{-# LANGUAGE CPP #-} #define FILEPATH_NAME AbstractFilePath #define OSSTRING_NAME OsString #define WORD_NAME OsChar #define CTOR OsString #define WTOR OsChar -- | -- Module : System.AbstractFilePath -- Copyright : © 2021 Julian Ospald -- License : MIT -- -- Maintainer : Julian Ospald <hasufell@posteo.de> -- Stability : experimental -- Portability : portable -- -- An implementation of the <https://gitlab.haskell.org/ghc/ghc/-/wikis/proposal/abstract-file-path Abstract FilePath Proposal>, -- which aims to supersede @type FilePath = String@ for various reasons: -- -- 1. it is more efficient (uses unpinned 'ShortByteString' under the hood) -- 2. is more type-safe (not a type synonym, but a newtype) -- 3. avoids round-tripping issues, by not converting to String (which loses the encoding) -- -- It is important to know that filenames\/filepaths have different representations across platforms: -- -- - On /Windows/, filepaths are expected to be in UTF16 as passed to -- syscalls (although there are other APIs, the <https://hackage.haskell.org/package/Win32 Win32> package uses the wide character one). -- This invariant is maintained by 'AbstractFilePath'. -- - On /Unix/, filepaths don't have a predefined encoding (although they -- are often interpreted as UTF8) as per the -- <https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap03.html#tag_03_170 POSIX specification> -- and are passed as @char[]@ to syscalls. 'AbstractFilePath' maintains no invariant -- here. Some functions however, such as 'toAbstractFilePath', may expect -- or produce UTF8. -- -- Apart from encoding, filepaths have additional restrictions per platform: -- -- - On /Windows/ the <https://docs.microsoft.com/en-us/windows/win32/fileio/naming-a-file#naming-conventions naming convention> may apply -- - On /Unix/, only @NUL@ bytes are disallowed as per the <https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap03.html#tag_03_170 POSIX specification> -- -- Use 'isValid' to check for these restrictions ('AbstractFilePath' doesn't -- maintain this invariant). -- -- Also note that these restrictions are -- not exhaustive and further filesystem specific restrictions may apply on -- all platforms. This library makes no attempt at satisfying these. -- Library users may need to account for that, depending -- on what filesystems they want to support. -- -- It is advised to follow these principles when dealing with filepaths\/filenames: -- -- 1. Avoid interpreting filenames that the OS returns, unless absolutely necessary. -- For example, the filepath separator is usually a predefined 'Word8', regardless of encoding. -- So even if we need to split filepaths, it might still not be necessary to understand the encoding -- of the filename. -- 2. When interpreting OS returned filenames consider that these might not be UTF8 on /unix/ -- or at worst don't have an ASCII compatible encoding. Some strategies here involve looking -- up the current locale and using that for decoding ('fromAbstractFilePathIO' does this). -- Otherwise it can be reasonable to assume UTF8 on unix ('fromAbstractFilePath' does that) if your application specifically -- mentions that it requires a UTF8 compatible system. These things should be documented. -- 3. When dealing with user input (e.g. on the command line) on /unix/ as e.g. @String@ the input -- encoding is lost. The output encoding (e.g. how we write a filename to disk) can then -- either follow the current locale again ('toAbstractFilePathIO') or a fixed encoding -- ('toAbstractFilePath'). The decision should be clearly documented. If the input is in the -- form of a @ByteString@, then 'bsToAFP' may be of interest, unless the input needs further -- interpretation. #include "AbstractFilePath/Common.hs"