfNot (isAttrs m) "module ${file} (${key}) does not look like a module."
{ _file = toString m._file or file;
_class = m._class or null;
2013-10-29 13:15:33 +00:00
key = toString m.key or key;
disabledModules = m.disabledModules or [];
2013-10-28 16:40:36 +00:00
imports = m.require or [] ++ m.imports or [];
options = {};
config = addFreeformType (removeAttrs m ["_class" "_file" "key" "disabledModules" "require" "imports" "freeformType"]);
};
applyModuleArgsIfFunction = key: f: args@{ config, ... }:
if isFunction f then applyModuleArgs key f args else f;
applyModuleArgs = key: f: args@{ config, ... }:
let
# Module arguments are resolved in a strict manner when attribute set
# deconstruction is used. As the arguments are now defined with the
# config._module.args option, the strictness used on the attribute
# set argument would cause an infinite loop, if the result of the
# option is given as argument.
#
# To work-around the strictness issue on the deconstruction of the
# attributes set argument, we create a new attribute set which is
# constructed to satisfy the expected set of attributes. Thus calling
# a module will resolve strictly the attributes used as argument but
# not their values. The values are forwarding the result of the
# evaluation of the option.
context = name: ''while evaluating the module argument `${name}' in "${key}":'';
extraArgs = mapAttrs (name: _:
addErrorContext (context name)
2021-05-06 02:59:27 +00:00
(args.${name} or config._module.args.${name})
) (functionArgs f);
# Note: we append in the opposite order such that we can add an error
2022-12-17 23:59:29 +00:00
# context on the explicit arguments of "args" too. This update
# operator is used to make the "args@{ ... }: with args.lib;" notation
# works.
in f (args // extraArgs);
/* Merge a list of modules. This will recurse over the option
declarations in all modules, combining them into a single set.
At the same time, for each option declaration, it will merge the
corresponding option definitions in all machines, returning them
in the value attribute of each option.
This returns a set like
{
# A recursive set of options along with their final values
matchedOptions = {
foo = { _type = "option"; value = "option value of foo"; ... };
bar.baz = { _type = "option"; value = "option value of bar.baz"; ... };
...
};
# A list of definitions that weren't matched by any option
unmatchedDefns = [
{ file = "file.nix"; prefix = [ "qux" ]; value = "qux"; }
...
];
}
*/
mergeModules = prefix: modules:
mergeModules' prefix modules
(concatMap (m: map (config: { file = m._file; inherit config; }) (pushDownProperties m.config)) modules);
nixos/modules: Add declarationPositions What it does: line and column level *declaration* position information: $ nix repl . nix-repl> :p nixosConfigurations.micro.options.environment.systemPackages.declarationPositions [ { column = 7; file = "/nix/store/24aj3k7fgqv3ly7qkbf98qvphasrw9nb-source/nixos/modules/config/system-path.nix"; line = 63; } ] Use cases: - ctags over NixOS options, as will be presented at NixCon 2023 ;) - improving the documentation pages to go to the exact line of the declarations. Related work: - https://github.com/NixOS/nixpkgs/pull/65024 This one does it for all *definitions* rather than declarations, and it was not followed through with due to performance worries. - https://github.com/NixOS/nixpkgs/pull/208173 The basis for this change. This change is just a rebase of that one. I split it out to add the capability before adding users of it, in order to simplify review. However, the ctags script in there is a sample user of this feature. Benchmarks: conducted by evaluating my own reasonably complex NixOS configuration with the command: `hyperfine -S none -w 1 -- "nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath"` ``` Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.971 s ± 0.254 s [User: 5.872 s, System: 1.388 s] Range (min … max): 8.574 s … 9.327 s 10 runs Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.766 s ± 0.160 s [User: 5.873 s, System: 1.346 s] Range (min … max): 8.496 s … 9.033 s 10 runs ``` Summary of results: it seems to be in the noise, this does not cause any visible regression in times.
2022-12-28 22:39:05 +00:00
mergeModules' = prefix: modules: configs:
let
# an attrset 'name' => list of submodules that declare name.
declsByName =
2023-07-11 10:38:49 +00:00
zipAttrsWith
(n: concatLists)
(map
(module: let subtree = module.options; in
if !(isAttrs subtree) then
throw ''
An option declaration for `${concatStringsSep "." prefix}' has type
`${typeOf subtree}' rather than an attribute set.
Did you mean to define this outside of `options'?
''
else
mapAttrs
(n: option:
[{ inherit (module) _file; pos = unsafeGetAttrPos n subtree; options = option; }]
)
subtree
2023-07-11 10:38:49 +00:00
)
nixos/modules: Add declarationPositions What it does: line and column level *declaration* position information: $ nix repl . nix-repl> :p nixosConfigurations.micro.options.environment.systemPackages.declarationPositions [ { column = 7; file = "/nix/store/24aj3k7fgqv3ly7qkbf98qvphasrw9nb-source/nixos/modules/config/system-path.nix"; line = 63; } ] Use cases: - ctags over NixOS options, as will be presented at NixCon 2023 ;) - improving the documentation pages to go to the exact line of the declarations. Related work: - https://github.com/NixOS/nixpkgs/pull/65024 This one does it for all *definitions* rather than declarations, and it was not followed through with due to performance worries. - https://github.com/NixOS/nixpkgs/pull/208173 The basis for this change. This change is just a rebase of that one. I split it out to add the capability before adding users of it, in order to simplify review. However, the ctags script in there is a sample user of this feature. Benchmarks: conducted by evaluating my own reasonably complex NixOS configuration with the command: `hyperfine -S none -w 1 -- "nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath"` ``` Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.971 s ± 0.254 s [User: 5.872 s, System: 1.388 s] Range (min … max): 8.574 s … 9.327 s 10 runs Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.766 s ± 0.160 s [User: 5.873 s, System: 1.346 s] Range (min … max): 8.496 s … 9.033 s 10 runs ``` Summary of results: it seems to be in the noise, this does not cause any visible regression in times.
2022-12-28 22:39:05 +00:00
modules);
# The root of any module definition must be an attrset.
checkedConfigs =
assert
all
(c:
# TODO: I have my doubts that this error would occur when option definitions are not matched.
# The implementation of this check used to be tied to a superficially similar check for
# options, so maybe that's why this is here.
isAttrs c.config || throw ''
In module `${c.file}', you're trying to define a value of type `${typeOf c.config}'
rather than an attribute set for the option
`${concatStringsSep "." prefix}'!
This usually happens if `${concatStringsSep "." prefix}' has option
definitions inside that are not matched. Please check how to properly define
this option by e.g. referring to `man 5 configuration.nix'!
''
)
configs;
configs;
# an attrset 'name' => list of submodules that define name.
pushedDownDefinitionsByName =
2023-07-11 10:38:49 +00:00
zipAttrsWith
(n: concatLists)
(map
(module:
mapAttrs
(n: value:
map (config: { inherit (module) file; inherit config; }) (pushDownProperties value)
)
module.config
)
checkedConfigs);
# extract the definitions for each loc
rawDefinitionsByName =
2023-07-11 10:38:49 +00:00
zipAttrsWith
(n: concatLists)
(map
(module:
mapAttrs
(n: value:
[{ inherit (module) file; inherit value; }]
)
module.config
)
checkedConfigs);
# Convert an option tree decl to a submodule option decl
optionTreeToOption = decl:
if isOption decl.options
then decl
else decl // {
options = mkOption {
type = types.submoduleWith {
modules = [ { options = decl.options; } ];
# `null` is not intended for use by modules. It is an internal
# value that means "whatever the user has declared elsewhere".
# This might become obsolete with https://github.com/NixOS/nixpkgs/issues/162398
shorthandOnlyDefinesConfig = null;
};
};
};
resultsByName = mapAttrs (name: decls:
# We're descending into attribute name.
let
2013-10-28 13:25:58 +00:00
loc = prefix ++ [name];
defns = pushedDownDefinitionsByName.${name} or [];
defns' = rawDefinitionsByName.${name} or [];
optionDecls = filter
(m: m.options?_type
&& (m.options._type == "option"
|| throwDeclarationTypeError loc m.options._type m._file
)
)
decls;
in
if length optionDecls == length decls then
let opt = fixupOptionType loc (mergeOptionDecls loc decls);
in {
matchedOptions = evalOptionValue loc opt defns';
unmatchedDefns = [];
}
else if optionDecls != [] then
if all (x: x.options.type.name or null == "submodule") optionDecls
# Raw options can only be merged into submodules. Merging into
# attrsets might be nice, but ambiguous. Suppose we have
# attrset as a `attrsOf submodule`. User declares option
# attrset.foo.bar, this could mean:
# a. option `bar` is only available in `attrset.foo`
# b. option `foo.bar` is available in all `attrset.*`
# c. reject and require "<name>" as a reminder that it behaves like (b).
# d. magically combine (a) and (c).
# All of the above are merely syntax sugar though.
then
let opt = fixupOptionType loc (mergeOptionDecls loc (map optionTreeToOption decls));
in {
matchedOptions = evalOptionValue loc opt defns';
unmatchedDefns = [];
}
else
let
nonOptions = filter (m: !isOption m.options) decls;
in
throw "The option `${showOption loc}' in module `${(head optionDecls)._file}' would be a parent of the following options, but its type `${(head optionDecls).options.type.description or "<no description>"}' does not support nested options.\n${
showRawDecls loc nonOptions
}"
else
mergeModules' loc decls defns) declsByName;
matchedOptions = mapAttrs (n: v: v.matchedOptions) resultsByName;
# an attrset 'name' => list of unmatched definitions for 'name'
unmatchedDefnsByName =
# Propagate all unmatched definitions from nested option sets
mapAttrs (n: v: v.unmatchedDefns) resultsByName
# Plus the definitions for the current prefix that don't have a matching option
// removeAttrs rawDefinitionsByName (attrNames matchedOptions);
in {
inherit matchedOptions;
# Transforms unmatchedDefnsByName into a list of definitions
unmatchedDefns =
if configs == []
then
# When no config values exist, there can be no unmatched config, so
# we short circuit and avoid evaluating more _options_ than necessary.
[]
else
concatLists (mapAttrsToList (name: defs:
map (def: def // {
# Set this so we know when the definition first left unmatched territory
prefix = [name] ++ (def.prefix or []);
}) defs
) unmatchedDefnsByName);
};
throwDeclarationTypeError = loc: actualTag: file:
let
name = lib.strings.escapeNixIdentifier (lib.lists.last loc);
path = showOption loc;
depth = length loc;
paragraphs = [
"In module ${file}: expected an option declaration at option path `${path}` but got an attribute set with type ${actualTag}"
] ++ optional (actualTag == "option-type") ''
When declaring an option, you must wrap the type in a `mkOption` call. It should look somewhat like:
${comment}
${name} = lib.mkOption {
description = ...;
type = <the type you wrote for ${name}>;
...
};
'';
# Ideally we'd know the exact syntax they used, but short of that,
# we can only reliably repeat the last. However, we repeat the
# full path in a non-misleading way here, in case they overlook
# the start of the message. Examples attract attention.
comment = optionalString (depth > 1) "\n # ${showOption loc}";
in
throw (concatStringsSep "\n\n" paragraphs);
/* Merge multiple option declarations into a single declaration. In
general, there should be only one declaration of each option.
The exception is the options attribute, which specifies
sub-options. These can be specified multiple times to allow one
module to add sub-options to an option declared somewhere else
(e.g. multiple modules define sub-options for fileSystems).
'loc' is the list of attribute names where the option is located.
'opts' is a list of modules. Each module has an options attribute which
correspond to the definition of 'loc' in 'opt.file'. */
mergeOptionDecls =
loc: opts:
foldl' (res: opt:
2016-09-07 01:03:32 +00:00
let t = res.type;
t' = opt.options.type;
mergedType = t.typeMerge t'.functor;
typesMergeable = mergedType != null;
typeSet = if (bothHave "type") && typesMergeable
then { type = mergedType; }
else {};
bothHave = k: opt.options ? ${k} && res ? ${k};
in
if bothHave "default" ||
bothHave "example" ||
bothHave "description" ||
bothHave "apply" ||
(bothHave "type" && (! typesMergeable))
then
throw "The option `${showOption loc}' in `${opt._file}' is already declared in ${showFiles res.declarations}."
else
let
getSubModules = opt.options.type.getSubModules or null;
submodules =
if getSubModules != null then map (setDefaultModuleLocation opt._file) getSubModules ++ res.options
else res.options;
in opt.options // res //
{ declarations = res.declarations ++ [opt._file];
nixos/modules: Add declarationPositions What it does: line and column level *declaration* position information: $ nix repl . nix-repl> :p nixosConfigurations.micro.options.environment.systemPackages.declarationPositions [ { column = 7; file = "/nix/store/24aj3k7fgqv3ly7qkbf98qvphasrw9nb-source/nixos/modules/config/system-path.nix"; line = 63; } ] Use cases: - ctags over NixOS options, as will be presented at NixCon 2023 ;) - improving the documentation pages to go to the exact line of the declarations. Related work: - https://github.com/NixOS/nixpkgs/pull/65024 This one does it for all *definitions* rather than declarations, and it was not followed through with due to performance worries. - https://github.com/NixOS/nixpkgs/pull/208173 The basis for this change. This change is just a rebase of that one. I split it out to add the capability before adding users of it, in order to simplify review. However, the ctags script in there is a sample user of this feature. Benchmarks: conducted by evaluating my own reasonably complex NixOS configuration with the command: `hyperfine -S none -w 1 -- "nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath"` ``` Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.971 s ± 0.254 s [User: 5.872 s, System: 1.388 s] Range (min … max): 8.574 s … 9.327 s 10 runs Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.766 s ± 0.160 s [User: 5.873 s, System: 1.346 s] Range (min … max): 8.496 s … 9.033 s 10 runs ``` Summary of results: it seems to be in the noise, this does not cause any visible regression in times.
2022-12-28 22:39:05 +00:00
# In the case of modules that are generated dynamically, we won't
# have exact declaration lines; fall back to just the file being
# evaluated.
declarationPositions = res.declarationPositions
++ (if opt.pos != null
then [opt.pos]
else [{ file = opt._file; line = null; column = null; }]);
options = submodules;
2016-09-07 01:03:32 +00:00
} // typeSet
nixos/modules: Add declarationPositions What it does: line and column level *declaration* position information: $ nix repl . nix-repl> :p nixosConfigurations.micro.options.environment.systemPackages.declarationPositions [ { column = 7; file = "/nix/store/24aj3k7fgqv3ly7qkbf98qvphasrw9nb-source/nixos/modules/config/system-path.nix"; line = 63; } ] Use cases: - ctags over NixOS options, as will be presented at NixCon 2023 ;) - improving the documentation pages to go to the exact line of the declarations. Related work: - https://github.com/NixOS/nixpkgs/pull/65024 This one does it for all *definitions* rather than declarations, and it was not followed through with due to performance worries. - https://github.com/NixOS/nixpkgs/pull/208173 The basis for this change. This change is just a rebase of that one. I split it out to add the capability before adding users of it, in order to simplify review. However, the ctags script in there is a sample user of this feature. Benchmarks: conducted by evaluating my own reasonably complex NixOS configuration with the command: `hyperfine -S none -w 1 -- "nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath"` ``` Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.971 s ± 0.254 s [User: 5.872 s, System: 1.388 s] Range (min … max): 8.574 s … 9.327 s 10 runs Benchmark 1: nix eval .#nixosConfigurations.snowflake.config.system.build.toplevel.outPath Time (mean ± σ): 8.766 s ± 0.160 s [User: 5.873 s, System: 1.346 s] Range (min … max): 8.496 s … 9.033 s 10 runs ``` Summary of results: it seems to be in the noise, this does not cause any visible regression in times.
2022-12-28 22:39:05 +00:00
) { inherit loc; declarations = []; declarationPositions = []; options = []; } opts;
/* Merge all the definitions of an option to produce the final
config value. */
evalOptionValue = loc: opt: defs:
let
# Add in the default value for this option, if any.
defs' =
(optional (opt ? default)
{ file = head opt.declarations; value = mkOptionDefault opt.default; }) ++ defs;
# Handle properties, check types, and merge everything together.
res =
if opt.readOnly or false && length defs' > 1 then
let
# For a better error message, evaluate all readOnly definitions as
# if they were the only definition.
separateDefs = map (def: def // {
value = (mergeDefinitions loc opt.type [ def ]).mergedValue;
}) defs';
in throw "The option `${showOption loc}' is read-only, but it's set multiple times. Definition values:${showDefs separateDefs}"
else
mergeDefinitions loc opt.type defs';
# Apply the 'apply' function to the merged value. This allows options to
# yield a value computed from the definitions
value = if opt ? apply then opt.apply res.mergedValue else res.mergedValue;
warnDeprecation =
warnIf (opt.type.deprecationMessage != null)
"The type `types.${opt.type.name}' of option `${showOption loc}' defined in ${showFiles opt.declarations} is deprecated. ${opt.type.deprecationMessage}";
in warnDeprecation opt //
{ value = addErrorContext "while evaluating the option `${showOption loc}':" value;
inherit (res.defsFinal') highestPrio;
definitions = map (def: def.value) res.defsFinal;
files = map (def: def.file) res.defsFinal;
definitionsWithLocations = res.defsFinal;
inherit (res) isDefined;
# This allows options to be correctly displayed using `${options.path.to.it}`
__toString = _: showOption loc;
};
2015-07-28 12:29:29 +00:00
# Merge definitions of a value of a given type.
mergeDefinitions = loc: type: defs: rec {
defsFinal' =
2015-07-28 12:29:29 +00:00
let
# Process mkMerge and mkIf properties.
defs' = concatMap (m:
map (value: { inherit (m) file; inherit value; }) (addErrorContext "while evaluating definitions from `${m.file}':" (dischargeProperties m.value))
2015-07-28 12:29:29 +00:00
) defs;
# Process mkOverride properties.
defs'' = filterOverrides' defs';
2015-07-28 12:29:29 +00:00
# Sort mkOrder properties.
defs''' =
2015-07-28 12:29:29 +00:00
# Avoid sorting if we don't have to.
if any (def: def.value._type or "" == "order") defs''.values
then sortProperties defs''.values
else defs''.values;
in {
values = defs''';
inherit (defs'') highestPrio;
};
defsFinal = defsFinal'.values;
2015-07-28 12:29:29 +00:00
# Type-check the remaining definitions, and merge them. Or throw if no definitions.
mergedValue =
if isDefined then
if all (def: type.check def.value) defsFinal then type.merge loc defsFinal
else let allInvalid = filter (def: ! type.check def.value) defsFinal;
in throw "A definition for option `${showOption loc}' is not of type `${type.description}'. Definition values:${showDefs allInvalid}"
else
# (nixos-option detects this specific error message and gives it special
# handling. If changed here, please change it there too.)
throw "The option `${showOption loc}' was accessed but has no value defined. Try setting the option.";
2015-07-28 12:29:29 +00:00
isDefined = defsFinal != [];
optionalValue =
if isDefined then { value = mergedValue; }
else {};
};
/* Given a config set, expand mkMerge properties, and push down the
2014-03-30 18:35:25 +00:00
other properties into the children. The result is a list of
config sets that do not have properties at top-level. For
example,
mkMerge [ { boot = set1; } (mkIf cond { boot = set2; services = set3; }) ]
is transformed into
[ { boot = set1; } { boot = mkIf cond set2; services = mkIf cond set3; } ].
This transform is the critical step that allows mkIf conditions
to refer to the full configuration without creating an infinite
recursion.
*/
pushDownProperties = cfg:
if cfg._type or "" == "merge" then
concatMap pushDownProperties cfg.contents
else if cfg._type or "" == "if" then
map (mapAttrs (n: v: mkIf cfg.condition v)) (pushDownProperties cfg.content)
2013-10-28 16:46:45 +00:00
else if cfg._type or "" == "override" then
map (mapAttrs (n: v: mkOverride cfg.priority v)) (pushDownProperties cfg.content)
2014-03-30 18:35:25 +00:00
else # FIXME: handle mkOrder?
[ cfg ];
/* Given a config value, expand mkMerge properties, and discharge
any mkIf conditions. That is, this is the place where mkIf
conditions are actually evaluated. The result is a list of
config values. For example, mkIf false x yields [],
mkIf true x yields [x], and
mkMerge [ 1 (mkIf true 2) (mkIf true (mkIf false 3)) ]
yields [ 1 2 ].
*/
dischargeProperties = def:
if def._type or "" == "merge" then
concatMap dischargeProperties def.contents
else if def._type or "" == "if" then
if isBool def.condition then
if def.condition then
dischargeProperties def.content
else
[ ]
else
throw "mkIf called with a non-Boolean condition"
else
[ def ];
2013-10-28 06:52:24 +00:00
/* Given a list of config values, process the mkOverride properties,
that is, return the values that have the highest (that is,
numerically lowest) priority, and strip the mkOverride
properties. For example,
2013-10-28 06:52:24 +00:00
[ { file = "/1"; value = mkOverride 10 "a"; }
{ file = "/2"; value = mkOverride 20 "b"; }
{ file = "/3"; value = "z"; }
{ file = "/4"; value = mkOverride 10 "d"; }
]
yields
2013-10-28 06:52:24 +00:00
[ { file = "/1"; value = "a"; }
{ file = "/4"; value = "d"; }
]
Note that "z" has the default priority 100.
*/
filterOverrides = defs: (filterOverrides' defs).values;
filterOverrides' = defs:
let
getPrio = def: if def.value._type or "" == "override" then def.value.priority else defaultOverridePriority;
highestPrio = foldl' (prio: def: min (getPrio def) prio) 9999 defs;
2013-10-28 04:23:10 +00:00
strip = def: if def.value._type or "" == "override" then def // { value = def.value.content; } else def;
in {
values = concatMap (def: if getPrio def == highestPrio then [(strip def)] else []) defs;
inherit highestPrio;
};
2014-03-30 18:35:25 +00:00
/* Sort a list of properties. The sort priority of a property is
defaultOrderPriority by default, but can be overridden by wrapping the property
2014-03-30 18:35:25 +00:00
using mkOrder. */
sortProperties = defs:
let
strip = def:
if def.value._type or "" == "order"
then def // { value = def.value.content; inherit (def.value) priority; }
else def;
defs' = map strip defs;
compare = a: b: (a.priority or defaultOrderPriority) < (b.priority or defaultOrderPriority);
2014-03-30 18:35:25 +00:00
in sort compare defs';
# This calls substSubModules, whose entire purpose is only to ensure that
# option declarations in submodules have accurate position information.
# TODO: Merge this into mergeOptionDecls
fixupOptionType = loc: opt:
if opt.type.getSubModules or null == null
then opt // { type = opt.type or types.unspecified; }
else opt // { type = opt.type.substSubModules opt.options; options = []; };
/*
Merge an option's definitions in a way that preserves the priority of the
individual attributes in the option value.
This does not account for all option semantics, such as readOnly.
Type:
option -> attrsOf { highestPrio, value }
*/
mergeAttrDefinitionsWithPrio = opt:
let
defsByAttr =
zipAttrs (
concatLists (
concatMap
({ value, ... }@def:
map
(mapAttrsToList (k: value: { ${k} = def // { inherit value; }; }))
(pushDownProperties value)
)
opt.definitionsWithLocations
)
);
in
assert opt.type.name == "attrsOf" || opt.type.name == "lazyAttrsOf";
mapAttrs
(k: v:
let merging = mergeDefinitions (opt.loc ++ [k]) opt.type.nestedTypes.elemType v;
in {
value = merging.mergedValue;
inherit (merging.defsFinal') highestPrio;
})
defsByAttr;
/* Properties. */
mkIf = condition: content:
{ _type = "if";
inherit condition content;
};
mkAssert = assertion: message: content:
mkIf
(if assertion then true else throw "\nFailed assertion: ${message}")
content;
mkMerge = contents:
{ _type = "merge";
inherit contents;
};
mkOverride = priority: content:
{ _type = "override";
inherit priority content;
};
mkOptionDefault = mkOverride 1500; # priority of option defaults
mkDefault = mkOverride 1000; # used in config sections of non-user modules to set a default
defaultOverridePriority = 100;
mkImageMediaOverride = mkOverride 60; # image media profiles can be derived by inclusion into host config, hence needing to override host config, but do allow user to mkForce
mkForce = mkOverride 50;
mkVMOverride = mkOverride 10; # used by nixos-rebuild build-vm
defaultPriority = warnIf (isInOldestRelease 2305) "lib.modules.defaultPriority is deprecated, please use lib.modules.defaultOverridePriority instead." defaultOverridePriority;
mkFixStrictness = warn "lib.mkFixStrictness has no effect and will be removed. It returns its argument unmodified, so you can just remove any calls." id;
2014-03-30 18:35:25 +00:00
mkOrder = priority: content:
{ _type = "order";
inherit priority content;
};
mkBefore = mkOrder 500;
defaultOrderPriority = 1000;
2014-03-30 18:35:25 +00:00
mkAfter = mkOrder 1500;
# Convenient property used to transfer all definitions and their
# properties from one option to another. This property is useful for
# renaming options, and also for including properties from another module
# system, including sub-modules.
#
# { config, options, ... }:
#
# {
# # 'bar' might not always be defined in the current module-set.
# config.foo.enable = mkAliasDefinitions (options.bar.enable or {});
#
# # 'barbaz' has to be defined in the current module-set.
# config.foobar.paths = mkAliasDefinitions options.barbaz.paths;
# }
#
# Note, this is different than taking the value of the option and using it
# as a definition, as the new definition will not keep the mkOverride /
# mkDefault properties of the previous option.
#
mkAliasDefinitions = mkAliasAndWrapDefinitions id;
mkAliasAndWrapDefinitions = wrap: option:
mkAliasIfDef option (wrap (mkMerge option.definitions));
# Similar to mkAliasAndWrapDefinitions but copies over the priority from the
# option as well.
#
# If a priority is not set, it assumes a priority of defaultOverridePriority.
mkAliasAndWrapDefsWithPriority = wrap: option:
let
prio = option.highestPrio or defaultOverridePriority;
defsWithPrio = map (mkOverride prio) option.definitions;
in mkAliasIfDef option (wrap (mkMerge defsWithPrio));
mkAliasIfDef = option:
mkIf (isOption option && option.isDefined);
/* Compatibility. */
fixMergeModules = modules: args: evalModules { inherit modules args; check = false; };
/* Return a module that causes a warning to be shown if the
specified option is defined. For example,
mkRemovedOptionModule [ "boot" "loader" "grub" "bootDevice" ] "<replacement instructions>"
causes a assertion if the user defines boot.loader.grub.bootDevice.
replacementInstructions is a string that provides instructions on
how to achieve the same functionality without the removed option,
or alternatively a reasoning why the functionality is not needed.
replacementInstructions SHOULD be provided!
*/
mkRemovedOptionModule = optionName: replacementInstructions:
{ options, ... }:
{ options = setAttrByPath optionName (mkOption {
visible = false;
apply = x: throw "The option `${showOption optionName}' can no longer be used since it's been removed. ${replacementInstructions}";
});
config.assertions =
let opt = getAttrFromPath optionName options; in [{
assertion = !opt.isDefined;
message = ''
The option definition `${showOption optionName}' in ${showFiles opt.files} no longer has any effect; please remove it.
${replacementInstructions}
'';
}];
};
/* Return a module that causes a warning to be shown if the
specified "from" option is defined; the defined value is however
forwarded to the "to" option. This can be used to rename options
while providing backward compatibility. For example,
mkRenamedOptionModule [ "boot" "copyKernels" ] [ "boot" "loader" "grub" "copyKernels" ]
forwards any definitions of boot.copyKernels to
boot.loader.grub.copyKernels while printing a warning.
This also copies over the priority from the aliased option to the
non-aliased option.
*/
mkRenamedOptionModule = from: to: doRename {
inherit from to;
visible = false;
warn = true;
use = trace "Obsolete option `${showOption from}' is used. It was renamed to `${showOption to}'.";
};
mkRenamedOptionModuleWith = {
/* Old option path as list of strings. */
from,
/* New option path as list of strings. */
to,
/*
Release number of the first release that contains the rename, ignoring backports.
Set it to the upcoming release, matching the nixpkgs/.version file.
*/
sinceRelease,
}: doRename {
inherit from to;
visible = false;
warn = isInOldestRelease sinceRelease;
use = warnIf (isInOldestRelease sinceRelease)
"Obsolete option `${showOption from}' is used. It was renamed to `${showOption to}'.";
};
/* Return a module that causes a warning to be shown if any of the "from"
option is defined; the defined values can be used in the "mergeFn" to set
the "to" value.
This function can be used to merge multiple options into one that has a
different type.
"mergeFn" takes the module "config" as a parameter and must return a value
of "to" option type.
mkMergedOptionModule
[ [ "a" "b" "c" ]
[ "d" "e" "f" ] ]
[ "x" "y" "z" ]
(config:
let value = p: getAttrFromPath p config;
in
if (value [ "a" "b" "c" ]) == true then "foo"
else if (value [ "d" "e" "f" ]) == true then "bar"
else "baz")
- options.a.b.c is a removed boolean option
- options.d.e.f is a removed boolean option
- options.x.y.z is a new str option that combines a.b.c and d.e.f
functionality
This show a warning if any a.b.c or d.e.f is set, and set the value of
x.y.z to the result of the merge function
*/
mkMergedOptionModule = from: to: mergeFn:
{ config, options, ... }:
{
options = foldl' recursiveUpdate {} (map (path: setAttrByPath path (mkOption {
visible = false;
# To use the value in mergeFn without triggering errors
default = "_mkMergedOptionModule";
})) from);
config = {
warnings = filter (x: x != "") (map (f:
let val = getAttrFromPath f config;
opt = getAttrFromPath f options;
in
optionalString
(val != "_mkMergedOptionModule")
"The option `${showOption f}' defined in ${showFiles opt.files} has been changed to `${showOption to}' that has a different type. Please read `${showOption to}' documentation and update your configuration accordingly."
) from);
} // setAttrByPath to (mkMerge
(optional
(any (f: (getAttrFromPath f config) != "_mkMergedOptionModule") from)
(mergeFn config)));
};
/* Single "from" version of mkMergedOptionModule.
Return a module that causes a warning to be shown if the "from" option is
defined; the defined value can be used in the "mergeFn" to set the "to"
value.
This function can be used to change an option into another that has a
different type.
"mergeFn" takes the module "config" as a parameter and must return a value of
"to" option type.
mkChangedOptionModule [ "a" "b" "c" ] [ "x" "y" "z" ]
(config:
let value = getAttrFromPath [ "a" "b" "c" ] config;
in
if value > 100 then "high"
else "normal")
- options.a.b.c is a removed int option
- options.x.y.z is a new str option that supersedes a.b.c
This show a warning if a.b.c is set, and set the value of x.y.z to the
result of the change function
*/
mkChangedOptionModule = from: to: changeFn:
mkMergedOptionModule [ from ] to changeFn;
/* Like mkRenamedOptionModule, but doesn't show a warning. */
mkAliasOptionModule = from: to: doRename {
inherit from to;
visible = true;
warn = false;
use = id;
};
/* Transitional version of mkAliasOptionModule that uses MD docs.
This function is no longer necessary and merely an alias of `mkAliasOptionModule`.
*/
mkAliasOptionModuleMD = mkAliasOptionModule;
/* mkDerivedConfig : Option a -> (a -> Definition b) -> Definition b
Create config definitions with the same priority as the definition of another option.
This should be used for option definitions where one option sets the value of another as a convenience.
For instance a config file could be set with a `text` or `source` option, where text translates to a `source`
value using `mkDerivedConfig options.text (pkgs.writeText "filename.conf")`.
It takes care of setting the right priority using `mkOverride`.
*/
# TODO: make the module system error message include information about `opt` in
# error messages about conflicts. E.g. introduce a variation of `mkOverride` which
# adds extra location context to the definition object. This will allow context to be added
# to all messages that report option locations "this value was derived from <full option name>
# which was defined in <locations>". It can provide a trace of options that contributed
# to definitions.
mkDerivedConfig = opt: f:
mkOverride
(opt.highestPrio or defaultOverridePriority)
(f opt.value);
/*
Return a module that help declares an option that has been renamed.
When a value is defined for the old option, it is forwarded to the `to` option.
*/
doRename = {
# List of strings representing the attribute path of the old option.
from,
# List of strings representing the attribute path of the new option.
to,
# Boolean, whether the old option is to be included in documentation.
visible,
# Whether to warn when a value is defined for the old option.
# NOTE: This requires the NixOS assertions module to be imported, so
# - this generally does not work in submodules
# - this may or may not work outside NixOS
warn,
# A function that is applied to the option value, to form the value
# of the old `from` option.
#
# For example, the identity function can be passed, to return the option value unchanged.
# ```nix
# use = x: x;
# ```
#
# To add a warning, you can pass the partially applied `warn` function.
# ```nix
# use = lib.warn "Obsolete option `${opt.old}' is used. Use `${opt.to}' instead.";
# ```
use,
# Legacy option, enabled by default: whether to preserve the priority of definitions in `old`.
withPriority ? true,
# A boolean that defines the `mkIf` condition for `to`.
# If the condition evaluates to `true`, and the `to` path points into an
# `attrsOf (submodule ...)`, then `doRename` would cause an empty module to
# be created, even if the `from` option is undefined.
# By setting this to an expression that may return `false`, you can inhibit
# this undesired behavior.
#
# Example:
#
# ```nix
# { config, lib, ... }:
# let
# inherit (lib) mkOption mkEnableOption types doRename;
# in
# {
# options = {
#
# # Old service
# services.foo.enable = mkEnableOption "foo";
#
# # New multi-instance service
# services.foos = mkOption {
# type = types.attrsOf (types.submodule …);
# };
# };
# imports = [
# (doRename {
# from = [ "services" "foo" "bar" ];
# to = [ "services" "foos" "" "bar" ];
# visible = true;
# warn = false;
# use = x: x;
# withPriority = true;
# # Only define services.foos."" if needed. (It's not just about `bar`)
# condition = config.services.foo.enable;
# })
# ];
# }
# ```
condition ? true
}:
{ config, options, ... }:
let
fromOpt = getAttrFromPath from options;
toOf = attrByPath to
2018-03-27 23:02:40 +00:00
(abort "Renaming error: option `${showOption to}' does not exist.");
toType = let opt = attrByPath to {} options; in opt.type or (types.submodule {});
in
{
options = setAttrByPath from (mkOption {
inherit visible;
description = "Alias of {option}`${showOption to}`.";
apply = x: use (toOf config);
} // optionalAttrs (toType != null) {
type = toType;
});
config = mkIf condition (mkMerge [
(optionalAttrs (options ? warnings) {
warnings = optional (warn && fromOpt.isDefined)
"The option `${showOption from}' defined in ${showFiles fromOpt.files} has been renamed to `${showOption to}'.";
})
(if withPriority
then mkAliasAndWrapDefsWithPriority (setAttrByPath to) fromOpt
else mkAliasAndWrapDefinitions (setAttrByPath to) fromOpt)
]);
};
/**
`importApply file arg :: Path -> a -> Module`, where `import file :: a -> Module`
`importApply` imports a Nix expression file much like the module system would,
after passing an extra positional argument to the function in the file.
This function should be used when declaring a module in a file that refers to
values from a different scope, such as that in a flake.
It solves the problems of alternative solutions:
- While `importApply file arg` is _mostly_ equivalent to
`import file arg`, the latter returns a module without a location,
as `import` only returns the contained expression. This leads to worse
error messages.
- Using `specialArgs` to provide arguments to all modules. This effectively
creates an incomplete module, and requires the user of the module to
manually pass the `specialArgs` to the configuration, which is error-prone,
verbose, and unnecessary.
The nix file must contain a function that returns a module.
A module may itself be a function, so the file is often a function with two
positional arguments instead of one. See the example below.
This function does not add support for deduplication and `disabledModules`,
although that could be achieved by wrapping the returned module and setting
the `_key` module attribute.
The reason for this omission is that the file path is not guaranteed to be
a unique identifier for the module, as two instances of the module may
reference different `arg`s in their closures.
Example
# lib.nix
imports = [
(lib.modules.importApply ./module.nix { bar = bar; })
];
# module.nix
{ bar }:
{ lib, config, ... }:
{
options = ...;
config = ... bar ...;
}
*/
importApply =
modulePath: staticArg:
lib.setDefaultModuleLocation modulePath (import modulePath staticArg);
/* Use this function to import a JSON file as NixOS configuration.
modules.importJSON :: path -> attrs
*/
importJSON = file: {
_file = file;
config = lib.importJSON file;
};
/* Use this function to import a TOML file as NixOS configuration.
modules.importTOML :: path -> attrs
*/
importTOML = file: {
_file = file;
config = lib.importTOML file;
};
2023-02-13 09:51:50 +00:00
private = mapAttrs
(k: warn "External use of `lib.modules.${k}` is deprecated. If your use case isn't covered by non-deprecated functions, we'd like to know more and perhaps support your use case well, instead of providing access to these low level functions. In this case please open an issue in https://github.com/nixos/nixpkgs/issues/.")
{
inherit
applyModuleArgsIfFunction
dischargeProperties
mergeModules
mergeModules'
pushDownProperties
unifyModuleSyntax
;
collectModules = collectModules null;
};
2023-02-13 09:51:50 +00:00
in
private //
2023-02-13 09:51:50 +00:00
{
# NOTE: not all of these functions are necessarily public interfaces; some
# are just needed by types.nix, but are not meant to be consumed
# externally.
2023-02-13 09:51:50 +00:00
inherit
defaultOrderPriority
defaultOverridePriority
defaultPriority
doRename
evalModules
evalOptionValue # for use by lib.types
2023-02-13 09:51:50 +00:00
filterOverrides
filterOverrides'
fixMergeModules
fixupOptionType # should be private?
importApply
2023-02-13 09:51:50 +00:00
importJSON
importTOML
mergeDefinitions
mergeAttrDefinitionsWithPrio
mergeOptionDecls # should be private?
2023-02-13 09:51:50 +00:00
mkAfter
mkAliasAndWrapDefinitions
mkAliasAndWrapDefsWithPriority
mkAliasDefinitions
mkAliasIfDef
mkAliasOptionModule
mkAliasOptionModuleMD
mkAssert
mkBefore
mkChangedOptionModule
mkDefault
mkDerivedConfig
mkFixStrictness
mkForce
mkIf
mkImageMediaOverride
mkMerge
mkMergedOptionModule
mkOptionDefault
mkOrder
mkOverride
mkRemovedOptionModule
mkRenamedOptionModule
mkRenamedOptionModuleWith
mkVMOverride
setDefaultModuleLocation
sortProperties;
}