Monadoc
The latest version of pathtype is 0.8.1.3-1.

pathtype

Version 0.7.0.1 revision 0 uploaded by HenningThielemann.

Package meta

Synopsis
Type-safe replacement for System.FilePath etc
Description

This package provides type-safe access to filepath manipulations.

System.Path is designed to be used instead of System.FilePath. (It is intended to provide versions of functions from that module which have equivalent functionality but are more typesafe). System.Path.Directory is a companion module providing a type-safe alternative to System.Directory.

The heart of this module is the Path ar fd abstract type which represents file and directory paths. The idea is that there are two type parameters - the first should be Abs or Rel, and the second File or Dir. A number of type synonyms are provided for common types:

type AbsFile     = Path Abs File
type RelFile     = Path Rel File
type AbsDir      = Path Abs Dir
type RelDir      = Path Rel Dir

type AbsPath  fd = Path Abs fd
type RelPath  fd = Path Rel fd
type FilePath ar = Path ar File
type DirPath  ar = Path ar Dir

The type of the combine (aka </>) function gives the idea:

(</>) :: DirPath ar -> RelPath fd -> Path ar fd

Together this enables us to give more meaningful types to a lot of the functions, and (hopefully) catch a bunch more errors at compile time.

You can use the construction functions as follows:

f :: Path.RelFile
f = relDir "tmp" </> relFile "someFile" <.> "ext"

or...

f :: Path.RelFile
f = dirPath "tmp" </> filePath "someFile" <.> "ext"

or...

f :: Path.RelFile
f = path "tmp" </> path "someFile" <.> "ext"

or just...

f :: Path.RelFile
f = relFile "tmp/someFile.ext"

The first and the last implementations force the most specific types and thus should be prefered.

Overloaded string literals are no longer supported, since this extension is intended for alternative text storage types. It would also decrease the type safety if you could omit the path type and let the compiler guess its type.

You will typically want to import as follows:

import qualified System.Path.Directory as Dir
import qualified System.Path.IO as PathIO
import qualified System.Path as Path
import System.Path ((</>))

System.Path.Generic provides all functions with the OS as type parameter. System.Path.Posix and System.Path.Windows offers only path constructors and destructors fixed to the corresponding operating system. System.Path exports either System.Path.Posix or System.Path.Windows depending on the host system and additionally the manipulation functions from System.Path.Generic. This mix should be appropriate for the average use and should free the user from writing type annotations.

The basic API (and properties satisfied) are heavily influenced by Neil Mitchell's System.FilePath module.

Some notes on how to choose proper type parameters:

The ar and the fd type parameters have quite different meaning. The types Abs and Rel refer to a property of the path, whereas the type File and Dir refers to a property of a disk object. You can decide whether a path is absolute or relative by just watching (the beginning of) the path string. In contrast to that, you have to access the disk in order to check the existence and type of an disk object. Even more, the disk object might change at any time, e.g. the user might delete a file and create a directory of the same name, or the disk object might not exist, and the purpose of the path is to create an according file or directory. That's why even if you have a path of type FilePath ar, every function accessing the file must check that the refered object exists and is a file. Conversely, there is not much sense in checking the disk object type and then chosing the path accordingly. Instead, you must choose the path type according to what type of disk object your application needs. The reality check must be performed and is performed by the standard functions for every access to the object. If an disk object is not of the type required by the path type then this is a runtime exception that must be handled at runtime but it is not a programming error.

Sometimes you have to change the type of a path as an intermediate step to construct a path for an object of different type. E.g. you may convert the path "pkg" from DirPath to FilePath because in the next step you like to extend it to "pkg.tar.gz". This is valid use of the Path type. E.g. the function dropExtensions reduces the FilePath "pkg.tar.gz" to the new FilePath "pkg" although no-one expects that there is or will be a file with name "pkg". Thus, if a function has a FilePath parameter then there is no warranty that it accesses the according file and does not touch related disk objects. It may well be that the function derives other file and directory names from the path and accesses them. That is, a FilePath or DirPath parameter is mainly for documentation purposes but it cannot prevent you seriously from any damage.

How to cope with user input? You may get a path from the user, e.g. as command-line argument. It might be either absolute or relative and it might refer to an actual file or directory or to something yet non-existing. In most cases it will not be important whether the path is absolute or relative, thus you should choose the AbsOrRel type parameter. If somewhere in the program an Abs path is needed then you can assert that the path is actually absolutized somewhere e.g. by dynamicMakeAbsolute. If you prevent usage of genericMakeAbsolute then you avoid to absolutize a path that is already absolutized.

The choice of the fd type parameter follows a different reasoning: Often you have a clear idea of whether the user must pass a file or directory path. The rule is: Just give the path the type you expect but do not perform any checking (unless you want to warn the user earlier about imminent danger). The disk object type must checked for every access to the object, anyway, so there is no point in checking it immediately. With your choice of the fd parameter you just document its intended use.

It might be that the path is only a base name used to construct other directory and file names. E.g. for an Audacity project named music you have to create the directory music_data and the file music.aup. In this case we recommend to give music the type FilePath. This type warrants that there is at least one final path component in contrast to a directory path that might be empty. You can easily convert a file path to a directory path using Path.dirFromFile. The reverse conversion is partial.

Some notes on file system links:

This package does not explicitly handle file system links. We treat a file path containing links like any other file path. The same holds for directory paths. A link is handled like any other path component.

Some notes on drive-relative paths on Windows:

We use the Rel type for paths that can be relative to any directory. We use the Abs type for all other paths, i.e. for paths with explicit locations or with restrictions on the set of locations. Windows has a notion of drives and maintains a current directory for every drive. E.g. the path "c:text.txt" refers to the current directory of drive C. Since it cannot be freely combined with other directories we treat this path like an absolute path. This is consistent with the behaviour of the filepath package. E.g. filepath evaluates all of the expressions "\\abs" </> "c:driverel", "c:\\abs" </> "c:driverel", "d:\\abs" </> "c:driverel" to "c:driverel". In our package you would have to use genericMakeAbsolute but we recommend to avoid its use.

Related packages:

  • path: Provides a wrapper type around FilePath and maps to functions from filepath package. This warrants consistency with filepath functions. Requires Template Haskell.

  • data-filepath: Requires Typeable and Template Haskell.

Author
Ben Moseley, Ben Millwood, Henning Thielemann
Bug reports
n/a
Category
System
Copyright
n/a
Homepage
http://hub.darcs.net/thielema/pathtype/
Maintainer
haskell@henning-thielemann.de, ben@moseley.name
Package URL
n/a
Stability
experimental

Components