nixpkgs/lib/path/default.nix

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

642 lines
22 KiB
Nix
Raw Normal View History

/* Functions for working with path values. */
# See ./README.md for internal docs
{ lib }:
let
inherit (builtins)
isString
isPath
2022-12-23 21:07:30 +01:00
split
match
2023-04-05 20:31:50 +02:00
typeOf
storeDir
;
2022-12-23 21:07:30 +01:00
inherit (lib.lists)
length
head
last
genList
elemAt
all
concatMap
foldl'
2023-04-05 20:31:50 +02:00
take
2023-06-16 00:19:33 +02:00
drop
2022-12-23 21:07:30 +01:00
;
listHasPrefix = lib.lists.hasPrefix;
inherit (lib.strings)
2022-12-23 21:07:30 +01:00
concatStringsSep
substring
;
inherit (lib.asserts)
assertMsg
;
inherit (lib.path.subpath)
isValid
;
# Return the reason why a subpath is invalid, or `null` if it's valid
subpathInvalidReason = value:
if ! isString value then
"The given value is of type ${builtins.typeOf value}, but a string was expected"
else if value == "" then
"The given string is empty"
else if substring 0 1 value == "/" then
"The given string \"${value}\" starts with a `/`, representing an absolute path"
# We don't support ".." components, see ./path.md#parent-directory
else if match "(.*/)?\\.\\.(/.*)?" value != null then
"The given string \"${value}\" contains a `..` component, which is not allowed in subpaths"
else null;
2022-12-23 21:07:30 +01:00
# Split and normalise a relative path string into its components.
# Error for ".." components and doesn't include "." components
splitRelPath = path:
let
# Split the string into its parts using regex for efficiency. This regex
# matches patterns like "/", "/./", "/././", with arbitrarily many "/"s
# together. These are the main special cases:
# - Leading "./" gets split into a leading "." part
# - Trailing "/." or "/" get split into a trailing "." or ""
# part respectively
#
# These are the only cases where "." and "" parts can occur
parts = split "/+(\\./+)*" path;
# `split` creates a list of 2 * k + 1 elements, containing the k +
# 1 parts, interleaved with k matches where k is the number of
# (non-overlapping) matches. This calculation here gets the number of parts
# back from the list length
# floor( (2 * k + 1) / 2 ) + 1 == floor( k + 1/2 ) + 1 == k + 1
partCount = length parts / 2 + 1;
# To assemble the final list of components we want to:
# - Skip a potential leading ".", normalising "./foo" to "foo"
# - Skip a potential trailing "." or "", normalising "foo/" and "foo/." to
# "foo". See ./path.md#trailing-slashes
skipStart = if head parts == "." then 1 else 0;
skipEnd = if last parts == "." || last parts == "" then 1 else 0;
# We can now know the length of the result by removing the number of
# skipped parts from the total number
componentCount = partCount - skipEnd - skipStart;
in
# Special case of a single "." path component. Such a case leaves a
# componentCount of -1 due to the skipStart/skipEnd not verifying that
# they don't refer to the same character
if path == "." then []
# Generate the result list directly. This is more efficient than a
# combination of `filter`, `init` and `tail`, because here we don't
# allocate any intermediate lists
else genList (index:
# To get to the element we need to add the number of parts we skip and
# multiply by two due to the interleaved layout of `parts`
elemAt parts ((skipStart + index) * 2)
) componentCount;
# Join relative path components together
joinRelPath = components:
# Always return relative paths with `./` as a prefix (./path.md#leading-dots-for-relative-paths)
"./" +
# An empty string is not a valid relative path, so we need to return a `.` when we have no components
(if components == [] then "." else concatStringsSep "/" components);
2023-04-05 20:31:50 +02:00
# Type: Path -> { root :: Path, components :: [ String ] }
#
# Deconstruct a path value type into:
# - root: The filesystem root of the path, generally `/`
# - components: All the path's components
#
# This is similar to `splitString "/" (toString path)` but safer
# because it can distinguish different filesystem roots
deconstructPath =
let
recurse = components: base:
# If the parent of a path is the path itself, then it's a filesystem root
if base == dirOf base then { root = base; inherit components; }
else recurse ([ (baseNameOf base) ] ++ components) (dirOf base);
in recurse [];
# The components of the store directory, typically [ "nix" "store" ]
storeDirComponents = splitRelPath ("./" + storeDir);
# The number of store directory components, typically 2
storeDirLength = length storeDirComponents;
# Type: [ String ] -> Bool
#
# Whether path components have a store path as a prefix, according to
# https://nixos.org/manual/nix/stable/store/store-path.html#store-path.
componentsHaveStorePathPrefix = components:
# path starts with the store directory (typically /nix/store)
listHasPrefix storeDirComponents components
# is not the store directory itself, meaning there's at least one extra component
&& storeDirComponents != components
# and the first component after the store directory has the expected format.
# NOTE: We could change the hash regex to be [0-9a-df-np-sv-z],
# because these are the actual ASCII characters used by Nix's base32 implementation,
# but this is not fully specified, so let's tie this too much to the currently implemented concept of store paths.
# Similar reasoning applies to the validity of the name part.
# We care more about discerning store path-ness on realistic values. Making it airtight would be fragile and slow.
&& match ".{32}-.+" (elemAt components storeDirLength) != null;
in /* No rec! Add dependencies on this file at the top. */ {
2023-08-08 00:09:08 +02:00
/*
Append a subpath string to a path.
Like `path + ("/" + string)` but safer, because it errors instead of returning potentially surprising results.
More specifically, it checks that the first argument is a [path value type](https://nixos.org/manual/nix/stable/language/values.html#type-path"),
and that the second argument is a [valid subpath string](#function-library-lib.path.subpath.isValid).
Laws:
- Not influenced by subpath [normalisation](#function-library-lib.path.subpath.normalise):
append p s == append p (subpath.normalise s)
Type:
append :: Path -> String -> Path
Example:
append /foo "bar/baz"
=> /foo/bar/baz
# subpaths don't need to be normalised
append /foo "./bar//baz/./"
=> /foo/bar/baz
# can append to root directory
append /. "foo/bar"
=> /foo/bar
# first argument needs to be a path value type
append "/foo" "bar"
=> <error>
# second argument needs to be a valid subpath string
append /foo /bar
=> <error>
append /foo ""
=> <error>
append /foo "/bar"
=> <error>
append /foo "../bar"
=> <error>
*/
append =
# The absolute path to append to
path:
# The subpath string to append
subpath:
assert assertMsg (isPath path) ''
lib.path.append: The first argument is of type ${builtins.typeOf path}, but a path was expected'';
assert assertMsg (isValid subpath) ''
lib.path.append: Second argument is not a valid subpath string:
${subpathInvalidReason subpath}'';
path + ("/" + subpath);
2023-04-05 20:31:50 +02:00
/*
2023-08-08 00:09:08 +02:00
Whether the first path is a component-wise prefix of the second path.
2023-04-05 20:31:50 +02:00
2023-08-08 00:09:08 +02:00
Laws:
2023-04-05 20:31:50 +02:00
2023-08-08 00:09:08 +02:00
- `hasPrefix p q` is only true if [`q == append p s`](#function-library-lib.path.append) for some [subpath](#function-library-lib.path.subpath.isValid) `s`.
2023-04-05 20:31:50 +02:00
2023-08-08 00:09:08 +02:00
- `hasPrefix` is a [non-strict partial order](https://en.wikipedia.org/wiki/Partially_ordered_set#Non-strict_partial_order) over the set of all path values.
2023-04-05 20:31:50 +02:00
2023-08-08 00:09:08 +02:00
Type:
hasPrefix :: Path -> Path -> Bool
2023-04-05 20:31:50 +02:00
2023-08-08 00:09:08 +02:00
Example:
hasPrefix /foo /foo/bar
=> true
hasPrefix /foo /foo
=> true
hasPrefix /foo/bar /foo
=> false
hasPrefix /. /foo
=> true
2023-04-05 20:31:50 +02:00
*/
hasPrefix =
path1:
assert assertMsg
(isPath path1)
"lib.path.hasPrefix: First argument is of type ${typeOf path1}, but a path was expected";
let
path1Deconstructed = deconstructPath path1;
in
path2:
assert assertMsg
(isPath path2)
"lib.path.hasPrefix: Second argument is of type ${typeOf path2}, but a path was expected";
let
path2Deconstructed = deconstructPath path2;
in
assert assertMsg
(path1Deconstructed.root == path2Deconstructed.root) ''
lib.path.hasPrefix: Filesystem roots must be the same for both paths, but paths with different roots were given:
first argument: "${toString path1}" with root "${toString path1Deconstructed.root}"
second argument: "${toString path2}" with root "${toString path2Deconstructed.root}"'';
take (length path1Deconstructed.components) path2Deconstructed.components == path1Deconstructed.components;
2023-06-16 00:19:33 +02:00
/*
2023-08-08 00:09:08 +02:00
Remove the first path as a component-wise prefix from the second path.
The result is a [normalised subpath string](#function-library-lib.path.subpath.normalise).
2023-06-16 00:19:33 +02:00
2023-08-08 00:09:08 +02:00
Laws:
2023-06-16 00:19:33 +02:00
2023-08-08 00:09:08 +02:00
- Inverts [`append`](#function-library-lib.path.append) for [normalised subpath string](#function-library-lib.path.subpath.normalise):
2023-06-16 00:19:33 +02:00
2023-08-08 00:09:08 +02:00
removePrefix p (append p s) == subpath.normalise s
2023-06-16 00:19:33 +02:00
2023-08-08 00:09:08 +02:00
Type:
removePrefix :: Path -> Path -> String
2023-06-16 00:19:33 +02:00
2023-08-08 00:09:08 +02:00
Example:
removePrefix /foo /foo/bar/baz
=> "./bar/baz"
removePrefix /foo /foo
=> "./."
removePrefix /foo/bar /foo
=> <error>
removePrefix /. /foo
=> "./foo"
2023-06-16 00:19:33 +02:00
*/
removePrefix =
path1:
assert assertMsg
(isPath path1)
"lib.path.removePrefix: First argument is of type ${typeOf path1}, but a path was expected.";
let
path1Deconstructed = deconstructPath path1;
path1Length = length path1Deconstructed.components;
in
path2:
assert assertMsg
(isPath path2)
"lib.path.removePrefix: Second argument is of type ${typeOf path2}, but a path was expected.";
let
path2Deconstructed = deconstructPath path2;
success = take path1Length path2Deconstructed.components == path1Deconstructed.components;
components =
if success then
drop path1Length path2Deconstructed.components
else
throw ''
lib.path.removePrefix: The first path argument "${toString path1}" is not a component-wise prefix of the second path argument "${toString path2}".'';
in
assert assertMsg
(path1Deconstructed.root == path2Deconstructed.root) ''
lib.path.removePrefix: Filesystem roots must be the same for both paths, but paths with different roots were given:
first argument: "${toString path1}" with root "${toString path1Deconstructed.root}"
second argument: "${toString path2}" with root "${toString path2Deconstructed.root}"'';
joinRelPath components;
2023-04-05 20:31:50 +02:00
/*
2023-08-08 00:09:08 +02:00
Split the filesystem root from a [path](https://nixos.org/manual/nix/stable/language/values.html#type-path).
The result is an attribute set with these attributes:
- `root`: The filesystem root of the path, meaning that this directory has no parent directory.
- `subpath`: The [normalised subpath string](#function-library-lib.path.subpath.normalise) that when [appended](#function-library-lib.path.append) to `root` returns the original path.
2023-08-08 00:09:08 +02:00
Laws:
- [Appending](#function-library-lib.path.append) the `root` and `subpath` gives the original path:
2023-08-08 00:09:08 +02:00
p ==
append
(splitRoot p).root
(splitRoot p).subpath
2023-08-08 00:09:08 +02:00
- Trying to get the parent directory of `root` using [`readDir`](https://nixos.org/manual/nix/stable/language/builtins.html#builtins-readDir) returns `root` itself:
2023-08-08 00:09:08 +02:00
dirOf (splitRoot p).root == (splitRoot p).root
2023-08-08 00:09:08 +02:00
Type:
splitRoot :: Path -> { root :: Path, subpath :: String }
2023-08-08 00:09:08 +02:00
Example:
splitRoot /foo/bar
=> { root = /.; subpath = "./foo/bar"; }
2023-08-08 00:09:08 +02:00
splitRoot /.
=> { root = /.; subpath = "./."; }
2023-08-08 00:09:08 +02:00
# Nix neutralises `..` path components for all path values automatically
splitRoot /foo/../bar
=> { root = /.; subpath = "./bar"; }
2023-08-08 00:09:08 +02:00
splitRoot "/foo/bar"
=> <error>
*/
splitRoot =
# The path to split the root off of
path:
assert assertMsg
(isPath path)
"lib.path.splitRoot: Argument is of type ${typeOf path}, but a path was expected";
let
deconstructed = deconstructPath path;
in {
root = deconstructed.root;
subpath = joinRelPath deconstructed.components;
};
/*
Whether a [path](https://nixos.org/manual/nix/stable/language/values.html#type-path)
has a [store path](https://nixos.org/manual/nix/stable/store/store-path.html#store-path)
as a prefix.
:::{.note}
As with all functions of this `lib.path` library, it does not work on paths in strings,
which is how you'd typically get store paths.
Instead, this function only handles path values themselves,
which occur when Nix files in the store use relative path expressions.
:::
Type:
hasStorePathPrefix :: Path -> Bool
Example:
# Subpaths of derivation outputs have a store path as a prefix
hasStorePathPrefix /nix/store/nvl9ic0pj1fpyln3zaqrf4cclbqdfn1j-foo/bar/baz
=> true
# The store directory itself is not a store path
hasStorePathPrefix /nix/store
=> false
# Derivation outputs are store paths themselves
hasStorePathPrefix /nix/store/nvl9ic0pj1fpyln3zaqrf4cclbqdfn1j-foo
=> true
# Paths outside the Nix store don't have a store path prefix
hasStorePathPrefix /home/user
=> false
# Not all paths under the Nix store are store paths
hasStorePathPrefix /nix/store/.links/10gg8k3rmbw8p7gszarbk7qyd9jwxhcfq9i6s5i0qikx8alkk4hq
=> false
# Store derivations are also store paths themselves
hasStorePathPrefix /nix/store/nvl9ic0pj1fpyln3zaqrf4cclbqdfn1j-foo.drv
=> true
*/
hasStorePathPrefix = path:
let
deconstructed = deconstructPath path;
in
assert assertMsg
(isPath path)
"lib.path.hasStorePathPrefix: Argument is of type ${typeOf path}, but a path was expected";
assert assertMsg
# This function likely breaks or needs adjustment if used with other filesystem roots, if they ever get implemented.
# Let's try to error nicely in such a case, though it's unclear how an implementation would work even and whether this could be detected.
# See also https://github.com/NixOS/nix/pull/6530#discussion_r1422843117
(deconstructed.root == /. && toString deconstructed.root == "/")
"lib.path.hasStorePathPrefix: Argument has a filesystem root (${toString deconstructed.root}) that's not /, which is currently not supported.";
componentsHaveStorePathPrefix deconstructed.components;
2023-08-08 00:09:08 +02:00
/*
Whether a value is a valid subpath string.
2023-08-08 00:09:08 +02:00
A subpath string points to a specific file or directory within an absolute base directory.
It is a stricter form of a relative path that excludes `..` components, since those could escape the base directory.
2023-08-08 00:09:08 +02:00
- The value is a string.
2023-08-08 00:09:08 +02:00
- The string is not empty.
2023-08-08 00:09:08 +02:00
- The string doesn't start with a `/`.
2023-08-08 00:09:08 +02:00
- The string doesn't contain any `..` path components.
2023-08-08 00:09:08 +02:00
Type:
subpath.isValid :: String -> Bool
2023-08-08 00:09:08 +02:00
Example:
# Not a string
subpath.isValid null
=> false
2023-08-08 00:09:08 +02:00
# Empty string
subpath.isValid ""
=> false
2023-08-08 00:09:08 +02:00
# Absolute path
subpath.isValid "/foo"
=> false
2023-08-08 00:09:08 +02:00
# Contains a `..` path component
subpath.isValid "../foo"
=> false
2023-08-08 00:09:08 +02:00
# Valid subpath
subpath.isValid "foo/bar"
=> true
2023-08-08 00:09:08 +02:00
# Doesn't need to be normalised
subpath.isValid "./foo//bar/"
=> true
*/
subpath.isValid =
# The value to check
value:
subpathInvalidReason value == null;
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
/*
Join subpath strings together using `/`, returning a normalised subpath string.
Like `concatStringsSep "/"` but safer, specifically:
- All elements must be [valid subpath strings](#function-library-lib.path.subpath.isValid).
- The result gets [normalised](#function-library-lib.path.subpath.normalise).
- The edge case of an empty list gets properly handled by returning the neutral subpath `"./."`.
Laws:
- Associativity:
subpath.join [ x (subpath.join [ y z ]) ] == subpath.join [ (subpath.join [ x y ]) z ]
- Identity - `"./."` is the neutral element for normalised paths:
subpath.join [ ] == "./."
subpath.join [ (subpath.normalise p) "./." ] == subpath.normalise p
subpath.join [ "./." (subpath.normalise p) ] == subpath.normalise p
- Normalisation - the result is [normalised](#function-library-lib.path.subpath.normalise):
subpath.join ps == subpath.normalise (subpath.join ps)
- For non-empty lists, the implementation is equivalent to [normalising](#function-library-lib.path.subpath.normalise) the result of `concatStringsSep "/"`.
Note that the above laws can be derived from this one:
ps != [] -> subpath.join ps == subpath.normalise (concatStringsSep "/" ps)
Type:
subpath.join :: [ String ] -> String
Example:
subpath.join [ "foo" "bar/baz" ]
=> "./foo/bar/baz"
# normalise the result
subpath.join [ "./foo" "." "bar//./baz/" ]
=> "./foo/bar/baz"
# passing an empty list results in the current directory
subpath.join [ ]
=> "./."
# elements must be valid subpath strings
subpath.join [ /foo ]
=> <error>
subpath.join [ "" ]
=> <error>
subpath.join [ "/foo" ]
=> <error>
subpath.join [ "../foo" ]
=> <error>
*/
subpath.join =
# The list of subpaths to join together
subpaths:
# Fast in case all paths are valid
if all isValid subpaths
then joinRelPath (concatMap splitRelPath subpaths)
else
# Otherwise we take our time to gather more info for a better error message
# Strictly go through each path, throwing on the first invalid one
# Tracks the list index in the fold accumulator
foldl' (i: path:
if isValid path
then i + 1
else throw ''
lib.path.subpath.join: Element at index ${toString i} is not a valid subpath string:
${subpathInvalidReason path}''
) 0 subpaths;
/*
2023-08-08 00:09:08 +02:00
Split [a subpath](#function-library-lib.path.subpath.isValid) into its path component strings.
Throw an error if the subpath isn't valid.
Note that the returned path components are also [valid subpath strings](#function-library-lib.path.subpath.isValid), though they are intentionally not [normalised](#function-library-lib.path.subpath.normalise).
2023-08-08 00:09:08 +02:00
Laws:
2023-08-08 00:09:08 +02:00
- Splitting a subpath into components and [joining](#function-library-lib.path.subpath.join) the components gives the same subpath but [normalised](#function-library-lib.path.subpath.normalise):
2023-08-08 00:09:08 +02:00
subpath.join (subpath.components s) == subpath.normalise s
2023-08-08 00:09:08 +02:00
Type:
subpath.components :: String -> [ String ]
2023-08-08 00:09:08 +02:00
Example:
subpath.components "."
=> [ ]
2023-08-08 00:09:08 +02:00
subpath.components "./foo//bar/./baz/"
=> [ "foo" "bar" "baz" ]
2023-08-08 00:09:08 +02:00
subpath.components "/foo"
=> <error>
*/
subpath.components =
# The subpath string to split into components
subpath:
assert assertMsg (isValid subpath) ''
lib.path.subpath.components: Argument is not a valid subpath string:
${subpathInvalidReason subpath}'';
splitRelPath subpath;
2023-08-08 00:09:08 +02:00
/*
Normalise a subpath. Throw an error if the subpath isn't [valid](#function-library-lib.path.subpath.isValid).
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Limit repeating `/` to a single one.
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Remove redundant `.` components.
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Remove trailing `/` and `/.`.
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Add leading `./`.
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
Laws:
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Idempotency - normalising multiple times gives the same result:
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
subpath.normalise (subpath.normalise p) == subpath.normalise p
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Uniqueness - there's only a single normalisation for the paths that lead to the same file system node:
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
subpath.normalise p != subpath.normalise q -> $(realpath ${p}) != $(realpath ${q})
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Don't change the result when [appended](#function-library-lib.path.append) to a Nix path value:
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
append base p == append base (subpath.normalise p)
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Don't change the path according to `realpath`:
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
$(realpath ${p}) == $(realpath ${subpath.normalise p})
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
- Only error on [invalid subpaths](#function-library-lib.path.subpath.isValid):
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
builtins.tryEval (subpath.normalise p)).success == subpath.isValid p
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
Type:
subpath.normalise :: String -> String
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
Example:
# limit repeating `/` to a single one
subpath.normalise "foo//bar"
=> "./foo/bar"
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# remove redundant `.` components
subpath.normalise "foo/./bar"
=> "./foo/bar"
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# add leading `./`
subpath.normalise "foo/bar"
=> "./foo/bar"
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# remove trailing `/`
subpath.normalise "foo/bar/"
=> "./foo/bar"
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# remove trailing `/.`
subpath.normalise "foo/bar/."
=> "./foo/bar"
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# Return the current directory as `./.`
subpath.normalise "."
=> "./."
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# error on `..` path components
subpath.normalise "foo/../bar"
=> <error>
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# error on empty string
subpath.normalise ""
=> <error>
2022-12-23 21:07:30 +01:00
2023-08-08 00:09:08 +02:00
# error on absolute path
subpath.normalise "/foo"
=> <error>
2022-12-23 21:07:30 +01:00
*/
subpath.normalise =
# The subpath string to normalise
subpath:
assert assertMsg (isValid subpath) ''
lib.path.subpath.normalise: Argument is not a valid subpath string:
${subpathInvalidReason subpath}'';
joinRelPath (splitRelPath subpath);
2022-12-23 21:07:30 +01:00
}