createDirectory dir creates a new directory dir which is initially empty, or as near to empty as the operating system allows.

The operation may fail with:

• isPermissionError The process has insufficient privileges to perform the operation. [EROFS, EACCES]
• isAlreadyExistsError The operand refers to a directory that already exists. [EEXIST]
• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument The operand is not a valid directory name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError There is no path to the directory. [ENOENT, ENOTDIR]
• isFullError Insufficient resources (virtual memory, process file descriptors, physical disk space, etc.) are available to perform the operation. [EDQUOT, ENOSPC, ENOMEM, EMLINK]
• InappropriateType The path refers to an existing non-directory object. [EEXIST]

createDirectoryIfMissing parents dir creates a new directory dir if it doesn't exist. If the first argument is True the function will also create all parent directories if they are missing.

removeDirectory dir removes an existing directory dir. The implementation may specify additional constraints which must be satisfied before a directory can be removed (e.g. the directory has to be empty, or may not be in use by other processes). It is not legal for an implementation to partially remove a directory unless the entire directory is removed. A conformant implementation need not support directory removal in all situations (e.g. removal of the root directory).

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument The operand is not a valid directory name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The directory does not exist. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EROFS, EACCES, EPERM]
• UnsatisfiedConstraints Implementation-dependent constraints are not satisfied. [EBUSY, ENOTEMPTY, EEXIST]
• UnsupportedOperation The implementation does not support removal in this situation. [EINVAL]
• InappropriateType The operand refers to an existing non-directory object. [ENOTDIR]

removeDirectoryRecursive dir removes an existing directory dir together with its contents and subdirectories. Within this directory, symbolic links are removed without affecting their targets.

On Windows, the operation fails if dir is a directory symbolic link.

This operation is reported to be flaky on Windows so retry logic may be advisable. See: https://github.com/haskell/directory/pull/108

Removes a file or directory at path together with its contents and subdirectories. Symbolic links are removed without affecting their targets. If the path does not exist, nothing happens.

Unlike other removal functions, this function will also attempt to delete files marked as read-only or otherwise made unremovable due to permissions. As a result, if the removal is incomplete, the permissions or attributes on the remaining files may be altered. If there are hard links in the directory, then permissions on all related hard links may be altered.

If an entry within the directory vanishes while removePathForcibly is running, it is silently ignored.

If an exception occurs while removing an entry, removePathForcibly will still try to remove as many entries as it can before failing with an exception. The first exception that it encountered is re-thrown.

renameDirectory old new changes the name of an existing directory from old to new. If the new directory already exists, it is atomically replaced by the old directory. If the new directory is neither the old directory nor an alias of the old directory, it is removed as if by removeDirectory. A conformant implementation need not support renaming directories in all situations (e.g. renaming to an existing directory, or across different physical devices), but the constraints must be documented.

On Win32 platforms, renameDirectory fails if the new directory already exists.

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument Either operand is not a valid directory name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The original directory does not exist, or there is no path to the target. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EROFS, EACCES, EPERM]
• isFullError Insufficient resources are available to perform the operation. [EDQUOT, ENOSPC, ENOMEM, EMLINK]
• UnsatisfiedConstraints Implementation-dependent constraints are not satisfied. [EBUSY, ENOTEMPTY, EEXIST]
• UnsupportedOperation The implementation does not support renaming in this situation. [EINVAL, EXDEV]
• InappropriateType Either path refers to an existing non-directory object. [ENOTDIR, EISDIR]

listDirectory dir returns a list of all entries in dir without the special entries (. and ..).

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument The operand is not a valid directory name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The directory does not exist. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EACCES]
• isFullError Insufficient resources are available to perform the operation. [EMFILE, ENFILE]
• InappropriateType The path refers to an existing non-directory object. [ENOTDIR]

Similar to listDirectory, but always includes the special entries (. and ..). (This applies to Windows as well.)

The operation may fail with the same exceptions as listDirectory.

Obtain the current working directory as an absolute path.

In a multithreaded program, the current working directory is a global state shared among all threads of the process. Therefore, when performing filesystem operations from multiple threads, it is highly recommended to use absolute rather than relative paths (see: makeAbsolute).

Note that getCurrentDirectory is not guaranteed to return the same path received by setCurrentDirectory. On POSIX systems, the path returned will always be fully dereferenced (not contain any symbolic links). For more information, refer to the documentation of getcwd.

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• isDoesNotExistError There is no path referring to the working directory. [EPERM, ENOENT, ESTALE...]
• isPermissionError The process has insufficient privileges to perform the operation. [EACCES]
• isFullError Insufficient resources are available to perform the operation.
• UnsupportedOperation The operating system has no notion of current working directory.

Change the working directory to the given path.

In a multithreaded program, the current working directory is a global state shared among all threads of the process. Therefore, when performing filesystem operations from multiple threads, it is highly recommended to use absolute rather than relative paths (see: makeAbsolute).

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument The operand is not a valid directory name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The directory does not exist. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EACCES]
• UnsupportedOperation The operating system has no notion of current working directory, or the working directory cannot be dynamically changed.
• InappropriateType The path refers to an existing non-directory object. [ENOTDIR]

Run an IO action with the given working directory and restore the original working directory afterwards, even if the given action fails due to an exception.

The operation may fail with the same exceptions as getCurrentDirectory and setCurrentDirectory.

Returns the current user's home directory.

The directory returned is expected to be writable by the current user, but note that it isn't generally considered good practice to store application-specific data here; use getXdgDirectory or getAppUserDataDirectory instead.

On Unix, getHomeDirectory behaves as follows:

• Returns $HOME env variable if set (including to an empty string).
• Otherwise uses home directory returned by getpwuid_r using the UID of the current proccesses user. This basically reads the etcpasswd file. An empty home directory field is considered valid.

On Windows, the system is queried for a suitable path; a typical path might be C:/Users/<user>.

The operation may fail with:

• UnsupportedOperation The operating system has no notion of home directory.
• isDoesNotExistError The home directory for the current user does not exist, or cannot be found.

Special directories for storing user-specific application data, configuration, and cache files, as specified by the XDG Base Directory Specification.

Note: On Windows, XdgData and XdgConfig usually map to the same directory.

Since: directory-1.2.3.0

Obtain the paths to special directories for storing user-specific application data, configuration, and cache files, conforming to the XDG Base Directory Specification. Compared with getAppUserDataDirectory, this function provides a more fine-grained hierarchy as well as greater flexibility for the user.

On Windows, XdgData and XdgConfig usually map to the same directory unless overridden.

Refer to the docs of XdgDirectory for more details.

The second argument is usually the name of the application. Since it will be integrated into the path, it must consist of valid path characters. Note: if the second argument is an absolute path, it will just return the second argument.

Note: The directory may not actually exist, in which case you would need to create it with file mode 700 (i.e. only accessible by the owner).

As of 1.3.5.0, the environment variable is ignored if set to a relative path, per revised XDG Base Directory Specification. See #100.

Search paths for various application data, as specified by the XDG Base Directory Specification.

The list of paths is split using searchPathSeparator, which on Windows is a semicolon.

Note: On Windows, XdgDataDirs and XdgConfigDirs usually yield the same result.

Since: directory-1.3.2.0

Similar to getXdgDirectory but retrieves the entire list of XDG directories.

On Windows, XdgDataDirs and XdgConfigDirs usually map to the same list of directories unless overridden.

Refer to the docs of XdgDirectoryList for more details.

Obtain the path to a special directory for storing user-specific application data (traditional Unix location). Newer applications may prefer the the XDG-conformant location provided by getXdgDirectory (migration guide).

The argument is usually the name of the application. Since it will be integrated into the path, it must consist of valid path characters.

• On Unix-like systems, the path is ~/.<app>.
• On Windows, the path is %APPDATA%/<app> (e.g. C:/Users/<user>/AppData/Roaming/<app>)

Note: the directory may not actually exist, in which case you would need to create it. It is expected that the parent directory exists and is writable.

The operation may fail with:

• UnsupportedOperation The operating system has no notion of application-specific data directory.
• isDoesNotExistError The home directory for the current user does not exist, or cannot be found.

Returns the current user's document directory.

The directory returned is expected to be writable by the current user, but note that it isn't generally considered good practice to store application-specific data here; use getXdgDirectory or getAppUserDataDirectory instead.

On Unix, getUserDocumentsDirectory returns the value of the HOME environment variable. On Windows, the system is queried for a suitable path; a typical path might be C:/Users/<user>/Documents.

The operation may fail with:

• UnsupportedOperation The operating system has no notion of document directory.
• isDoesNotExistError The document directory for the current user does not exist, or cannot be found.

Returns the current directory for temporary files.

On Unix, getTemporaryDirectory returns the value of the TMPDIR environment variable or "/tmp" if the variable isn't defined. On Windows, the function checks for the existence of environment variables in the following order and uses the first path found:

• TMP environment variable.
• TEMP environment variable.
• USERPROFILE environment variable.
• The Windows directory

The operation may fail with:

• UnsupportedOperation The operating system has no notion of temporary directory.

The function doesn't verify whether the path exists.

removeFile file removes the directory entry for an existing file file, where file is not itself a directory. The implementation may specify additional constraints which must be satisfied before a file can be removed (e.g. the file may not be in use by other processes).

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument The operand is not a valid file name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The file does not exist. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EROFS, EACCES, EPERM]
• UnsatisfiedConstraints Implementation-dependent constraints are not satisfied. [EBUSY]
• InappropriateType The operand refers to an existing directory. [EPERM, EINVAL]

renameFile old new changes the name of an existing file system object from old to new. If the new object already exists, it is replaced by the old object. Neither path may refer to an existing directory.

A conformant implementation need not support renaming files in all situations (e.g. renaming across different physical devices), but the constraints must be documented. On Windows, this does not support renaming across different physical devices; if you are looking to do so, consider using copyFileWithMetadata and removeFile.

On Windows, this calls MoveFileEx with MOVEFILE_REPLACE_EXISTING set, which is not guaranteed to be atomic (https://github.com/haskell/directory/issues/109).

On other platforms, this operation is atomic.

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument Either operand is not a valid file name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The original file does not exist, or there is no path to the target. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EROFS, EACCES, EPERM]
• isFullError Insufficient resources are available to perform the operation. [EDQUOT, ENOSPC, ENOMEM, EMLINK]
• UnsatisfiedConstraints Implementation-dependent constraints are not satisfied. [EBUSY]
• UnsupportedOperation The implementation does not support renaming in this situation. [EXDEV]
• InappropriateType Either path refers to an existing directory. [ENOTDIR, EISDIR, EINVAL, EEXIST, ENOTEMPTY]

Rename a file or directory. If the destination path already exists, it is replaced atomically. The destination path must not point to an existing directory. A conformant implementation need not support renaming files in all situations (e.g. renaming across different physical devices), but the constraints must be documented.

The operation may fail with:

• HardwareFault A physical I/O error has occurred. [EIO]
• InvalidArgument Either operand is not a valid file name. [ENAMETOOLONG, ELOOP]
• isDoesNotExistError The original file does not exist, or there is no path to the target. [ENOENT, ENOTDIR]
• isPermissionError The process has insufficient privileges to perform the operation. [EROFS, EACCES, EPERM]
• isFullError Insufficient resources are available to perform the operation. [EDQUOT, ENOSPC, ENOMEM, EMLINK]
• UnsatisfiedConstraints Implementation-dependent constraints are not satisfied. [EBUSY]
• UnsupportedOperation The implementation does not support renaming in this situation. [EXDEV]
• InappropriateType Either the destination path refers to an existing directory, or one of the parent segments in the destination path is not a directory. [ENOTDIR, EISDIR, EINVAL, EEXIST, ENOTEMPTY]

Copy a file with its permissions. If the destination file already exists, it is replaced atomically. Neither path may refer to an existing directory. No exceptions are thrown if the permissions could not be copied.

Copy a file with its associated metadata. If the destination file already exists, it is overwritten. There is no guarantee of atomicity in the replacement of the destination file. Neither path may refer to an existing directory. If the source and/or destination are symbolic links, the copy is performed on the targets of the links.

On Windows, it behaves like the Win32 function CopyFile, which copies various kinds of metadata including file attributes and security resource properties.

On Unix-like systems, permissions, access time, and modification time are preserved. If possible, the owner and group are also preserved. Note that the very act of copying can change the access time of the source file, hence the access times of the two files may differ after the operation completes.

Obtain the size of a file in bytes.

Make a path absolute, normalize the path, and remove as many indirections from it as possible. Any trailing path separators are discarded via dropTrailingPathSeparator. Additionally, on Windows the letter case of the path is canonicalized.

Note: This function is a very big hammer. If you only need an absolute path, makeAbsolute is sufficient for removing dependence on the current working directory.

Indirections include the two special directories . and .., as well as any symbolic links (and junction points on Windows). The input path need not point to an existing file or directory. Canonicalization is performed on the longest prefix of the path that points to an existing file or directory. The remaining portion of the path that does not point to an existing file or directory will still be normalized, but case canonicalization and indirection removal are skipped as they are impossible to do on a nonexistent path.

Most programs should not worry about the canonicity of a path. In particular, despite the name, the function does not truly guarantee canonicity of the returned path due to the presence of hard links, mount points, etc.

If the path points to an existing file or directory, then the output path shall also point to the same file or directory, subject to the condition that the relevant parts of the file system do not change while the function is still running. In other words, the function is definitively not atomic. The results can be utterly wrong if the portions of the path change while this function is running.

Since some indirections (symbolic links on all systems, .. on non-Windows systems, and junction points on Windows) are dependent on the state of the existing filesystem, the function can only make a conservative attempt by removing such indirections from the longest prefix of the path that still points to an existing file or directory.

Note that on Windows parent directories .. are always fully expanded before the symbolic links, as consistent with the rest of the Windows API (such as GetFullPathName). In contrast, on POSIX systems parent directories .. are expanded alongside symbolic links from left to right. To put this more concretely: if L is a symbolic link for R/P, then on Windows L\.. refers to ., whereas on other operating systems L/.. refers to R.

Similar to normalise, passing an empty path is equivalent to passing the current directory.

canonicalizePath can resolve at least 64 indirections in a single path, more than what is supported by most operating systems. Therefore, it may return the fully resolved path even though the operating system itself would have long given up.

On Windows XP or earlier systems, junction expansion is not performed due to their lack of GetFinalPathNameByHandle.

Changes since 1.2.3.0: The function has been altered to be more robust and has the same exception behavior as makeAbsolute.

Changes since 1.3.0.0: The function no longer preserves the trailing path separator. File symbolic links that appear in the middle of a path are properly dereferenced. Case canonicalization and symbolic link expansion are now performed on Windows.

Convert a path into an absolute path. If the given path is relative, the current directory is prepended and then the combined result is normalized. If the path is already absolute, the path is simply normalized. The function preserves the presence or absence of the trailing path separator unless the path refers to the root directory /.

If the path is already absolute, the operation never fails. Otherwise, the operation may fail with the same exceptions as getCurrentDirectory.

Construct a path relative to the current directory, similar to makeRelative.

The operation may fail with the same exceptions as getCurrentDirectory.

Test whether the given path points to an existing filesystem object. If the user lacks necessary permissions to search the parent directories, this function may return false even if the file does actually exist. This operation traverses symbolic links, so it can return either True or False for them.

The operation doesFileExist returns True if the argument file exists and is not a directory, and False otherwise. This operation traverses symbolic links, so it can return either True or False for them.

The operation doesDirectoryExist returns True if the argument file exists and is either a directory or a symbolic link to a directory, and False otherwise. This operation traverses symbolic links, so it can return either True or False for them.

Given the name or path of an executable file, findExecutable searches for such a file in a list of system-defined locations, which generally includes PATH and possibly more. The full path to the executable is returned if found. For example, (findExecutable "ghc") would normally give you the path to GHC.

The path returned by findExecutable name corresponds to the program that would be executed by createProcess when passed the same string (as a RawCommand, not a ShellCommand), provided that name is not a relative path with more than one segment.

On Windows, findExecutable calls the Win32 function SearchPath, which may search other places before checking the directories in the PATH environment variable. Where it actually searches depends on registry settings, but notably includes the directory containing the current executable.

On non-Windows platforms, the behavior is equivalent to findFileWith using the search directories from the PATH environment variable and testing each file for executable permissions. Details can be found in the documentation of findFileWith.

Search for executable files in a list of system-defined locations, which generally includes PATH and possibly more.

On Windows, this only returns the first occurrence, if any. Its behavior is therefore equivalent to findExecutable.

On non-Windows platforms, the behavior is equivalent to findExecutablesInDirectories using the search directories from the PATH environment variable. Details can be found in the documentation of findExecutablesInDirectories.

Given a name or path, findExecutable appends the exeExtension to the query and searches for executable files in the list of given search directories and returns all occurrences.

The behavior is equivalent to findFileWith using the given search directories and testing each file for executable permissions. Details can be found in the documentation of findFileWith.

Unlike other similarly named functions, findExecutablesInDirectories does not use SearchPath from the Win32 API. The behavior of this function on Windows is therefore equivalent to those on non-Windows platforms.

Search through the given list of directories for the given file.

The behavior is equivalent to findFileWith, returning only the first occurrence. Details can be found in the documentation of findFileWith.

Search through the given list of directories for the given file and returns all paths where the given file exists.

The behavior is equivalent to findFilesWith. Details can be found in the documentation of findFilesWith.

Search through a given list of directories for a file that has the given name and satisfies the given predicate and return the path of the first occurrence. The directories are checked in a left-to-right order.

This is essentially a more performant version of findFilesWith that always returns the first result, if any. Details can be found in the documentation of findFilesWith.

findFilesWith predicate dirs name searches through the list of directories (dirs) for files that have the given name and satisfy the given predicate and returns the paths of those files. The directories are checked in a left-to-right order and the paths are returned in the same order.

If the name is a relative path, then for every search directory dir, the function checks whether dir </> name exists and satisfies the predicate. If so, dir </> name is returned as one of the results. In other words, the returned paths can be either relative or absolute depending on the search directories were used. If there are no search directories, no results are ever returned.

If the name is an absolute path, then the function will return a single result if the file exists and satisfies the predicate and no results otherwise. This is irrespective of what search directories were given.

Filename extension for executable files (including the dot if any) (usually "" on POSIX systems and ".exe" on Windows or OS/2).

Create a file symbolic link. The target path can be either absolute or relative and need not refer to an existing file. The order of arguments follows the POSIX convention.

To remove an existing file symbolic link, use removeFile.

Although the distinction between file symbolic links and directory symbolic links does not exist on POSIX systems, on Windows this is an intrinsic property of every symbolic link and cannot be changed without recreating the link. A file symbolic link that actually points to a directory will fail to dereference and vice versa. Moreover, creating symbolic links on Windows may require privileges unavailable to users outside the Administrators group. Portable programs that use symbolic links should take both into consideration.

On Windows, the function is implemented using CreateSymbolicLink. Since 1.3.3.0, the SYMBOLIC_LINK_FLAG_ALLOW_UNPRIVILEGED_CREATE flag is included if supported by the operating system. On POSIX, the function uses symlink and is therefore atomic.

Windows-specific errors: This operation may fail with permissionErrorType if the user lacks the privileges to create symbolic links. It may also fail with illegalOperationErrorType if the file system does not support symbolic links.

Create a directory symbolic link. The target path can be either absolute or relative and need not refer to an existing directory. The order of arguments follows the POSIX convention.

To remove an existing directory symbolic link, use removeDirectoryLink.

Although the distinction between file symbolic links and directory symbolic links does not exist on POSIX systems, on Windows this is an intrinsic property of every symbolic link and cannot be changed without recreating the link. A file symbolic link that actually points to a directory will fail to dereference and vice versa. Moreover, creating symbolic links on Windows may require privileges unavailable to users outside the Administrators group. Portable programs that use symbolic links should take both into consideration.

On Windows, the function is implemented using CreateSymbolicLink with SYMBOLIC_LINK_FLAG_DIRECTORY. Since 1.3.3.0, the SYMBOLIC_LINK_FLAG_ALLOW_UNPRIVILEGED_CREATE flag is also included if supported by the operating system. On POSIX, this is an alias for createFileLink and is therefore atomic.

Windows-specific errors: This operation may fail with permissionErrorType if the user lacks the privileges to create symbolic links. It may also fail with illegalOperationErrorType if the file system does not support symbolic links.

Remove an existing directory symbolic link.

On Windows, this is an alias for removeDirectory. On POSIX systems, this is an alias for removeFile.

See also: removeFile, which can remove an existing file symbolic link.

Check whether an existing path is a symbolic link. If path is a regular file or directory, False is returned. If path does not exist or is otherwise inaccessible, an exception is thrown (see below).

On Windows, this checks for FILE_ATTRIBUTE_REPARSE_POINT. In addition to symbolic links, the function also returns true on junction points. On POSIX systems, this checks for S_IFLNK.

The operation may fail with:

• isDoesNotExistError if the symbolic link does not exist; or
• isPermissionError if the user is not permitted to read the symbolic link.

Retrieve the target path of either a file or directory symbolic link. The returned path may not be absolute, may not exist, and may not even be a valid path.

On Windows systems, this calls DeviceIoControl with FSCTL_GET_REPARSE_POINT. In addition to symbolic links, the function also works on junction points. On POSIX systems, this calls readlink.

Windows-specific errors: This operation may fail with illegalOperationErrorType if the file system does not support symbolic links.

Get the permissions of a file or directory.

On Windows, the writable permission corresponds to the "read-only" attribute. The executable permission is set if the file extension is of an executable file type. The readable permission is always set.

On POSIX systems, this returns the result of access.

The operation may fail with:

• isPermissionError if the user is not permitted to access the permissions, or
• isDoesNotExistError if the file or directory does not exist.

Set the permissions of a file or directory.

On Windows, this is only capable of changing the writable permission, which corresponds to the "read-only" attribute. Changing the other permissions has no effect.

On POSIX systems, this sets the owner permissions.

The operation may fail with:

• isPermissionError if the user is not permitted to set the permissions, or
• isDoesNotExistError if the file or directory does not exist.

Copy the permissions of one file to another. This reproduces the permissions more accurately than using getPermissions followed by setPermissions.

On Windows, this copies only the read-only attribute.

On POSIX systems, this is equivalent to stat followed by chmod.

Obtain the time at which the file or directory was last accessed.

The operation may fail with:

• isPermissionError if the user is not permitted to read the access time; or
• isDoesNotExistError if the file or directory does not exist.

Caveat for POSIX systems: This function returns a timestamp with sub-second resolution only if this package is compiled against unix-2.6.0.0 or later and the underlying filesystem supports them.

Obtain the time at which the file or directory was last modified.

The operation may fail with:

• isPermissionError if the user is not permitted to read the modification time; or
• isDoesNotExistError if the file or directory does not exist.

Caveat for POSIX systems: This function returns a timestamp with sub-second resolution only if this package is compiled against unix-2.6.0.0 or later and the underlying filesystem supports them.

Change the time at which the file or directory was last accessed.

The operation may fail with:

• isPermissionError if the user is not permitted to alter the access time; or
• isDoesNotExistError if the file or directory does not exist.

Some caveats for POSIX systems:

• Not all systems support utimensat, in which case the function can only emulate the behavior by reading the modification time and then setting both the access and modification times together. On systems where utimensat is supported, the access time is set atomically with nanosecond precision.
• If compiled against a version of unix prior to 2.7.0.0, the function would not be able to set timestamps with sub-second resolution. In this case, there would also be loss of precision in the modification time.

Change the time at which the file or directory was last modified.

The operation may fail with:

• isPermissionError if the user is not permitted to alter the modification time; or
• isDoesNotExistError if the file or directory does not exist.
• InvalidArgument on FAT32 file system if the time is before DOS Epoch (1 January 1980).

Some caveats for POSIX systems:

• Not all systems support utimensat, in which case the function can only emulate the behavior by reading the access time and then setting both the access and modification times together. On systems where utimensat is supported, the modification time is set atomically with nanosecond precision.
• If compiled against a version of unix prior to 2.7.0.0, the function would not be able to set timestamps with sub-second resolution. In this case, there would also be loss of precision in the access time.