filediff-2.0.0: src/Filediff.hs
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE InstanceSigs #-}
-- | The module exposing the functionality of this package
module Filediff
( -- * lists
diffLists
, applyListDiff
-- * files
, diffFiles
, applyFileDiff
-- * directories
, diffDirectories
, diffDirectoriesWithIgnoredSubdirs
, applyDirectoryDiff
) where
import Debug.Trace
import qualified Data.HashMap as HMap
import Control.Concurrent (forkIO)
import Control.Concurrent.Thread as Thread (Result(..), result)
import Control.Concurrent.Thread.Group as ThreadGroup (new, forkIO, wait)
import qualified System.IO as IO
import qualified System.Directory as D
import qualified Data.Text as T
import qualified Data.Text.IO as TIO
import qualified Data.MemoCombinators as Memo
import Data.Either.Combinators
-- function imports
import Data.Maybe (isJust, fromJust, catMaybes)
import Data.List ((\\), intersect)
import Data.Monoid
import Control.Applicative
import Control.Monad
import Control.Monad.Trans.Either
import Control.Monad.IO.Class (liftIO)
-- Filediff imports
import Filediff.Types
import Filediff.Utils
( (</>)
, (<.>)
, getFileDirectory
, removeDotDirs
, createFileWithContents
, dropUntil
, removeFirstPathComponent
, removePathComponents
, getDirectoryContentsRecursiveSafe
, isPrefix
, dropPrefix
, dropInitialSlash
, dropTrailingSlash )
-- * basic operations
-- | /O(mn)/. Compute the difference between the two files (more
-- specifically, the minimal number of changes to make to transform the
-- file residing at the location specified by the first
-- parameter into the second). Throws an exception if either or both of
-- the parameters point to a directory, not a file.
--
-- Files are allowed to not exist at either or both of the parameters.
diffFiles :: FilePath -> FilePath -> IO Filediff
diffFiles a b = do
aIsDir <- D.doesDirectoryExist a
bIsDir <- D.doesDirectoryExist b
when (aIsDir || bIsDir) $ error $ "One or both of " ++ a ++ " and " ++ b ++ "is not a file, but a directory."
aExists <- D.doesFileExist a
bExists <- D.doesFileExist b
case (aExists, bExists) of
(False, False) -> return $ Filediff a b mempty
(False, True ) -> addCase
(True , False) -> delCase
(True , True ) -> modCase
where
addCase :: IO Filediff
addCase = do
let aLines = []
bLines <- T.lines <$> TIO.readFile b
return Filediff
{ base = a
, comp = b
, change = Add $ diffLists aLines bLines }
delCase :: IO Filediff
delCase = do
aLines <- T.lines <$> TIO.readFile a
let bLines = []
return Filediff
{ base = a
, comp = b
, change = Del $ diffLists aLines bLines }
modCase :: IO Filediff
modCase = do
aLines <- T.lines <$> TIO.readFile a
bLines <- T.lines <$> TIO.readFile b
return Filediff
{ base = a
, comp = b
, change = Mod $ diffLists aLines bLines }
-- | Compute the difference between the two directories (more
-- specifically, the minimal number of changes to make to transform the
-- directory residing at the location specified by the first
-- parameter into the second). Throws an exception if either or both of
-- the parameters point to a file, not a directory.
diffDirectories :: FilePath -> FilePath -> IO Diff
diffDirectories a b = diffDirectoriesWithIgnoredSubdirs a b [] []
-- | Diff two directories, ignoring some subdirectories. The first
-- `[FilePath]` parameter refers to the first `FilePath` parameter,
-- and same for the second, respectively.
diffDirectoriesWithIgnoredSubdirs :: FilePath -> FilePath -> [FilePath] -> [FilePath] -> IO Diff
diffDirectoriesWithIgnoredSubdirs a' b' aToIgnore bToIgnore = do
aIsFile <- D.doesFileExist a
bIsFile <- D.doesFileExist b
when (aIsFile || bIsFile) $ error $ "One or both of " ++ a ++ " and " ++ b ++ "is not a directory, but a file."
aContents' <- getDirectoryContentsRecursiveSafe a
bContents' <- getDirectoryContentsRecursiveSafe b
let aContents = filter (not . shouldIgnore aToIgnore) aContents'
let bContents = filter (not . shouldIgnore bToIgnore) bContents'
intersectionDiffs <- getDiffs $ intersect aContents bContents
aOnlyDiffs <- getDiffs $ aContents \\ bContents
bOnlyDiffs <- getDiffs $ bContents \\ aContents
let allDiffs = map makeRelative $ intersectionDiffs ++ aOnlyDiffs ++ bOnlyDiffs
return $ Diff allDiffs
where
a :: FilePath
a = dropTrailingSlash a'
b :: FilePath
b = dropTrailingSlash b'
-- | `x` is the prefix of the "base" of the diff; `y` is the
-- | "compare".
getDiffs :: [FilePath] -> IO [Filediff]
getDiffs filepaths
= filter (not . isIdentityFileDiff)
<$> mapM (\fp -> diffFiles (a </> fp) (b </> fp)) filepaths
isIdentityFileDiff :: Filediff -> Bool
isIdentityFileDiff = (==) mempty . change
makeRelative :: Filediff -> Filediff
makeRelative (Filediff base comp change) =
Filediff
(dropInitialSlash $ dropPrefix a base)
(dropInitialSlash $ dropPrefix b comp)
change
shouldIgnore :: [FilePath] -> FilePath -> Bool
shouldIgnore toIgnore filepath = any (flip isPrefix $ filepath) toIgnore
-- | /O(n)/. Apply a diff to a file. Returns the fail state if
-- application fails. For more on how diff application can fail,
-- see 'applyListDiff'.
applyFileDiff :: Filediff -> FilePath -> EitherT Error IO [Line]
applyFileDiff (Filediff _ _ change) filepath = do
case change of
Del _ -> delCase
Mod listdiff -> modCase listdiff
Add listdiff -> addCase listdiff
where
delCase :: EitherT Error IO [Line]
delCase = liftIO (D.removeFile filepath) >> right []
addCase :: ListDiff Line -> EitherT Error IO [Line]
addCase listDiff = liftIO (createFileWithContents filepath "") >> modCase listDiff
modCase :: ListDiff Line -> EitherT Error IO [Line]
modCase listDiff = do
-- Data.Text.IO.readFile is strict, which is what we
-- need, here (because of the write right after)
file <- liftIO (TIO.readFile filepath)
result <- hoistEither (applyListDiff listDiff . T.lines $ file)
liftIO (TIO.writeFile filepath (safeInit . T.unlines $ result))-- `init` for trailing \n
right result
safeInit :: T.Text -> T.Text
safeInit x = if T.null x then x else T.init x
-- | Applies a `Diff` to a directory. Returns the fail state if
-- any file application fails, but because of the parallelism in the
-- implementation, all file diffs will be attempted to be applied, so
-- if this fails, your directory will be left in an inconsistent state.
-- For more on how diff application can fail, see 'applyListDiff'.
applyDirectoryDiff :: Diff -> FilePath -> EitherT Error IO [[Line]]
applyDirectoryDiff (Diff filediffs) filepath
= EitherT
. fmap sequence
$ mapMParallelWaitForAll (runEitherT . apply) filediffs >>= mapM Thread.result
where
apply :: Filediff -> EitherT Error IO [Line]
apply diff@(Filediff base compare _)
= applyFileDiff diff (filepath </> base)
-- | forks the given 'IO' action for each element in the given list,
-- but waits for all to finish before returning.
mapMParallelWaitForAll :: (a -> IO b) -> [a] -> IO [Thread.Result b]
mapMParallelWaitForAll f list = do
group <- ThreadGroup.new
ioResults <- mapM (fmap snd . ThreadGroup.forkIO group . f) list
ThreadGroup.wait group
sequence ioResults
-- | Computes the minimal number of additions and deletions needed to
-- transform the first parameter into the second.
--
-- > λ diffLists "abcdefg" "wabxyze"
-- > ListDiff {dels = [(2,'c'),(3,'d'),(5,'f'),(6,'g')], adds = [(0,'w'),(3,'x'),(4,'y'),(5,'z')]}
diffLists :: forall a. (Eq a) => [a] -> [a] -> ListDiff a
diffLists a b = ListDiff
(map (\i -> (i, a !! i)) $ nonSubsequenceIndices common a)
(getProgressiveIndicesToAdd common b)
where
common :: [a]
common = longestCommonSubsequenceWrapper a b
-- | > λ add
-- > [(0,"w"),(3,"x"),(4,"y")]
-- > λ common
-- > ["a","b","e"]
getProgressiveIndicesToAdd :: [a] -> [a] -> [(Int, a)]
getProgressiveIndicesToAdd sub super =
map (\i -> (i, super !! i)) $ nonSubsequenceIndices sub super
-- | Applies a list diff. For example,
--
-- > ListDiff {dels = [(2,'c'),(3,'d'),(5,'f'),(6,'g')], adds = [(0,'w'),(3,'x'),(4,'y'),(5,'z')]}
-- > λ applyListDiff it "abcdefg"
-- > Right "wabxyze"
--
-- Returns a fail state if the diff cannot be applied. This can happen
-- for two reasons: first, the diff calls for a deletion at an index
-- but the element at that index doesn't match the element believed by
-- the to be diff at that index. Second, it can happen if the diff calls
-- for an element to be added at an index too large for the given input.
-- Here are respective examples of inputs that would trigger this case:
--
-- > let base = "abcdefg"
-- > let faultyBase = "ab*defg"
-- > let comp = "wabxyze"
-- > let listDiff = F.diffLists base comp
-- > F.applyListDiff listDiff faultyBase -- fails
--
-- and
--
-- > let base = "abcdefg"
-- > let faultyBase = "abcde"
-- > let comp = "wabxyzefgq"
-- > let listDiff = F.diffLists base comp
-- > F.applyListDiff listDiff faultyBase -- fails
applyListDiff :: forall a. (Eq a) => ListDiff a -> [a] -> Either Error [a]
applyListDiff l@(ListDiff dels adds) list =
removeAtIndices dels list >>= insertAtProgressiveIndices adds
-- | Best explained by example:
--
-- > λ insertAtProgressiveIndices [(1,'a'),(3,'b')] "def"
-- > Just "daebf"
insertAtProgressiveIndices :: [(Int, a)] -> [a] -> Either Error [a]
insertAtProgressiveIndices = insertAtProgressiveIndices' 0
insertAtProgressiveIndices' :: Int -> [(Int, a)] -> [a] -> Either Error [a]
insertAtProgressiveIndices' _ [] dest = Right dest
insertAtProgressiveIndices' curr src@((i,s):src') dest =
if i == curr
then (:) s <$> insertAtProgressiveIndices' (succ curr) src' dest
else case dest of
(d:dest') -> (:) d <$> insertAtProgressiveIndices' (succ curr) src dest'
[] -> Left "Fatal: couldn't apply list diff (application requires inserting at an index larger than the length of the list to which to apply the diff)."
-- all functions below are not exposed
-- don't hit the memotable if not necessary
-- | A wrapper around `longestCommonSubsequence`. It gives a bit of a
-- performance boost; it avoids hitting the memo table to an extent
-- (exactly how much depends on the arguments).
longestCommonSubsequenceWrapper :: forall a. (Eq a) => [a] -> [a] -> [a]
longestCommonSubsequenceWrapper xs ys =
if xs == ys
then xs -- (WLOG) don't want to return xs ++ xs
else commonPrefix
++ longestCommonSubsequence (getMiddle xs) (getMiddle ys)
++ commonSuffix
where
commonPrefix :: [a]
commonPrefix = getCommonPrefix xs ys
-- drop (length commonPrefix) to prevent the "abc" vs. "abc*abc" case
commonSuffix :: [a]
commonSuffix = reverse
(getCommonPrefix
(reverse (drop (length commonPrefix) xs))
(reverse (drop (length commonPrefix) ys)))
getCommonPrefix :: [a] -> [a] -> [a]
getCommonPrefix as bs = map fst . takeWhile (uncurry (==)) $ zip as bs
-- xs = abcd***efg
-- ys = abcd???????efg
-- getMiddle xs == ****
-- getMiddle ys = ??????
getMiddle :: [a] -> [a]
getMiddle elems = take (length elems - length commonPrefix - length commonSuffix) . drop (length commonPrefix) $ elems
-- | Compute the longest common (potentially noncontiguous) subsequence
-- between two sequences. Element type is fixed because memoization
-- requires a static type.
longestCommonSubsequence :: forall a. (Eq a) => [a] -> [a] -> [a]
longestCommonSubsequence xs ys = longestCommonSubsequence' 0 0
where
-- TODO: UArray?
xs' :: HMap.Map Int a
xs' = foldl update HMap.empty (zip [0..] xs)
ys' :: HMap.Map Int a
ys' = foldl update HMap.empty (zip [0..] ys)
update :: HMap.Map Int a -> (Int, a) -> HMap.Map Int a
update hmap (i, a) = HMap.insert i a hmap
xsLength :: Int
xsLength = length xs
ysLength :: Int
ysLength = length ys
longestCommonSubsequence' :: Int -> Int -> [a]
longestCommonSubsequence' = Memo.memo2 Memo.integral Memo.integral longestCommonSubsequence''
longestCommonSubsequence'' :: Int -> Int -> [a]
longestCommonSubsequence'' i j
| i == xsLength = []
| j == ysLength = []
| x == y = x : longestCommonSubsequence' (i + 1) (j + 1) -- WLOG
| length caseX > length caseY = caseX
| otherwise = caseY
where
x :: a
x = xs' HMap.! i
y :: a
y = ys' HMap.! j
caseX :: [a]
caseX = longestCommonSubsequence' (i + 1) j
caseY :: [a]
caseY = longestCommonSubsequence' i (j + 1)
-- | When `sub` is a (not necessarily contiguous) subsequence of `super`,
-- get the index at which each element of `sub` appears. E.g.
--
-- > λ subsequenceIndices "abe" "abcdefg"
-- > [0,1,4]
subsequenceIndices :: (Eq a) => [a] -> [a] -> [Int]
subsequenceIndices [] _ = []
subsequenceIndices _ [] = error "`sub` was not a subsequence of `super`"
subsequenceIndices sub@(a:sub') super@(b:super') =
if a == b
then 0 : map succ (subsequenceIndices sub' super')
else map succ (subsequenceIndices sub super')
-- | When `sub` is a (not necessarily contiguous) subsequence of `super`,
-- get the indices at which elements of `sub` do *not* appear. E.g.
--
-- > λ nonSubsequenceIndices "abe" "abcdefg"
-- > [2,3,5,6]
nonSubsequenceIndices :: (Eq a) => [a] -> [a] -> [Int]
nonSubsequenceIndices sub super =
[0..(length super - 1)] \\ (subsequenceIndices sub super)
-- | /O(n)/. `indices` parameter *must* be sorted in increasing order,
-- and indices must all exist.
removeAtIndices :: forall a. (Eq a) => [(Int, a)] -> [a] -> Either Error [a]
removeAtIndices dels list = if not matches
then Left "Fatal: couldn't apply list diff (application requires removing an element where the diff calls for a different element residing at that index)."
else Right (removeAtIndices' 0 (map fst dels) list)
where
matches :: Bool
matches = all (< length list) (map fst dels)
&& all (\(i, ch) -> (list !! i) == ch) dels
removeAtIndices' :: Int -> [Int] -> [a] -> [a]
removeAtIndices' _ [] xs = xs
removeAtIndices' curr (i:is) (x:xs) =
if curr == i
then removeAtIndices' (succ curr) is xs
else x : removeAtIndices' (succ curr) (i:is) xs