hpath/src/HPath.hs

480 lines
12 KiB
Haskell
Raw Normal View History

Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- |
-- Module : HPath
-- Copyright : © 20152016 FP Complete, 2016 Julian Ospald
-- License : BSD 3 clause
--
-- Maintainer : Julian Ospald <hasufell@posteo.de>
-- Stability : experimental
-- Portability : portable
--
-- Support for well-typed paths.
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
{-# LANGUAGE DeriveDataTypeable #-}
{-# LANGUAGE EmptyDataDecls #-}
{-# LANGUAGE PatternSynonyms #-}
2016-03-30 00:47:42 +00:00
{-# LANGUAGE ForeignFunctionInterface #-}
{-# LANGUAGE OverloadedStrings #-}
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
{-# OPTIONS_HADDOCK ignore-exports #-}
module HPath
(
-- * Types
Abs
,Path
,Rel
2016-03-30 00:47:42 +00:00
,Fn
,PathParseException
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- * PatternSynonyms/ViewPatterns
,pattern Path
-- * Path Parsing
2016-03-30 00:47:42 +00:00
,parseAbs
,parseFn
,parseRel
-- * Path Conversion
,canonicalizePath
,fromAbs
,fromRel
,normalize
,toFilePath
-- * Path Operations
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
,(</>)
,basename
,dirname
,isParentOf
2016-04-03 14:20:38 +00:00
,getAllParents
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
,stripDir
-- * ByteString/Word8 constants
,nullByte
,pathDot
,pathDot'
,pathSeparator
,pathSeparator'
-- * ByteString operations
,addTrailingPathSeparator
,combine
,dropFileName
,dropTrailingPathSeparator
,dropWhileEnd
,joinPath
,normalise
,splitDirectories
,splitFileName
,splitPath
,stripPrefix
,takeDirectory
2016-03-30 00:47:42 +00:00
-- * Queries
,hasDot
,hasDoublePS
,hasLeadingPathSeparator
2016-03-30 00:47:42 +00:00
,hasParentDir
,hasTrailingPathSeparator
,isAbsolute
2016-03-30 00:47:42 +00:00
,isFileName
,isRelative
,isValid
2016-03-30 00:47:42 +00:00
-- * String based functions
,realPath
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
)
where
import Control.Exception (Exception)
2016-03-30 00:47:42 +00:00
import Control.Monad(void)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
import Control.Monad.Catch (MonadThrow(..))
import Data.ByteString(ByteString)
import qualified Data.ByteString as B
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
import Data.Data
import qualified Data.List as L
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
import Data.Maybe
2016-04-04 16:46:25 +00:00
import Data.Word8
2016-03-30 00:47:42 +00:00
import Foreign.C.Error
import Foreign.C.String
import Foreign.Marshal.Alloc(allocaBytes)
import HPath.Foreign
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
import HPath.Internal
--------------------------------------------------------------------------------
-- Types
-- | An absolute path.
data Abs deriving (Typeable)
-- | A relative path; one without a root.
data Rel deriving (Typeable)
2016-03-30 00:47:42 +00:00
-- | A filename, without any '/'.
data Fn deriving (Typeable)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Exception when parsing a location.
data PathParseException
= InvalidAbs ByteString
| InvalidRel ByteString
| InvalidFn ByteString
| Couldn'tStripPrefixTPS ByteString ByteString
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
deriving (Show,Typeable)
instance Exception PathParseException
2016-03-30 00:47:42 +00:00
instance RelC Rel
instance RelC Fn
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
--------------------------------------------------------------------------------
-- PatternSynonyms
pattern Path x <- (MkPath x)
--------------------------------------------------------------------------------
-- Path Parsers
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
2016-03-30 00:47:42 +00:00
-- | Get a location for an absolute path.
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
--
-- Throws: 'PathParseException'
--
2016-03-30 00:47:42 +00:00
parseAbs :: MonadThrow m
=> ByteString -> m (Path Abs)
2016-03-30 00:47:42 +00:00
parseAbs filepath =
if isAbsolute filepath &&
not (B.null filepath) &&
isValid filepath
2016-03-30 00:47:42 +00:00
then return (MkPath filepath)
else throwM (InvalidAbs filepath)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Get a location for a relative path. Produces a normalized
-- path which always ends in a path separator.
--
-- Note that @filepath@ may contain any number of @./@ but may not consist
-- solely of @./@. It also may not contain a single @..@ anywhere.
--
-- Throws: 'PathParseException'
--
2016-03-30 00:47:42 +00:00
parseRel :: MonadThrow m
=> ByteString -> m (Path Rel)
2016-03-30 00:47:42 +00:00
parseRel filepath =
if not (isAbsolute filepath) &&
not (B.null filepath) &&
isValid filepath
2016-03-30 00:47:42 +00:00
then return (MkPath filepath)
else throwM (InvalidRel filepath)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
2016-03-30 00:47:42 +00:00
parseFn :: MonadThrow m
=> ByteString -> m (Path Fn)
2016-03-30 00:47:42 +00:00
parseFn filepath =
if not (isAbsolute filepath) &&
not (B.null filepath) &&
2016-03-30 00:47:42 +00:00
isFileName filepath &&
isValid filepath
2016-03-30 00:47:42 +00:00
then return (MkPath filepath)
else throwM (InvalidFn filepath)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
--------------------------------------------------------------------------------
-- Path Conversion
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Convert to a ByteString type.
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
--
-- All TPS data types have a trailing slash, so if you want no trailing
-- slash, you can use 'System.FilePath.dropTrailingPathSeparator' from
-- the filepath package.
toFilePath :: Path b -> ByteString
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
toFilePath (MkPath l) = l
fromAbs :: Path Abs -> ByteString
2016-03-30 00:47:42 +00:00
fromAbs = toFilePath
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
fromRel :: RelC r => Path r -> ByteString
2016-03-30 00:47:42 +00:00
fromRel = toFilePath
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
2016-03-30 00:47:42 +00:00
normalize :: Path t -> Path t
normalize (MkPath l) = MkPath $ normalise l
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
2016-03-30 00:47:42 +00:00
-- | May fail on `realPath`.
canonicalizePath :: Path Abs -> IO (Path Abs)
canonicalizePath (MkPath l) = do
nl <- realPath l
return $ MkPath nl
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
--------------------------------------------------------------------------------
-- Path Operations
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Append two paths.
--
-- The second argument must always be a relative path, which ensures
-- that undefinable things like `"/abc" </> "/def"` cannot happen.
--
-- Technically, the first argument can be a path that points to a non-directory,
-- because this library is IO-agnostic and makes no assumptions about
-- file types.
2016-03-30 00:47:42 +00:00
(</>) :: RelC r => Path b -> Path r -> Path b
(</>) (MkPath a) (MkPath b) = MkPath (a' `B.append` b)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
where
a' = addTrailingPathSeparator a
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Strip directory from path, making it relative to that directory.
-- Throws 'Couldn'tStripPrefixDir' if directory is not a parent of the path.
--
-- The bases must match.
--
stripDir :: MonadThrow m
2016-03-30 00:47:42 +00:00
=> Path b -> Path b -> m (Path Rel)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
stripDir (MkPath p) (MkPath l) =
case stripPrefix p' l of
Nothing -> throwM (Couldn'tStripPrefixTPS p' l)
Just "" -> throwM (Couldn'tStripPrefixTPS p' l)
Just ok -> return (MkPath ok)
where
p' = addTrailingPathSeparator p
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Is p a parent of the given location? Implemented in terms of
-- 'stripDir'. The bases must match.
2016-03-30 00:47:42 +00:00
isParentOf :: Path b -> Path b -> Bool
isParentOf p l = isJust (stripDir p l :: Maybe (Path Rel))
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
2016-04-03 14:20:38 +00:00
getAllParents :: Path Abs -> [Path Abs]
getAllParents (MkPath p) =
case np of
(MkPath "/") -> []
_ -> dirname np : getAllParents (dirname np)
where
np = MkPath . dropTrailingPathSeparator . normalise $ p
2016-04-03 14:20:38 +00:00
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Extract the directory name of a path.
--
-- The following properties hold:
--
-- @dirname (p \<\/> a) == dirname p@
--
2016-03-30 00:47:42 +00:00
dirname :: Path Abs -> Path Abs
dirname (MkPath fp) = MkPath (takeDirectory $ dropTrailingPathSeparator fp)
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
-- | Extract the file part of a path.
--
--
-- The following properties hold:
--
-- @basename (p \<\/> a) == basename a@
--
2016-03-30 00:47:42 +00:00
-- Except when "/" is passed in which case the filename "." is returned.
basename :: Path b -> Path Fn
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
basename (MkPath l)
| not (isAbsolute rl) = MkPath rl
2016-03-30 00:47:42 +00:00
| otherwise = MkPath "."
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
where
rl = last . splitPath . dropTrailingPathSeparator $ l
--------------------------------------------------------------------------------
-- ByteString/Word8 constants
pathSeparator :: Word8
2016-04-04 16:46:25 +00:00
pathSeparator = _slash
pathSeparator' :: ByteString
2016-04-04 16:46:25 +00:00
pathSeparator' = B.singleton pathSeparator
pathDot :: Word8
2016-04-04 16:46:25 +00:00
pathDot = _period
pathDot' :: ByteString
2016-04-04 16:46:25 +00:00
pathDot' = B.singleton pathDot
nullByte :: Word8
2016-04-04 16:46:25 +00:00
nullByte = _nul
--------------------------------------------------------------------------------
-- ByteString Operations
dropWhileEnd :: (Word8 -> Bool) -> ByteString -> ByteString
dropWhileEnd p = B.reverse . B.dropWhile p . B.reverse
dropTrailingPathSeparator :: ByteString -> ByteString
dropTrailingPathSeparator filepath =
if hasTrailingPathSeparator filepath
then let filepath' = dropWhileEnd (== pathSeparator) filepath
in if B.null filepath' then B.singleton . B.last $ filepath else filepath'
else filepath
addTrailingPathSeparator :: ByteString -> ByteString
addTrailingPathSeparator filepath
| B.null filepath = filepath
| filepath == pathSeparator' = filepath
| not (hasTrailingPathSeparator filepath)
= filepath `B.append` pathSeparator'
| otherwise = filepath
normalise :: ByteString -> ByteString
normalise filepath = result `B.append`
(if addPathSeparator then pathSeparator' else B.empty)
where
(drv, pth) = splitDrive filepath
result = joinDrive' (normaliseDrive drv) (f pth)
joinDrive' d p
| d == "" && p == "" = B.singleton pathDot
| otherwise = joinDrive d p
addPathSeparator = isDirPath pth && not (hasTrailingPathSeparator result)
isDirPath xs = hasTrailingPathSeparator xs
|| not (B.null xs) && B.last xs == pathDot
&& hasTrailingPathSeparator (B.init xs)
normaliseDrive p
| p == "" = ""
| otherwise = B.singleton pathSeparator
f = joinPath . dropDots . propSep . splitDirectories
propSep :: [ByteString] -> [ByteString]
propSep (x:xs)
| B.all (== pathSeparator) x = pathSeparator' : xs
| otherwise = x : xs
propSep [] = []
dropDots = filter (pathDot' /=)
splitPath :: ByteString -> [ByteString]
splitPath filepath = [drv | drv /= ""] ++ f pth
where
(drv, pth) = splitDrive filepath
f p
| p == "" = []
| otherwise = (a `B.append` c) : f d
where
(a, b) = B.break (== pathSeparator) p
(c, d) = splitDrive b
joinPath :: [ByteString] -> ByteString
joinPath = foldr combine ""
splitDrive :: ByteString -> (ByteString, ByteString)
splitDrive = B.span (== pathSeparator)
joinDrive :: ByteString -> ByteString -> ByteString
joinDrive = combineAlways
splitDirectories :: ByteString -> [ByteString]
splitDirectories = map dropTrailingPathSeparator . splitPath
combine :: ByteString -> ByteString -> ByteString
combine d p
| hasLeadingPathSeparator p = p
| otherwise = combineAlways d p
combineAlways :: ByteString -> ByteString -> ByteString
combineAlways d p
| B.null d = p
| B.null p = d
| hasTrailingPathSeparator d = d `B.append` p
| otherwise = d `B.append` B.singleton pathSeparator `B.append` p
takeDirectory :: ByteString -> ByteString
takeDirectory = dropTrailingPathSeparator . dropFileName
dropFileName :: ByteString -> ByteString
dropFileName = fst . splitFileName
splitFileName :: ByteString -> (ByteString, ByteString)
splitFileName filepath = (if B.null dir then "./" else dir, name)
where
(dir, name) = splitFileName_ filepath
splitFileName_ p = (drv `B.append` dir', file)
where
(drv, pth) = splitDrive p
(dir', file) = B.breakEnd (== pathSeparator) pth
stripPrefix :: ByteString -> ByteString -> Maybe ByteString
stripPrefix a b = B.pack `fmap` L.stripPrefix (B.unpack a) (B.unpack b)
2016-03-30 00:47:42 +00:00
2016-04-03 14:20:38 +00:00
2016-03-30 00:47:42 +00:00
--------------------------------------------------------------------------------
-- ByteString Query functions
2016-03-30 00:47:42 +00:00
-- | Helper function: check if the filepath has any parent directories in it.
hasParentDir :: ByteString -> Bool
2016-03-30 00:47:42 +00:00
hasParentDir filepath =
("/.." `B.isSuffixOf` filepath) ||
("/../" `B.isInfixOf` filepath) ||
("../" `B.isPrefixOf` filepath)
2016-03-30 00:47:42 +00:00
hasDot :: ByteString -> Bool
2016-03-30 00:47:42 +00:00
hasDot filepath =
("/." `B.isSuffixOf` filepath) ||
("/./" `B.isInfixOf` filepath) ||
("./" `B.isPrefixOf` filepath)
2016-03-30 00:47:42 +00:00
hasDoublePS :: ByteString -> Bool
2016-03-30 00:47:42 +00:00
hasDoublePS filepath =
("//" `B.isInfixOf` filepath)
hasTrailingPathSeparator :: ByteString -> Bool
hasTrailingPathSeparator filepath
| B.null filepath = False
| B.last filepath == pathSeparator = True
| otherwise = False
2016-03-30 00:47:42 +00:00
hasLeadingPathSeparator :: ByteString -> Bool
hasLeadingPathSeparator filepath
| B.null filepath = False
| B.head filepath == pathSeparator = True
| otherwise = False
isFileName :: ByteString -> Bool
2016-03-30 00:47:42 +00:00
isFileName filepath =
not ("/" `B.isInfixOf` filepath)
isAbsolute :: ByteString -> Bool
isAbsolute filepath
| B.null filepath = False
| B.head filepath == pathSeparator = True
| otherwise = False
isRelative :: ByteString -> Bool
isRelative = not . isAbsolute
isValid :: ByteString -> Bool
isValid filepath
| B.null filepath = False
| filepath == "" = False
| nullByte `B.elem` filepath = False
| otherwise = True
2016-03-30 00:47:42 +00:00
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00
--------------------------------------------------------------------------------
2016-03-30 00:47:42 +00:00
-- String based path functions
foreign import ccall "realpath"
c_realpath :: CString -> CString -> IO CString
-- | return the canonicalized absolute pathname
--
-- like canonicalizePath, but uses realpath(3)
realPath :: ByteString -> IO ByteString
realPath inp =
2016-04-04 01:06:22 +00:00
allocaBytes pathMax $ \tmp -> do
void $ B.useAsCString inp
2016-04-04 01:06:22 +00:00
$ \cstr -> throwErrnoIfNull "realpath"
$ c_realpath cstr tmp
B.packCString tmp
Fork chrisdone's path library I wasn't happy with the way it dealt with Dir vs File things. In his version of the library, a `Path b Dir` always ends with a trailing path separator and `Path b File` never ends with a trailing path separator. IMO, it is nonsensical to make a Dir vs File distinction on path level, although it first seems nice. Some of the reasons are: * a path is just that: a path. It is completely disconnected from IO level and even if a `Dir`/`File` type theoretically allows us to say "this path ought to point to a file", there is literally zero guarantee that it will hold true at runtime. So this basically gives a false feeling of a type-safe file distinction. * it's imprecise about Dir vs File distinction, which makes it even worse, because a directory is also a file (just not a regular file). Add symlinks to that and the confusion is complete. * it makes the API oddly complicated for use cases where we basically don't care (yet) whether something turns out to be a directory or not Still, it comes also with a few perks: * it simplifies some functions, because they now have guarantees whether a path ends in a trailing path separator or not * it may be safer for interaction with other library functions, which behave differently depending on a trailing path separator (like probably shelly) Not limited to, but also in order to fix my remarks without breaking any benefits, I did: * rename the `Dir`/`File` types to `TPS`/`NoTPS`, so it's clear we are only giving information about trailing path separators and not actual file types we don't know about yet * add a `MaybeTPS` type, which does not mess with trailing path separators and also gives no guarantees about them... then added `toNoTPS` and `toTPS` to allow type-safe conversion * make some functions accept more general types, so we don't unnecessarily force paths with trailing separators for `(</>)` for example... instead these functions now examine the paths to still have correct behavior. This is really minor overhead. You might say now "but then I can append filepath to filepath". Well, as I said... we don't know whether it's a "filepath" at all. * merge `filename` and `dirname` into `basename` and make `parent` be `dirname`, so the function names match the name of the POSIX ones, which do (almost) the same... * fix a bug in `basename` (formerly `dirname`) which broke the type guarantees * add a pattern synonym for easier pattern matching without exporting the internal Path constructor
2016-03-08 21:53:42 +00:00