path-0.5.9/src/0000755000000000000000000000000012745423334011445 5ustar0000000000000000path-0.5.9/src/Path/0000755000000000000000000000000012730471672012343 5ustar0000000000000000path-0.5.9/test/0000755000000000000000000000000012745423573011642 5ustar0000000000000000path-0.5.9/src/Path.hs0000644000000000000000000002634212745423334012704 0ustar0000000000000000-- | -- Module : Path -- Copyright : © 2015–2016 FP Complete -- License : BSD 3 clause -- -- Maintainer : Chris Done -- Stability : experimental -- Portability : portable -- -- Support for well-typed paths. {-# LANGUAGE CPP #-} {-# LANGUAGE TemplateHaskell #-} {-# LANGUAGE DeriveDataTypeable #-} {-# LANGUAGE EmptyDataDecls #-} {-# LANGUAGE FlexibleInstances #-} module Path (-- * Types Path ,Abs ,Rel ,File ,Dir -- * Parsing ,parseAbsDir ,parseRelDir ,parseAbsFile ,parseRelFile ,PathParseException -- * Constructors ,mkAbsDir ,mkRelDir ,mkAbsFile ,mkRelFile -- * Operations ,() ,stripDir ,isParentOf ,parent ,filename ,dirname -- * Conversion ,toFilePath ,fromAbsDir ,fromRelDir ,fromAbsFile ,fromRelFile ) where import Control.Exception (Exception) import Control.Monad.Catch (MonadThrow(..)) import Data.Aeson (FromJSON (..)) import qualified Data.Aeson.Types as Aeson import Data.Data import Data.List import Data.Maybe import Language.Haskell.TH import Path.Internal import qualified System.FilePath as FilePath -------------------------------------------------------------------------------- -- Types -- | An absolute path. data Abs deriving (Typeable) -- | A relative path; one without a root. Note that a @.@ as well as any path -- starting with a @..@ is not a valid relative path. In other words, a -- relative path is always strictly under the directory tree to which it is -- relative. data Rel deriving (Typeable) -- | A file path. data File deriving (Typeable) -- | A directory path. data Dir deriving (Typeable) instance FromJSON (Path Abs File) where parseJSON = parseJSONWith parseAbsFile {-# INLINE parseJSON #-} instance FromJSON (Path Rel File) where parseJSON = parseJSONWith parseRelFile {-# INLINE parseJSON #-} instance FromJSON (Path Abs Dir) where parseJSON = parseJSONWith parseAbsDir {-# INLINE parseJSON #-} instance FromJSON (Path Rel Dir) where parseJSON = parseJSONWith parseRelDir {-# INLINE parseJSON #-} parseJSONWith :: (Show e, FromJSON a) => (a -> Either e b) -> Aeson.Value -> Aeson.Parser b parseJSONWith f x = do fp <- parseJSON x case f fp of Right p -> return p Left e -> fail (show e) {-# INLINE parseJSONWith #-} -- | Exception when parsing a location. data PathParseException = InvalidAbsDir FilePath | InvalidRelDir FilePath | InvalidAbsFile FilePath | InvalidRelFile FilePath | Couldn'tStripPrefixDir FilePath FilePath deriving (Show,Typeable) instance Exception PathParseException -------------------------------------------------------------------------------- -- Parsers -- | Convert an absolute 'FilePath' to a normalized absolute dir 'Path'. -- -- Throws: 'PathParseException' when the supplied path: -- -- * is not an absolute path -- * contains a @..@ anywhere in the path -- * is not a valid path (See 'System.FilePath.isValid') -- parseAbsDir :: MonadThrow m => FilePath -> m (Path Abs Dir) parseAbsDir filepath = if FilePath.isAbsolute filepath && not (hasParentDir filepath) && FilePath.isValid filepath then return (Path (normalizeDir filepath)) else throwM (InvalidAbsDir filepath) -- | Convert a relative 'FilePath' to a normalized relative dir 'Path'. -- -- Throws: 'PathParseException' when the supplied path: -- -- * is not a relative path -- * is any of @""@, @.@ or @..@ -- * contains @..@ anywhere in the path -- * is not a valid path (See 'System.FilePath.isValid') -- parseRelDir :: MonadThrow m => FilePath -> m (Path Rel Dir) parseRelDir filepath = if not (FilePath.isAbsolute filepath) && not (hasParentDir filepath) && not (null filepath) && filepath /= "." && (normalizeFilePath filepath) /= curDirNormalizedFP && filepath /= ".." && FilePath.isValid filepath then return (Path (normalizeDir filepath)) else throwM (InvalidRelDir filepath) -- | Convert an absolute 'FilePath' to a normalized absolute file 'Path'. -- -- Throws: 'PathParseException' when the supplied path: -- -- * is not an absolute path -- * has a trailing path separator -- * contains @..@ anywhere in the path -- * is not a valid path (See 'System.FilePath.isValid') -- parseAbsFile :: MonadThrow m => FilePath -> m (Path Abs File) parseAbsFile filepath = if FilePath.isAbsolute filepath && not (FilePath.hasTrailingPathSeparator filepath) && not (hasParentDir filepath) && FilePath.isValid filepath then return (Path (normalizeFilePath filepath)) else throwM (InvalidAbsFile filepath) -- | Convert a relative 'FilePath' to a normalized relative file 'Path'. -- -- Throws: 'PathParseException' when the supplied path: -- -- * is not a relative path -- * has a trailing path separator -- * is @""@, @.@ or @..@ -- * contains @..@ anywhere in the path -- * is not a valid path (See 'System.FilePath.isValid') -- parseRelFile :: MonadThrow m => FilePath -> m (Path Rel File) parseRelFile filepath = if not (FilePath.isAbsolute filepath || FilePath.hasTrailingPathSeparator filepath) && not (null filepath) && not (hasParentDir filepath) && filepath /= "." && filepath /= ".." && FilePath.isValid filepath then return (Path (normalizeFilePath filepath)) else throwM (InvalidRelFile filepath) -- | Helper function: check if the filepath has any parent directories in it. -- This handles the logic of checking for different path separators on Windows. hasParentDir :: FilePath -> Bool hasParentDir filepath' = ("/.." `isSuffixOf` filepath) || ("/../" `isInfixOf` filepath) || ("../" `isPrefixOf` filepath) where filepath = case FilePath.pathSeparator of '/' -> filepath' x -> map (\y -> if x == y then '/' else y) filepath' -------------------------------------------------------------------------------- -- Constructors -- | Make a 'Path Abs Dir'. -- -- Remember: due to the nature of absolute paths this (e.g. @\/home\/foo@) -- may compile on your platform, but it may not compile on another -- platform (Windows). mkAbsDir :: FilePath -> Q Exp mkAbsDir s = case parseAbsDir s of Left err -> error (show err) Right (Path str) -> [|Path $(return (LitE (StringL str))) :: Path Abs Dir|] -- | Make a 'Path Rel Dir'. mkRelDir :: FilePath -> Q Exp mkRelDir s = case parseRelDir s of Left err -> error (show err) Right (Path str) -> [|Path $(return (LitE (StringL str))) :: Path Rel Dir|] -- | Make a 'Path Abs File'. -- -- Remember: due to the nature of absolute paths this (e.g. @\/home\/foo@) -- may compile on your platform, but it may not compile on another -- platform (Windows). mkAbsFile :: FilePath -> Q Exp mkAbsFile s = case parseAbsFile s of Left err -> error (show err) Right (Path str) -> [|Path $(return (LitE (StringL str))) :: Path Abs File|] -- | Make a 'Path Rel File'. mkRelFile :: FilePath -> Q Exp mkRelFile s = case parseRelFile s of Left err -> error (show err) Right (Path str) -> [|Path $(return (LitE (StringL str))) :: Path Rel File|] -------------------------------------------------------------------------------- -- Conversion -- | Convert to a 'FilePath' type. -- -- All directories have a trailing slash, so if you want no trailing -- slash, you can use 'System.FilePath.dropTrailingPathSeparator' from -- the filepath package. toFilePath :: Path b t -> FilePath toFilePath (Path l) = l -- | Convert absolute path to directory to 'FilePath' type. fromAbsDir :: Path Abs Dir -> FilePath fromAbsDir = toFilePath -- | Convert relative path to directory to 'FilePath' type. fromRelDir :: Path Rel Dir -> FilePath fromRelDir = toFilePath -- | Convert absolute path to file to 'FilePath' type. fromAbsFile :: Path Abs File -> FilePath fromAbsFile = toFilePath -- | Convert relative path to file to 'FilePath' type. fromRelFile :: Path Rel File -> FilePath fromRelFile = toFilePath -------------------------------------------------------------------------------- -- Operations -- | Append two paths. -- -- The following cases are valid and the equalities hold: -- -- @$(mkAbsDir x) \<\/> $(mkRelDir y) = $(mkAbsDir (x ++ \"/\" ++ y))@ -- -- @$(mkAbsDir x) \<\/> $(mkRelFile y) = $(mkAbsFile (x ++ \"/\" ++ y))@ -- -- @$(mkRelDir x) \<\/> $(mkRelDir y) = $(mkRelDir (x ++ \"/\" ++ y))@ -- -- @$(mkRelDir x) \<\/> $(mkRelFile y) = $(mkRelFile (x ++ \"/\" ++ y))@ -- -- The following are proven not possible to express: -- -- @$(mkAbsFile …) \<\/> x@ -- -- @$(mkRelFile …) \<\/> x@ -- -- @x \<\/> $(mkAbsFile …)@ -- -- @x \<\/> $(mkAbsDir …)@ -- () :: Path b Dir -> Path Rel t -> Path b t () (Path a) (Path b) = Path (a ++ b) -- | Strip directory from path, making it relative to that directory. -- Throws 'Couldn'tStripPrefixDir' if directory is not a parent of the path. -- -- The following properties hold: -- -- @stripDir x (x \<\/> y) = y@ -- -- Cases which are proven not possible: -- -- @stripDir (a :: Path Abs …) (b :: Path Rel …)@ -- -- @stripDir (a :: Path Rel …) (b :: Path Abs …)@ -- -- In other words the bases must match. -- stripDir :: MonadThrow m => Path b Dir -> Path b t -> m (Path Rel t) stripDir (Path p) (Path l) = case stripPrefix p l of Nothing -> throwM (Couldn'tStripPrefixDir p l) Just "" -> throwM (Couldn'tStripPrefixDir p l) Just ok -> return (Path ok) -- | Is p a parent of the given location? Implemented in terms of -- 'stripDir'. The bases must match. -- -- The following properties hold: -- -- @not (x `isParentOf` x)@ -- -- @x `isParentOf` (x \<\/\> y)@ -- isParentOf :: Path b Dir -> Path b t -> Bool isParentOf p l = isJust (stripDir p l) -- | Take the absolute parent directory from the absolute path. -- -- The following properties hold: -- -- @parent (x \<\/> y) == x@ -- -- On the root, getting the parent is idempotent: -- -- @parent (parent \"\/\") = \"\/\"@ -- parent :: Path Abs t -> Path Abs Dir parent (Path fp) = Path (normalizeDir (FilePath.takeDirectory (FilePath.dropTrailingPathSeparator fp))) -- | Extract the file part of a path. -- -- The following properties hold: -- -- @filename (p \<\/> a) == filename a@ -- filename :: Path b File -> Path Rel File filename (Path l) = Path (FilePath.takeFileName l) -- | Extract the last directory name of a path. -- -- The following properties hold: -- -- @dirname (p \<\/> a) == dirname a@ -- dirname :: Path b Dir -> Path Rel Dir dirname (Path l) = Path (last (FilePath.splitPath l)) -------------------------------------------------------------------------------- -- Internal functions curDirNormalizedFP :: FilePath curDirNormalizedFP = '.' : [FilePath.pathSeparator] -- | Internal use for normalizing a directory. normalizeDir :: FilePath -> FilePath normalizeDir = FilePath.addTrailingPathSeparator . normalizeFilePath normalizeFilePath :: FilePath -> FilePath #if defined(mingw32_HOST_OS) || defined(__MINGW32__) || MIN_VERSION_filepath(1,4,0) normalizeFilePath = FilePath.normalise #else normalizeFilePath = normalizeLeadingSeparators . FilePath.normalise where sep = FilePath.pathSeparator normalizeLeadingSeparators (x1:x2:xs) | x1 == sep && x2 == sep = normalizeLeadingSeparators (sep:xs) normalizeLeadingSeparators x = x #endif path-0.5.9/src/Path/Internal.hs0000644000000000000000000000254512730471672014461 0ustar0000000000000000{-# LANGUAGE CPP #-} {-# LANGUAGE DeriveDataTypeable #-} -- | Internal types and functions. module Path.Internal (Path(..)) where import Control.DeepSeq (NFData (..)) import Data.Aeson (ToJSON (..)) import Data.Data -- | Path of some base and type. -- -- Internally is a string. The string can be of two formats only: -- -- 1. File format: @file.txt@, @foo\/bar.txt@, @\/foo\/bar.txt@ -- 2. Directory format: @foo\/@, @\/foo\/bar\/@ -- -- All directories end in a trailing separator. There are no duplicate -- path separators @\/\/@, no @..@, no @.\/@, no @~\/@, etc. newtype Path b t = Path FilePath deriving (Typeable) -- | String equality. -- -- The following property holds: -- -- @show x == show y ≡ x == y@ instance Eq (Path b t) where (==) (Path x) (Path y) = x == y -- | String ordering. -- -- The following property holds: -- -- @show x \`compare\` show y ≡ x \`compare\` y@ instance Ord (Path b t) where compare (Path x) (Path y) = compare x y -- | Same as 'Path.toFilePath'. -- -- The following property holds: -- -- @x == y ≡ show x == show y@ instance Show (Path b t) where show (Path x) = show x instance NFData (Path b t) where rnf (Path x) = rnf x instance ToJSON (Path b t) where toJSON (Path x) = toJSON x {-# INLINE toJSON #-} #if MIN_VERSION_aeson(0,10,0) toEncoding (Path x) = toEncoding x {-# INLINE toEncoding #-} #endif path-0.5.9/test/Main.hs0000644000000000000000000002045612745423573013071 0ustar0000000000000000{-# LANGUAGE TemplateHaskell #-} -- | Test suite. module Main where import Control.Applicative import Control.Monad import Data.Aeson import qualified Data.ByteString.Lazy.Char8 as LBS import Data.Maybe import Path import Path.Internal import Test.Hspec -- | Test suite entry point, returns exit failure if any test fails. main :: IO () main = hspec spec -- | Test suite. spec :: Spec spec = do describe "Parsing: Path Abs Dir" parseAbsDirSpec describe "Parsing: Path Rel Dir" parseRelDirSpec describe "Parsing: Path Abs File" parseAbsFileSpec describe "Parsing: Path Rel File" parseRelFileSpec describe "Operations: ()" operationAppend describe "Operations: stripDir" operationStripDir describe "Operations: isParentOf" operationIsParentOf describe "Operations: parent" operationParent describe "Operations: filename" operationFilename describe "Restrictions" restrictions describe "Aeson Instances" aesonInstances -- | Restricting the input of any tricks. restrictions :: Spec restrictions = do -- These ~ related ones below are now lifted: -- https://github.com/chrisdone/path/issues/19 parseSucceeds "~/" (Path "~/") parseSucceeds "~/foo" (Path "~/foo/") parseSucceeds "~/foo/bar" (Path "~/foo/bar/") -- parseFails "../" parseFails ".." parseFails "." parseFails "/.." parseFails "/foo/../bar/" parseFails "/foo/bar/.." where parseFails x = it (show x ++ " should be rejected") (isNothing (void (parseAbsDir x) <|> void (parseRelDir x) <|> void (parseAbsFile x) <|> void (parseRelFile x))) parseSucceeds x with = parserTest parseRelDir x (Just with) -- | The 'filename' operation. operationFilename :: Spec operationFilename = do it "filename ($(mkAbsDir parent) filename $(mkRelFile filename)) == $(mkRelFile filename)" (filename ($(mkAbsDir "/home/chris/") filename $(mkRelFile "bar.txt")) == $(mkRelFile "bar.txt")) it "filename ($(mkRelDir parent) filename $(mkRelFile filename)) == $(mkRelFile filename)" (filename ($(mkRelDir "home/chris/") filename $(mkRelFile "bar.txt")) == $(mkRelFile "bar.txt")) -- | The 'parent' operation. operationParent :: Spec operationParent = do it "parent (parent child) == parent" (parent ($(mkAbsDir "/foo") $(mkRelDir "bar")) == $(mkAbsDir "/foo")) it "parent \"\" == \"\"" (parent $(mkAbsDir "/") == $(mkAbsDir "/")) it "parent (parent \"\") == \"\"" (parent (parent $(mkAbsDir "/")) == $(mkAbsDir "/")) -- | The 'isParentOf' operation. operationIsParentOf :: Spec operationIsParentOf = do it "isParentOf parent (parent child)" (isParentOf $(mkAbsDir "///bar/") ($(mkAbsDir "///bar/") $(mkRelFile "bar/foo.txt"))) it "isParentOf parent (parent child)" (isParentOf $(mkRelDir "bar/") ($(mkRelDir "bar/") $(mkRelFile "bob/foo.txt"))) -- | The 'stripDir' operation. operationStripDir :: Spec operationStripDir = do it "stripDir parent (parent child) = child" (stripDir $(mkAbsDir "///bar/") ($(mkAbsDir "///bar/") $(mkRelFile "bar/foo.txt")) == Just $(mkRelFile "bar/foo.txt")) it "stripDir parent (parent child) = child" (stripDir $(mkRelDir "bar/") ($(mkRelDir "bar/") $(mkRelFile "bob/foo.txt")) == Just $(mkRelFile "bob/foo.txt")) it "stripDir parent parent = _|_" (stripDir $(mkAbsDir "/home/chris/foo") $(mkAbsDir "/home/chris/foo") == Nothing) -- | The '' operation. operationAppend :: Spec operationAppend = do it "AbsDir + RelDir = AbsDir" ($(mkAbsDir "/home/") $(mkRelDir "chris") == $(mkAbsDir "/home/chris/")) it "AbsDir + RelFile = AbsFile" ($(mkAbsDir "/home/") $(mkRelFile "chris/test.txt") == $(mkAbsFile "/home/chris/test.txt")) it "RelDir + RelDir = RelDir" ($(mkRelDir "home/") $(mkRelDir "chris") == $(mkRelDir "home/chris")) it "RelDir + RelFile = RelFile" ($(mkRelDir "home/") $(mkRelFile "chris/test.txt") == $(mkRelFile "home/chris/test.txt")) -- | Tests for the tokenizer. parseAbsDirSpec :: Spec parseAbsDirSpec = do failing "" failing "./" failing "foo.txt" succeeding "/" (Path "/") succeeding "//" (Path "/") succeeding "///foo//bar//mu/" (Path "/foo/bar/mu/") succeeding "///foo//bar////mu" (Path "/foo/bar/mu/") succeeding "///foo//bar/.//mu" (Path "/foo/bar/mu/") where failing x = parserTest parseAbsDir x Nothing succeeding x with = parserTest parseAbsDir x (Just with) -- | Tests for the tokenizer. parseRelDirSpec :: Spec parseRelDirSpec = do failing "" failing "/" failing "//" succeeding "~/" (Path "~/") -- https://github.com/chrisdone/path/issues/19 failing "/" failing "./" failing "././" failing "//" failing "///foo//bar//mu/" failing "///foo//bar////mu" failing "///foo//bar/.//mu" succeeding "..." (Path ".../") succeeding "foo.bak" (Path "foo.bak/") succeeding "./foo" (Path "foo/") succeeding "././foo" (Path "foo/") succeeding "./foo/./bar" (Path "foo/bar/") succeeding "foo//bar//mu//" (Path "foo/bar/mu/") succeeding "foo//bar////mu" (Path "foo/bar/mu/") succeeding "foo//bar/.//mu" (Path "foo/bar/mu/") where failing x = parserTest parseRelDir x Nothing succeeding x with = parserTest parseRelDir x (Just with) -- | Tests for the tokenizer. parseAbsFileSpec :: Spec parseAbsFileSpec = do failing "" failing "./" failing "~/" failing "./foo.txt" failing "/" failing "//" failing "///foo//bar//mu/" succeeding "/..." (Path "/...") succeeding "/foo.txt" (Path "/foo.txt") succeeding "///foo//bar////mu.txt" (Path "/foo/bar/mu.txt") succeeding "///foo//bar/.//mu.txt" (Path "/foo/bar/mu.txt") where failing x = parserTest parseAbsFile x Nothing succeeding x with = parserTest parseAbsFile x (Just with) -- | Tests for the tokenizer. parseRelFileSpec :: Spec parseRelFileSpec = do failing "" failing "/" failing "//" failing "~/" failing "/" failing "./" failing "//" failing "///foo//bar//mu/" failing "///foo//bar////mu" failing "///foo//bar/.//mu" succeeding "..." (Path "...") succeeding "foo.txt" (Path "foo.txt") succeeding "./foo.txt" (Path "foo.txt") succeeding "././foo.txt" (Path "foo.txt") succeeding "./foo/./bar.txt" (Path "foo/bar.txt") succeeding "foo//bar//mu.txt" (Path "foo/bar/mu.txt") succeeding "foo//bar////mu.txt" (Path "foo/bar/mu.txt") succeeding "foo//bar/.//mu.txt" (Path "foo/bar/mu.txt") where failing x = parserTest parseRelFile x Nothing succeeding x with = parserTest parseRelFile x (Just with) -- | Parser test. parserTest :: (Show a1,Show a,Eq a1) => (a -> Maybe a1) -> a -> Maybe a1 -> SpecWith () parserTest parser input expected = it ((case expected of Nothing -> "Failing: " Just{} -> "Succeeding: ") ++ "Parsing " ++ show input ++ " " ++ case expected of Nothing -> "should fail." Just x -> "should succeed with: " ++ show x) (actual `shouldBe` expected) where actual = parser input -- | Tests for the 'ToJSON' and 'FromJSON' instances -- -- Can't use overloaded strings due to some weird issue with bytestring-0.9.2.1 / ghc-7.4.2: -- https://travis-ci.org/sjakobi/path/jobs/138399072#L989 aesonInstances :: Spec aesonInstances = do it "Decoding \"[\"/foo/bar\"]\" as a [Path Abs Dir] should succeed." $ eitherDecode (LBS.pack "[\"/foo/bar\"]") `shouldBe` Right [Path "/foo/bar/" :: Path Abs Dir] it "Decoding \"[\"/foo/bar\"]\" as a [Path Rel Dir] should fail." $ decode (LBS.pack "[\"/foo/bar\"]") `shouldBe` (Nothing :: Maybe [Path Rel Dir]) it "Encoding \"[\"/foo/bar/mu.txt\"]\" should succeed." $ encode [Path "/foo/bar/mu.txt" :: Path Abs File] `shouldBe` (LBS.pack "[\"/foo/bar/mu.txt\"]") path-0.5.9/LICENSE0000644000000000000000000000272412666307154011673 0ustar0000000000000000Copyright (c) 2015–2016, FP Complete All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: * Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. * Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. * Neither the name of paths nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. path-0.5.9/Setup.hs0000644000000000000000000000005612624136666012320 0ustar0000000000000000import Distribution.Simple main = defaultMain path-0.5.9/path.cabal0000644000000000000000000000223712745423627012607 0ustar0000000000000000name: path version: 0.5.9 synopsis: Support for well-typed paths description: Support for well-typed paths. license: BSD3 license-file: LICENSE author: Chris Done maintainer: Chris Done copyright: 2015–2016 FP Complete category: Filesystem build-type: Simple cabal-version: >=1.8 extra-source-files: README.md, CHANGELOG library hs-source-dirs: src/ ghc-options: -Wall -O2 exposed-modules: Path, Path.Internal build-depends: base >= 4 && <5 , exceptions , filepath < 1.2.0.1 || >= 1.3 , template-haskell , deepseq , aeson test-suite test type: exitcode-stdio-1.0 main-is: Main.hs hs-source-dirs: test build-depends: HUnit , aeson , base , bytestring , hspec , mtl , path source-repository head type: git location: https://github.com/chrisdone/path.git path-0.5.9/README.md0000644000000000000000000004075412666307154012152 0ustar0000000000000000# Path Support for well-typed paths in Haskell. * [Motivation](#motivation) * [Approach](#approach) * [Solution](#solution) * [Implementation](#implementation) * [The data types](#the-data-types) * [Parsers](#parsers) * [Smart constructors](#smart-constructors) * [Overloaded stings](#overloaded-strings) * [Operations](#operations) * [Review](#review) * [Relative vs absolute confusion](#relative-vs-absolute-confusion) * [The equality problem](#the-equality-problem) * [Unpredictable concatenation issues](#unpredictable-concatenation-issues) * [Confusing files and directories](#confusing-files-and-directories) * [Self-documentation](#self-documentation) * [In practice](#in-practice) * [Doing I/O](#doing-io) * [Doing textual manipulations](#doing-textual-manipulations) * [Accepting user input](#accepting-user-input) * [Comparing with existing path libraries](#comparing-with-existing-path-libraries) * [filepath and system-filepath](#filepath-and-system-filepath) * [system-canonicalpath, canonical-filepath, directory-tree](#system-canonicalpath-canonical-filepath-directory-tree) * [pathtype](#pathtype) * [data-filepath](#data-filepath) * [Summary](#summary) ## Motivation It was after working on a number of projects at FP Complete that use file paths in various ways. We used the system-filepath package, which was supposed to solve many path problems by being an opaque path type. It occurred to me that the same kind of bugs kept cropping up: * Expected a path to be absolute but it was relative, or vice-versa. * Expected two equivalent paths to be equal or order the same, but they did not (`/home//foo` vs `/home/foo/` vs `/home/bar/../foo`, etc.). * Unpredictable behaviour with regards to concatenating paths. * Confusing files and directories. * Not knowing whether a path was a file or directory or relative or absolute based on the type alone was a drag. All of these bugs are preventable. ## Approach My approach to problems like this is to make a type that encodes the properties I want and then make it impossible to let those invariants be broken, without compromise or backdoors to let the wrong value “slip in”. Once I have a path, I want to be able to trust it fully. This theme will be seen throughout the things I lay out below. ## Solution After having to fix bugs due to these in our software, I put my foot down and made: * An opaque `Path` type (a newtype wrapper around `String`). * Smart constructors which are very stringent in the parsing. * Make the parsers highly normalizing. * Leave equality and concatenation to basic string equality and concatenation. * Include relativity (absolute/relative) and type (directory/file) in the type itself. * Use the already cross-platform [filepath](http://hackage.haskell.org/package/filepath) package for implementation details. ## Implementation ### The data types Here is the type: ```haskell newtype Path b t = Path FilePath deriving (Typeable) ``` The type variables are: * `b` — base, the base location of the path; absolute or relative. * `t` — type, whether file or directory. The base types can be filled with these: ```haskell data Abs deriving (Typeable) data Rel deriving (Typeable) ``` And the type can be filled with these: ```haskell data File deriving (Typeable) data Dir deriving (Typeable) ``` (Why not use data kinds like `data Type = File | Dir`? Because that imposes an extension overhead of adding `{-# LANGUAGE DataKinds #-}` to every module you might want to write out a path type in. Given that one cannot construct paths of types other than these, via the operations in the module, it’s not a concern for me.) There is a conversion function to give you back the filepath: ```haskell toFilePath :: Path b t -> FilePath toFilePath (Path l) = l ``` Beginning from version 0.5.3, there are type-constrained versions of `toFilePath` with the following signatures: ```haskell fromAbsDir :: Path Abs Dir -> FilePath fromRelDir :: Path Rel Dir -> FilePath fromAbsFile :: Path Abs File -> FilePath fromRelFile :: Path Rel File -> FilePath ``` ### Parsers To get a `Path` value, you need to use one of the four parsers: ```haskell parseAbsDir :: MonadThrow m => FilePath -> m (Path Abs Dir) parseRelDir :: MonadThrow m => FilePath -> m (Path Rel Dir) parseAbsFile :: MonadThrow m => FilePath -> m (Path Abs File) parseRelFile :: MonadThrow m => FilePath -> m (Path Rel File) ``` The following properties apply: * Absolute parsers will reject non-absolute paths. * The only delimiter syntax accepted is the path separator; `/` on POSIX and `\` on Windows. * Any other delimiter is rejected; `..`, `~/`, `/./`, etc. * All parsers normalize into single separators: `/home//foo` → `/home/foo`. * Directory parsers always normalize with a final trailing `/`. So `/home/foo` parses into the string `/home/foo/`. It was discussed briefly whether we should just have a class for parsing rather than four separate parsing functions. In my experience so far, I have had type errors where I wrote something `like x <- parseAbsDir someAbsDirString` because `x` was then passed to a place that expected a relative directory. In this way, overloading the return value would’ve just been accepted. So I don’t think having a class is a good idea. Being explicit here doesn’t exactly waste our time, either. Why are these functions in `MonadThrow`? Because it means I can have it return an `Either`, or a `Maybe`, if I’m in pure code, and if I’m in `IO`, and I don’t expect parsing to ever fail, I can use it in IO like this: ```haskell do x <- parseRelFile (fromCabalFileName x) foo x … ``` That’s really convenient and we take advantage of this at FP Complete a lot. The instances Equality, ordering and printing are simply re-using the `String` instances: ```haskell instance Eq (Path b t) where (==) (Path x) (Path y) = x == y instance Ord (Path b t) where compare (Path x) (Path y) = compare x y instance Show (Path b t) where show (Path x) = show x ``` Which gives us for free the following equational properties: ```haskell toFilePath x == toFilePath y ≡ x == y -- Eq instance toFilePath x `compare` toFilePath y ≡ x `compare` y -- Ord instance toFilePath x == toFilePath y ≡ show x == show y -- Show instance ``` In other words, the representation and the path you get out at the end are the same. Two paths that are equal will always give you back the same thing. ### Smart constructors For when you know what a path will be at compile-time, there are constructors for that: ```haskell $(mkAbsDir "/home/chris") $(mkRelDir "chris") $(mkAbsFile "/home/chris/x.txt") $(mkRelFile "chris/x.txt") ``` These will run at compile-time and underneath use the appropriate parser. ### Overloaded strings No `IsString` instance is provided, because that has no way to statically determine whether the path is correct, and would otherwise have to be a partial function. In practice I have written the wrong path format in a `$(mk… "")` and been thankful it was caught early. ### Operations There is path concatenation: ```haskell () :: Path b Dir -> Path Rel t -> Path b t ``` Get the parent directory of a path: ```haskell parent :: Path Abs t -> Path Abs Dir ``` Get the filename of a file path: ```haskell filename :: Path b File -> Path Rel File ``` Get the directory name of a directory path: ```haskell dirname :: Path b Dir -> Path Rel Dir ``` Stripping the parent directory from a path: ```haskell stripDir :: MonadThrow m => Path b Dir -> Path b t -> m (Path Rel t) ``` ## Review Let’s review my initial list of complaints and see if they’ve been satisfied. ### Relative vs absolute confusion Paths now distinguish in the type system whether they are relative or absolute. You can’t append two absolute paths, for example: ```haskell λ> $(mkAbsDir "/home/chris") $(mkAbsDir "/home/chris") :23:31-55: Couldn't match type ‘Abs’ with ‘Rel’ ``` ### The equality problem Paths are now stringently normalized. They have to be a valid path, and they only support single path separators, and all directories are suffixed with a trailing path separator: ```haskell λ> $(mkAbsDir "/home/chris//") == $(mkAbsDir "/./home//chris") True λ> toFilePath $(mkAbsDir "/home/chris//") == toFilePath $(mkAbsDir "/./home//chris") True λ> ($(mkAbsDir "/home/chris//"),toFilePath $(mkAbsDir "/./home//chris")) ("/home/chris/","/home/chris/") ``` ### Unpredictable concatenation issues Because of the stringent normalization, path concatenation, as seen above, is simply string concatenation. This is about as predictable as it can get: ```haskell λ> toFilePath $(mkAbsDir "/home/chris//") "/home/chris/" λ> toFilePath $(mkRelDir "foo//bar") "foo/bar/" λ> $(mkAbsDir "/home/chris//") $(mkRelDir "foo//bar") "/home/chris/foo/bar/" ``` ### Confusing files and directories Now that the path type is encoded in the type system, our `` operator prevents improper appending: ```haskell λ> $(mkAbsDir "/home/chris/") $(mkRelFile "foo//bar") "/home/chris/foo/bar" λ> $(mkAbsFile "/home/chris") $(mkRelFile "foo//bar") :35:1-26: Couldn't match type ‘File’ with ‘Dir’ ``` ### Self-documentation Now I can read the path like: ```haskell { fooPath :: Path Rel Dir, ... } ``` And know that this refers to the directory relative to some other path, meaning I should be careful to consider the current directory when using this in IO, or that I’ll probably need a parent to append to it at some point. ## In practice We’ve been using this at FP Complete in a number of packages for some months now, it’s turned out surprisingly sufficient for most of our path work with only one bug found. We weren’t sure initially whether it would just be too much of a pain to use, but really it’s quite acceptable given the advantages. You can see its use all over the [`stack`](https://github.com/commercialhaskell/stack) codebase. ## Doing I/O Currently any operations involving I/O can be done by using the existing I/O library: ```haskell doesFileExist (toFilePath fp) readFile (toFilePath fp) ``` etc. This has problems with respect to accidentally running something like: ```haskell doesFileExist $(mkRelDir "foo") ``` But I/O is currently outside the scope of what this package solves. Once you leave the realm of the `Path` type invariants are back to your responsibility. As with the original version of this library, we’re currently building up a set of functions in a `Path.IO` module over time that fits our real-world use-cases. It may or may not appear in the path package eventually. It’ll need cleaning up and considering what should really be included. **Edit:** There is now [`path-io`](https://hackage.haskell.org/package/path-io) package that complements the `path` library and includes complete well-typed interface to [`directory`](https://hackage.haskell.org/package/directory) and [`temporary`](https://hackage.haskell.org/package/temporary). There is work to add more generally useful functions from Stack's `Path.IO` to it and make Stack depend on the `path-io` package. ## Doing textual manipulations One problem that crops up sometimes is wanting to manipulate paths. Currently the way we do it is via the filepath library and re-parsing the path: ```haskell parseAbsFile . addExtension "/directory/path" "ext" . toFilePath ``` It doesn’t happen too often, in our experience, to the extent this needs to be more convenient. ## Accepting user input Sometimes you have user input that contains `../`. The solution we went with is to have a function like `resolveDir`: ```haskell resolveDir :: (MonadIO m, MonadThrow m) => Path Abs Dir -> FilePath -> m (Path Abs Dir) ``` Which will call `canonicalizePath` which collapses and normalizes a path and then we parse with regular old `parseAbsDir` and we’re cooking with gas. This and others like it might get added to the `path` package. ## Comparing with existing path libraries ### filepath and system-filepath The [filepath](http://hackage.haskell.org/package/filepath) package is intended as the complimentary package to be used before parsing into a Path value, and/or after printing from a Path value. The package itself contains no type-safety, instead contains a range of cross-platform textual operations. Definitely reach for this library when you want to do more involved manipulations. The `system-filepath` package is deprecated in favour of `filepath`. ### system-canonicalpath, canonical-filepath, directory-tree The [`system-canonicalpath`](http://hackage.haskell.org/package/system-canonicalpath) and the [`canonical-filepath`](http://hackage.haskell.org/package/canonical-filepath) packages both are a kind of subset of `path`. They canonicalize a string into an opaque path, but neither distinguish directories from files or absolute/relative. Useful if you just want a canonical path but doesn’t do anything else. The [`directory-tree`](http://hackage.haskell.org/package/directory-tree) package contains a sum type of dir/file/etc but doesn’t distinguish in its operations relativity or path type. ### pathtype Finally, we come to a path library that path is similar to: the [`pathtype`](http://hackage.haskell.org/package/pathtype) library. There are the same types of `Path Abs File` / `Path Rel Dir`, etc. The points where this library isn’t enough for me are: * There is an `IsString` instance, which means people will use it, and will make mistakes. * Paths are not normalized into a predictable format, leading to me being unsure when equality will succeed. This is the same problem I encountered in `system-filepath`. The equality function normalizes, but according to what properties I can reason about? I don’t know. ```haskell System.Path.Posix> ("/tmp//" :: Path a Dir) == ("/tmp" :: Path a Dir) True System.Path.Posix> ("tmp" :: Path a Dir) == ("/tmp" :: Path a Dir) True System.Path.Posix> ("/etc/passwd/" :: Path a b) == ("/etc/passwd" :: Path a b) True System.Path.Posix> ("/tmp//" :: Path Abs Dir) == ("/tmp/./" :: Path Abs Dir) False System.Path.Posix> ("/tmp/../" :: Path Abs Dir) == ("/" :: Path Abs Dir) False ``` * Empty string should not be allowed, and introduction of `.` due to that gets weird: ```haskell System.Path.Posix> fmap getPathString (Right ("." :: Path Rel File)) Right "." System.Path.Posix> fmap getPathString (mkPathAbsOrRel "") Right "." System.Path.Posix> (Right ("." :: Path Rel File)) == (mkPathAbsOrRel "") False System.Path.Posix> takeDirectory ("tmp" :: Path Rel Dir) . System.Path.Posix> (getPathString ("." :: Path Rel File) == getPathString ("" :: Path Rel File)) True System.Path.Posix> (("." :: Path Rel File) == ("" :: Path Rel File)) False ``` * It has functions like `<.>/addExtension` which lets you insert an arbitrary string into a path. * Some functions let you produce nonsense (could be prevented by a stricter type), for example: ```haskell System.Path.Posix> takeFileName ("/tmp/" :: Path Abs Dir) tmp ``` I’m being a bit picky here, a bit unfair. But the point is really to show the kind of things I tried to avoid in `path`. In summary, it’s just hard to know where things can go wrong, similar to what was going on in `system-filepath`. ### data-filepath The [`data-filepath`](https://hackage.haskell.org/package/data-filepath) is also very similar, I discovered it after writing my own at work and was pleased to see it’s mostly the same. The main differences are: * Uses `DataKinds` for the relative/absolute and file/dir distinction which as I said above is an overhead. * Uses a GADT for the path type, which is fine. In my case I wanted to retain the original string which functions that work on the `FilePath` (`String`) type already deal with well. It does change the parsing step somewhat, because it parses into segments. * It’s more lenient at parsing (allowing `..` and trailing `.`). The API is a bit awkward to just parse a directory, requires a couple functions to get it (going via `WeakFilePath`), returning only an `Either`, and there are no functions like parent. But there’s not much to complain about. It’s a fine library, but I didn’t feel the need to drop my own in favor of it. Check it out and decide for yourself. ## Summary There’s a growing interest in making practical use of well-typed file path handling. I think everyone’s wanted it for a while, but few people have really committed to it in practice. Now that I’ve been using `path` for a while, I can’t really go back. It’ll be interesting to see what new packages crop up in the coming year, I expect there’ll be more. path-0.5.9/CHANGELOG0000644000000000000000000000134512745423651012075 0ustar00000000000000000.5.9: * Lifted ~ restriction from parser https://github.com/chrisdone/path/issues/19 0.5.8 * Add Aeson instances. 0.5.7: * Fix haddock problem. 0.5.6: * Reject only .. and . 0.5.5: * Use filepath's isValid function for additional sanity checks 0.5.4: * Disable parsing of path consisting only of "." * Add NFData instance for Path * Some typo/docs improvements * Add standard headers to modules 0.5.3: * Added conversion functions. 0.2.0: * Rename parentAbs to simply parent. * Add dirname. 0.3.0: * Removed Generic instance. 0.4.0: * Implemented stricter parsing, disabling use of "..". * Made stripDir generic over MonadThrow 0.5.0: * Fix stripDir p p /= Nothing bug. 0.5.2: * Removed unused DeriveGeneric.