a bit of finalization

README.md

@@ -0,0 +1,170 @@
+
+# adiff (arbitrary-tokens diff, patch and merge)
+
+This is a half-working pre-alpha version, use with care.
+
+### Short summary
+
+The main aim of this toolbox is to help with finding differences in text
+formats that do not have a fixed "line-by-line" semantics, as assumed by
+standard unix `diff` and related tools.
+
+The problem was previously tackled by Arek Antoniewicz on MFF CUNI, who
+produced a working software package in C++, and designed the Regex-edged DFAs
+(REDFAs) that were used for user-specifiable tokenization of the input. The
+work on the corresponding thesis is finished.
+
+This started as a simple Haskell port of that work, and packed some relatively
+orthogonal improvements (mainly the histogram-style diffing). I later got rid
+of the REDFA concept -- while super-interesting and useful in theory, I didn't
+find a sufficiently universal way to build good lexers from user-specified
+strings.  Having a proper Regex representation library (so that e.g.
+reconstructing Flex is easy) would help a lot.
+
+### TODO list
+
+- Implement `patch` functionality, mainly patchfile parsing and fuzzy matching
+  of hunk context. `diff` and `diff3` works.
+- Implement a splitting heuristic for diffs, so that diffing of large files
+  doesn't take aeons
+- check if we can have external lexers, unix-style
+
+# How-To
+
+Install using `cabal`. The `adiff` program has 3 sub-commands that work like
+`diff`, `patch` and `diff3`.
+
+## Example
+
+Let's have a file `orig`:
+```
+Roses are red. Violets are blue.
+Patch is quite hard. I cannot rhyme.
+```
+
+and a modified file `mine`:
+```
+Roses are red. Violets are blue.
+Patching is hard. I still cannot rhyme.
+```
+
+Let's use the `words` lexer, which marks everything whitespace-ish as
+whitespace, and picks up groups of non-whitespace "content" characters.
+
+Diffing the 2 files gets done as such:
+```
+ $ cabal run adiff -- -l words diff orig mine
+```
+
+You should get something like this:
+```
+@@ -7 +7 @@
+ . 
+ |are
+ . 
+ |blue.
+ .\n
+-|Patch
++|Patching
+ . 
+ |is
+-. 
+-|quite
+ . 
+ |hard.
+ . 
+ |I
++. 
++|still
+ . 
+ |cannot
+ . 
+ |rhyme.
+ .\n
+```
+
+Let's pretend someone has sent us a new version, with a better formated verse
+and some other improvements, in file `yours`:
+```
+Roses are red.
+Violets are blue.
+Patch is quite hard.
+I cannot do verses.
+```
+
+We can run `diff3` to get a patch with both changes, optionally with reduced
+context:
+```
+ $ cabal run adiff -- -l words diff3 mine orig yours -C1
+```
+...which outputs:
+```
+@@ -4 +4 @@
+ |red.
+-. 
++.\n
+ |Violets
+@@ -11 +11 @@
+ .\n
+-|Patch
++|Patching
+ . 
+ |is
+-. 
+-|quite
+ . 
+ |hard.
+-. 
++.\n
+ |I
++. 
++|still
+ . 
+@@ -23 +23 @@
+ . 
+-|rhyme.
++|do
++. 
++|verses.
+ .\n
+```
+
+...or get a merged output right away, using the `-m`/`--merge` option:
+```
+Roses are red.
+Violets are blue.
+Patching is hard.
+I still cannot do verses.
+```
+
+...or completely ignore whatever whitespace changes that the people decided to
+do for whatever reason, with `-i`/`--ignore-whitespace` (also works without
+`-m`):
+```
+Roses are red. Violets are blue.
+Patching is hard. I still cannot do verses.
+```
+
+If there's a conflict (substituing the `Patch` to `Merging` in file `yours`), it gets highlighted in the merged diff as such:
+```
+[...]
+ . 
+ |blue.
+ .\n
+<|Patching
+=|Patch
+>|Merging
+ . 
+ |is
+-. 
+-|quite
+[...]
+```
+
+and using the standard conflict marks in the merged output:
+```
+Roses are red.
+Violets are blue.
+<<<<<<<Patching|||||||Patch=======Merging>>>>>>> is hard.
+I still cannot do verses.
+```

adiff.cabal

@@ -42,6 +42,8 @@ executable adiff
   -- .hs or .lhs file containing the Main module.
   main-is: Main.hs
 
+  ghc-options: -O2 -Wall
+
   -- Modules included in this executable, other than Main.
   other-modules: Diff,
                  Diff3,
@@ -49,8 +51,7 @@ executable adiff
                  Hunks,
                  Merge,
                  Patch,
-                 Redfa,
+                 Tokenizers,
-                 Substr,
                  Types,
                  Version
 
@@ -58,13 +59,15 @@ executable adiff
   other-extensions: CPP
 
   -- Other library packages from which modules are imported.
-  build-depends: base ^>=4.13.0.0,
+  build-depends: base ^>=4.15.0.0,
+                 attoparsec ^>=0.14,
                  extra ^>= 1.7,
                  mmap ^>=0.5,
-                 regex-tdfa ^>= 1.3,
                  optparse-applicative ^>=0.16,
-                 bytestring ^>= 0.10.12,
+                 bytestring ^>= 0.11.2,
                  vector ^>=0.12,
+                 word8 ^>=0.1,
+                 unicode-data ^>=0.3,
                  utf8-string ^>=1.0
 
   -- Directories containing source files.

src/Diff.hs

@@ -4,18 +4,8 @@ module Diff
   ( diffToks
   ) where
 
-import Control.Monad
 import qualified Data.ByteString as B
-import qualified Data.ByteString.Builder as BB
-import qualified Data.ByteString.Internal as BI
-import qualified Data.ByteString.Lazy as BL
-import Data.ByteString.UTF8 (fromString)
-import Data.Function (on)
-import Data.List (groupBy, mapAccumL)
-import Data.List.Extra (split, takeEnd)
 import qualified Data.Vector as V
-import qualified Data.Vector.Unboxed.Mutable as M
-import Substr
 import Types
 
 data DiffEnv =
@@ -30,10 +20,11 @@ data DiffEnv =
     , deB :: Int
     , deVS :: V.Vector (Int, Int)
     , deVE :: V.Vector (Int, Int)
+    , deTokPrio :: Tok -> Int
     , deTrans :: Bool
     }
-  deriving (Show)
 
+toksMatch :: Int -> Int -> DiffEnv -> Bool
 toksMatch x y DiffEnv {deT1 = t1, deT2 = t2} = t1 V.! x == t2 V.! y
 
 stripEqToks :: TV -> TV -> (Diff, Diff, TV, TV)
@@ -56,10 +47,19 @@ stripEqToks t1 t2 = (pre, post, t1', t2')
     t1' = V.slice b (l1 - e - b) t1
     t2' = V.slice b (l2 - e - b) t2
 
+makePrios :: TV -> TV -> (Bool, BS) -> Int
+makePrios _ _ = get
+  where
+    get (isToken, str) =
+      if isToken
+        then B.length str
+        else 0
+
 diffToks :: TV -> TV -> Diff
 diffToks t1' t2' = pre ++ res ++ post
   where
     (pre, post, t1, t2) = stripEqToks t1' t2'
+    stats = makePrios t1' t2'
     res
       | V.null t1 = map (Add, ) (V.toList t2)
       | V.null t2 = map (Remove, ) (V.toList t1)
@@ -76,6 +76,7 @@ diffToks t1' t2' = pre ++ res ++ post
           , deB = V.length t2
           , deVS = V.fromList $ zip [0 .. V.length t2] (repeat 0)
           , deVE = V.fromList $ reverse $ zip [0 .. V.length t2] (repeat 0)
+          , deTokPrio = stats
           , deTrans = False
           }
       | otherwise =
@@ -91,9 +92,11 @@ diffToks t1' t2' = pre ++ res ++ post
           , deB = V.length t1
           , deVS = V.fromList $ zip [0 .. V.length t1] (repeat 0)
           , deVE = V.fromList $ reverse $ zip [0 .. V.length t1] (repeat 0)
+          , deTokPrio = stats
           , deTrans = True
           }
 
+minIndexFwd :: V.Vector (Int, Int) -> Int
 minIndexFwd =
   V.minIndexBy
     (\x y ->
@@ -102,6 +105,7 @@ minIndexFwd =
          else GT --basically normal V.minIndex
      )
 
+minIndexRev :: V.Vector (Int, Int) -> Int
 minIndexRev =
   V.minIndexBy
     (\x y ->
@@ -117,12 +121,7 @@ diffToks' de@DiffEnv {deS = s, deE = e, deA = a, deB = b, deTrans = trans} =
     mid = quot (s + e) 2
     vecSmid = vecS mid
     vecEmid = vecE mid
-    extraScore i =
+    prio i = negate . deTokPrio de $ deT1 de V.! i
-      if isToken
-        then -(B.length s)
-        else 0
-      where
-        (isToken, s) = deT1 de V.! i
     vecS = vec -- "forward" operation
       where
         vec i
@@ -138,7 +137,7 @@ diffToks' de@DiffEnv {deS = s, deE = e, deA = a, deB = b, deTrans = trans} =
                     (iupleft, supleft) = v V.! pred j
                     keep
                       | toksMatch (pred i) (pred j) de =
-                        min (iupleft, supleft + extraScore (pred i))
+                        min (iupleft, supleft + prio (pred i))
                       | otherwise = id
                     res = keep $ min (succ iup, sup) (succ ileft, sleft)
                  in res : go (succ j) res
@@ -157,7 +156,7 @@ diffToks' de@DiffEnv {deS = s, deE = e, deA = a, deB = b, deTrans = trans} =
                     (idownright, sdownright) = v V.! succ j
                     keep
                       | toksMatch i j de =
-                        min (idownright, sdownright + extraScore i)
+                        min (idownright, sdownright + prio i)
                       | otherwise = id
                     res = keep $ min (succ idown, sdown) (succ iright, sright)
                  in res : go (pred j) res
@@ -220,7 +219,7 @@ diffToks' de@DiffEnv {deS = s, deE = e, deA = a, deB = b, deTrans = trans} =
                 , fst (vecLS V.! i) == fst sCost - a + i
                 , sumL V.! i == totalCost
                 , if doKeep
-                    then scoreAdd (vecLS V.! i) (0, extraScore s) ==
+                    then scoreAdd (vecLS V.! i) (0, prio s) ==
                          vecRS V.! succ i
                     else scoreAdd (vecLS V.! i) (1, 0) == vecRS V.! i
                 , if doKeep
@@ -230,7 +229,7 @@ diffToks' de@DiffEnv {deS = s, deE = e, deA = a, deB = b, deTrans = trans} =
                 ]
             jumpEnd =
               if doKeep
-                then jumpPos + 1
+                then succ jumpPos
                 else jumpPos
          in map
               (\i ->

src/Diff3.hs

@@ -1,56 +1,47 @@
 {-# LANGUAGE TupleSections #-}
-module Diff3 where
+
+module Diff3
+  ( diff3Toks
+  ) where
 
 import Diff
-import Types
 import Merge
+import Types
 
-data Origin
+stable :: (Origin, a) -> Bool
-  = Stable
+stable (Stable, _) = True
-  | Mine
+stable _ = False
-  | Your
+
-  deriving (Show, Eq)
+applySplit :: (a -> Bool) -> ([a] -> [b]) -> ([a] -> [b]) -> [a] -> [b]
+applySplit cond f g xs = f (takeWhile cond xs) ++ g (dropWhile cond xs)
+
+conflict :: MergeOpts -> [(Origin, (Op, Tok))] -> [(Op, Tok)]
+conflict mo = go
+  where
+    go [] = []
+    go as@(a:_)
+      | stable a = applySplit stable (map snd) go as
+      | otherwise = applySplit (not . stable) (merge mo) go as
+
+align :: Diff -> Diff -> [(Origin, (Op, Tok))]
+align ((Keep, a):as) ((Keep, _):bs) = (Stable, (Keep, a)) : align as bs
+align ((Add, a):as) ((Add, b):bs) =
+  (Mine, (Add, a)) : (Your, (Add, b)) : align as bs
+align ((Remove, a):as) ((Remove, b):bs) =
+  (Mine, (Remove, a)) : (Your, (Remove, b)) : align as bs
+align ((Add, a):as) bs@((Keep, _):_) = (Mine, (Add, a)) : align as bs
+align as@((Keep, _):_) ((Add, b):bs) = (Your, (Add, b)) : align as bs
+align ((Remove, a):as) ((Keep, b):bs) =
+  (Mine, (Remove, a)) : (Your, (Keep, b)) : align as bs
+align ((Keep, a):as) ((Remove, b):bs) =
+  (Mine, (Keep, a)) : (Your, (Remove, b)) : align as bs
+align ((Add, a):as) bs@((Remove, _):_) = (Mine, (Add, a)) : align as bs
+align as@((Remove, _):_) ((Add, b):bs) = (Your, (Add, b)) : align as bs
+align [] [] = []
+align as@((Add, _):_) [] = map (Mine, ) as
+align [] bs@((Add, _):_) = map (Your, ) bs
+align _ _ = error "Internal failure: diffstreams seem broken, cannot align"
 
 diff3Toks :: MergeOpts -> TV -> TV -> TV -> Diff
 diff3Toks mo tMine tOrig tYour =
-  conflict $ align (diffToks tOrig tMine) (diffToks tOrig tYour)
+  conflict mo $ align (diffToks tOrig tMine) (diffToks tOrig tYour)
-  where
-    align :: Diff -> Diff -> [(Origin, (Op, Tok))]
-    align ((Keep, a):as) ((Keep, _):bs) = (Stable, (Keep, a)) : align as bs
-    align ((Add, a):as) ((Add, b):bs) =
-      (Mine, (Add, a)) : (Your, (Add, b)) : align as bs
-    align ((Remove, a):as) ((Remove, b):bs) =
-      (Mine, (Remove, a)) : (Your, (Remove, b)) : align as bs
-    align ((Add, a):as) bs@((Keep, _):_) = (Mine, (Add, a)) : align as bs
-    align as@((Keep, _):_) ((Add, b):bs) = (Your, (Add, b)) : align as bs
-    align ((Remove, a):as) ((Keep, b):bs) =
-      (Mine, (Remove, a)) : (Your, (Keep, b)) : align as bs
-    align ((Keep, a):as) ((Remove, b):bs) =
-      (Mine, (Keep, a)) : (Your, (Remove, b)) : align as bs
-    align ((Add, a):as) bs@((Remove, _):_) = (Mine, (Add, a)) : align as bs
-    align as@((Remove, _):_) ((Add, b):bs) = (Your, (Add, b)) : align as bs
-    align [] [] = []
-    align as@((Add, _):_) [] = map (Mine,) as
-    align [] bs@((Add, _):_) = map (Your,) bs
-    align _ _ = error "Internal failure: diffstreams seem broken, cannot align"
-    conflict :: [(Origin, (Op, Tok))] -> Diff
-    conflict [] = []
-    conflict as@(a:_)
-      | stable a = applySplit stable (map snd) conflict as
-      | unstable a = applySplit unstable merge conflict as
-    applySplit :: (a -> Bool) -> ([a] -> [b]) -> ([a] -> [b]) -> [a] -> [b]
-    applySplit cond f g xs = f (takeWhile cond xs) ++ g (dropWhile cond xs)
-    merge :: [(Origin, (Op,Tok))] -> Diff
-    merge cs =
-      let mys = map (\a -> map snd $ filter ((== a) . fst) cs) [Mine, Your]
-          [tokOrigsMine, tokOrigsYour] =
-            map (map snd.filter ((/= Add) . fst)) mys
-          [tokMine, tokYour] = map (map snd.filter ((/= Remove) . fst)) mys
-       in if tokOrigsMine /= tokOrigsYour
-            then error "Internal failure: merge origins differ"
-            else map (MineChanged,) tokMine ++
-                 map (Original,) tokOrigsMine ++
-                 map (YourChanged,) tokYour
-    stable (Stable, _) = True
-    stable _ = False
-    unstable = not . stable

src/Format.hs

@@ -1,43 +1,128 @@
-{-# LANGUAGE OverloadedStrings #-}
+module Format
-
+  ( pprHunks
-module Format where
+  , pprHunk
+  , parsePatch
+  ) where
 
 import Types
 
+import Control.Applicative
+import Control.Monad
+import Data.Attoparsec.ByteString as A
 import qualified Data.ByteString as B
 import qualified Data.ByteString.Builder as BB
 import qualified Data.ByteString.Internal as BI
 import qualified Data.ByteString.Lazy as BL
 import Data.String
-import Substr
+import Data.Word8 as W8
 
-pprHunkHdr i j = fromString $ "@@ -" ++ show i ++ " +" ++ show j ++ " @@" 
+backslash :: Word8
+backslash = BI.c2w '\\'
 
-lineSep = fromString "\n"
+newline :: Word8
+newline = BI.c2w '\n'
 
 pprHunks :: [Hunk] -> BB.Builder
 pprHunks = mconcat . map pprHunk
 
+lineSep :: BB.Builder
+lineSep = BB.word8 newline
+
+pprHunkHdr :: Int -> Int -> BB.Builder
+pprHunkHdr i j =
+  (fromString $ "@@ -" ++ show i ++ " +" ++ show j ++ " @@") <> lineSep
+
 pprHunk :: Hunk -> BB.Builder
-pprHunk ((i, j), toks) = mconcat ((pprHunkHdr i j <> lineSep) : map pprDiff1 toks)
+pprHunk ((i, j), toks) = mconcat (pprHunkHdr i j : map pprDiff1 toks)
 
 pprDiff1 :: (Op, Tok) -> BB.Builder
-pprDiff1 (op, (tok, s)) =
+pprDiff1 (op, (tok, s)) = fromString pfx <> escNewlines s <> lineSep
-  fromString pfx <> escNewlines s <> lineSep
   where
-    pfx = [opc,tc]
+    pfx = [opc, tc]
-    opc = case op of
+    opc =
+      case op of
         Add -> '+'
         Keep -> ' '
         Remove -> '-'
         MineChanged -> '<'
         Original -> '='
         YourChanged -> '>'
-    tc = if tok then '|' else '.'
+    tc =
+      if tok
+        then '|'
+        else '.'
 
 escNewlines :: BS -> BB.Builder
 escNewlines s
   | B.null s = mempty
-  | B.head s == BI.c2w '\n' = BB.stringUtf8 "\\n" <> escNewlines (B.tail s)
+  | B.head s == newline =
-  | B.head s == BI.c2w '\\' = BB.stringUtf8 "\\\\" <> escNewlines (B.tail s)
+    BB.word8 backslash <> BB.word8 (BI.c2w 'n') <> escNewlines (B.tail s)
+  | B.head s == backslash =
+    BB.word8 backslash <> BB.word8 backslash <> escNewlines (B.tail s)
   | otherwise = BB.word8 (B.head s) <> escNewlines (B.tail s)
+
+--parsePatch :: BS -> Either String [Hunk]
+parsePatch = parseOnly parseHunks
+
+parseHunks :: Parser [Hunk]
+parseHunks = many parseHunk <* endOfInput
+
+parseHunk :: Parser Hunk
+parseHunk = liftA2 (,) parseHunkHdr (many parseDiff1)
+
+parseInt :: Parser Int
+parseInt = read . map BI.w2c <$> some (satisfy W8.isDigit)
+
+eol :: Parser ()
+eol = void $ word8 newline
+
+parseHunkHdr :: Parser (Int, Int)
+parseHunkHdr = do
+  void . string $ fromString "@@ -"
+  i <- parseInt
+  void . string $ fromString " +"
+  j <- parseInt
+  void . string $ fromString " @@"
+  eol
+  return (i, j)
+
+pairs2parsers :: [(a, Char)] -> [Parser a]
+pairs2parsers = map (\(x, ch) -> x <$ word8 (BI.c2w ch))
+
+parseOpList :: [Parser Op]
+parseOpList =
+  pairs2parsers
+    [ (Add, '+')
+    , (Keep, ' ')
+    , (Remove, '-')
+    , (MineChanged, '<')
+    , (Original, '=')
+    , (YourChanged, '>')
+    ]
+
+parseOp :: Parser Op
+parseOp = choice parseOpList
+
+parseTokMarkList :: [Parser Bool]
+parseTokMarkList = pairs2parsers [(True, '|'), (False, '.')]
+
+parseTokMark :: Parser Bool
+parseTokMark = choice parseTokMarkList
+
+parseTokBS :: Parser BS
+parseTokBS =
+  (BL.toStrict . BB.toLazyByteString . mconcat <$> many parseTokChar) <* eol
+
+parseTokChar :: Parser BB.Builder
+parseTokChar =
+  choice
+    [ BB.word8 newline <$ string (fromString "\\n")
+    , BB.word8 backslash <$ string (fromString "\\\\")
+    , BB.word8 <$> satisfy (\w -> w /= backslash && w /= newline)
+    ]
+
+parseTok :: Parser Tok
+parseTok = liftA2 (,) parseTokMark parseTokBS
+
+parseDiff1 :: Parser (Op, Tok)
+parseDiff1 = liftA2 (,) parseOp parseTok

src/Hunks.hs

@@ -21,7 +21,7 @@ hunks ctxt d =
       zipWith (<=) (padEnd ctxt edits) (replicate (1 + 2 * ctxt) 0 ++ edits)
     addNums = snd . mapAccumL countTok (0, 0)
     stripNums = (,) <$> fst . head <*> map snd
-    countTok x@(i, j) d@(op, _) =
+    countTok x@(i, j) d'@(op, _) =
       (,)
         (case op of
            Remove -> (i + 1, j)
@@ -30,4 +30,4 @@ hunks ctxt d =
            MineChanged -> (i, j)
            Original -> (i + 1, j + 1)
            YourChanged -> (i, j))
-        (x, d)
+        (x, d')

src/Main.hs

@@ -1,9 +1,10 @@
 module Main where
 
+import Control.Exception
+import Control.Monad
 import qualified Data.ByteString as B
 import qualified Data.ByteString.Builder as BB
-import qualified Data.ByteString.Char8 as B8
+import Data.Foldable (traverse_)
-import Data.ByteString.UTF8 (fromString)
 import qualified Data.Vector as V
 import Diff
 import Diff3
@@ -11,14 +12,17 @@ import Format
 import Hunks
 import Merge
 import Options.Applicative
-import Redfa
+import Patch
-import System.IO (stdout)
+import System.Exit
+import System.IO (hPutStrLn, stderr, stdout)
 import System.IO.MMap
+import Tokenizers
+import Types
 import Version
 
 data ADiffOptions =
   ADiffOptions
-    { adiffRedfaOpt :: RedfaOption
+    { adiffTokOpts :: TokOpts
     , adiffCmdOpts :: ADiffCommandOpts
     }
   deriving (Show)
@@ -31,11 +35,17 @@ data ADiffCommandOpts
       }
   | CmdPatch
       { patchDryRun :: Bool
-      , patchInDir :: Maybe String
+      --, patchInDir :: Maybe String
-      , patchInput :: String
+      --, patchPathStrip :: Int
+      , patchInputPatch :: String
+      , patchOutput :: String
+      , patchReject :: String --todo convert to Maybes with optional
+      , patchBackup :: String
       , patchReverse :: Bool
-      , patchPathStrip :: Int
+      , patchScanRange :: Int
+      , context :: Int
       , patchMergeOpts :: MergeOpts
+      , patchInput :: String
       }
   | CmdDiff3
       { context :: Int
@@ -46,50 +56,90 @@ data ADiffCommandOpts
       }
   deriving (Show)
 
-contextOpt =
+contextOpt :: Bool -> Parser Int
+contextOpt forPatch =
+  check <$>
   option
     auto
     (metavar "CONTEXT" <>
      short 'C' <>
      long "context" <>
-     value 5 <> help "How many tokens around the change to include in the patch")
+     value
+       (if forPatch
+          then 4
+          else 8) <>
+     help
+       (if forPatch
+          then "Maximum number of context tokens that may be discarded from the beginning and end of the hunk when attempting to find a match"
+          else "How many tokens around the change to include in the patch"))
+  where
+    check c
+      | c < 0 = error "Negative context"
+      | otherwise = c
 
+diffCmdOptions :: Parser ADiffCommandOpts
 diffCmdOptions =
-  CmdDiff <$> contextOpt <*> strArgument (metavar "FROMFILE") <*>
+  CmdDiff <$> contextOpt False <*> strArgument (metavar "FROMFILE") <*>
   strArgument (metavar "TOFILE")
 
+patchCmdOptions :: Parser ADiffCommandOpts
 patchCmdOptions =
   CmdPatch <$>
   switch
     (short 'n' <>
      long "dry-run" <>
      help "Do not patch anything, just print what would happen") <*>
-  optional
+  -- optional (strOption $ short 'd' <> long "directory" <> metavar "DIR" <> help "Work in DIR") <*>
-    (strOption $
+  -- option auto (short 'p' <> long "strip" <> metavar "NUM" <> help "Strip NUM leading components from the paths" <> value 0) <*>
-     short 'd' <> long "directory" <> metavar "DIR" <> help "Work in DIR") <*>
   strOption
     (short 'i' <>
      long "input" <>
-     metavar "INPUT" <>
+     metavar "PATCHFILE" <>
-     help "Read the patchfile from INPUT, defaults to `-' for STDIN" <>
+     help "Read the patchfile from PATCHFILE, defaults to `-' for STDIN" <>
      value "-") <*>
+  strOption
+    (short 'o' <>
+     long "output" <>
+     metavar "OUTPUT" <>
+     help
+       "Write the patched file to OUTPUT, use `-' for STDOUT. By default, INPUT is rewritten." <>
+     value "") <*>
+  strOption
+    (short 'r' <>
+     long "reject" <>
+     metavar "REJECTS" <>
+     help
+       "Write the rejected hunks file to file REJECTS, instead of default `OUTPUT.rej'. Use `-' to discard rejects." <>
+     value "") <*>
+  strOption
+    (short 'b' <>
+     long "backup" <>
+     metavar "BACKUP" <>
+     help
+       "When rewriting INPUT after a partially applied or otherwise suspicious patch, back up the original file in BACKUP instead of default `INPUT.orig'. Use `-' to discard backups." <>
+     value "") <*>
   switch (short 'R' <> long "reverse" <> help "Unapply applied patches") <*>
   option
     auto
-    (short 'p' <>
+    (short 'S' <>
-     long "strip" <>
+     long "scan-range" <>
-     metavar "NUM" <>
+     metavar "RANGE" <>
-     help "Strip NUM leading components from the paths" <> value 0) <*>
+     help
-  mergeOption True
+       "Maximum distance from the indended patch position (in tokens) for fuzzy matching of hunks" <>
+     value 42) <*>
+  contextOpt True <*>
+  mergeOption True <*>
+  strArgument (metavar "INPUT")
 
+diff3CmdOptions :: Parser ADiffCommandOpts
 diff3CmdOptions =
-  CmdDiff3 <$> contextOpt <*> strArgument (metavar "MYFILE") <*>
+  CmdDiff3 <$> contextOpt False <*> strArgument (metavar "MYFILE") <*>
   strArgument (metavar "OLDFILE") <*>
   strArgument (metavar "YOURFILE") <*>
   mergeOption False
 
-actionOption :: Parser ADiffCommandOpts
+actionOptions :: Parser ADiffCommandOpts
-actionOption =
+actionOptions =
   hsubparser $
   mconcat
     [ command "diff" $ info diffCmdOptions $ progDesc "Compare two files"
@@ -98,35 +148,138 @@ actionOption =
       info diff3CmdOptions $ progDesc "Compare and merge three files"
     ]
 
-adiffOptions = ADiffOptions <$> redfaOption <*> actionOption
+adiffOptions :: Parser ADiffOptions
+adiffOptions = ADiffOptions <$> tokOptions <*> actionOptions
 
-loadToks redfa f =
+-- TODO: load in case it's not a regular file
-  V.fromList <$> (mmapFileByteString f Nothing >>= redfaTokenize redfa)
+loadToksMM :: TokOpts -> FilePath -> IO TV
+loadToksMM topt fn = loadToksWith topt fn (mmapFileByteString fn Nothing)
 
-main :: IO ()
+loadToksR :: TokOpts -> FilePath -> IO TV
-main =
+loadToksR topt fn = loadToksWith topt fn (B.readFile fn)
+
+loadToksWith :: TokOpts -> FilePath -> IO BS -> IO TV
+loadToksWith topt fn bs = V.fromList <$> (bs >>= tokenize topt fn)
+
+doDiff :: TokOpts -> ADiffCommandOpts -> IO ()
+doDiff topt (CmdDiff {diffFile1 = f1, diffFile2 = f2, context = ctxt}) = do
+  [toks1, toks2] <- traverse (loadToksMM topt) [f1, f2]
+  let output = hunks (max 0 ctxt) $ diffToks toks1 toks2
+  BB.hPutBuilder stdout $ pprHunks output
+  unless (null output) $ exitWith (ExitFailure 1)
+doDiff _ _ = error "dispatch failure"
+
+doDiff3 :: TokOpts -> ADiffCommandOpts -> IO ()
+doDiff3 topt (CmdDiff3 ctxt f1 f2 f3 mo) = do
+  [toksMine, toksOld, toksYour] <- traverse (loadToksMM topt) [f1, f2, f3]
+  let d3 = diff3Toks mo toksMine toksOld toksYour
+      isConflict (MineChanged, _) = True
+      isConflict (YourChanged, _) = True
+      isConflict _ = False
+      hasConflict = any isConflict d3
+  BB.hPutBuilder stdout $
+    if mergeDoMerge mo
+      then do
+        fmtMerged mo d3
+      else pprHunks $ hunks (max 0 ctxt) d3
+  when hasConflict $ exitWith (ExitFailure 1)
+doDiff3 _ _ = error "dispatch failure"
+
+note :: String -> IO ()
+note = hPutStrLn stderr
+
+doPatch :: TokOpts -> ADiffCommandOpts -> IO ()
+doPatch topt o = do
+  toksIn <- loadToksR topt (patchInput o)
+  hs' <-
+    parsePatch <$>
+    case (patchInputPatch o) of
+      "-" -> B.getContents
+      fn -> B.readFile fn
+  hs <-
+    case hs' of
+      Left _ -> ioError $ userError "PATCHFILE parsing failed"
+      Right x -> pure x
+  let (toks, rej, warns) =
+        patchToks
+          toksIn
+          hs
+          (patchReverse o)
+          (patchScanRange o)
+          (context o)
+          (patchMergeOpts o)
+      sus = not (null warns)
+      dry = patchDryRun o
+      rewritingInput = null (patchOutput o)
+      outputStdout = patchOutput o == "-"
+      rejFile
+        | rewritingInput || outputStdout = patchInput o ++ ".rej"
+        | otherwise = patchOutput o ++ ".rej"
+      backupFile
+        | patchBackup o == "-" = ""
+        | patchBackup o == "" && rewritingInput = patchInput o ++ ".orig"
+        | otherwise = patchBackup o
+      outFile
+        | rewritingInput = patchInput o
+        | otherwise = patchOutput o
+  traverse_ (note . pprPatchWarn) warns
+  when dry $
+    note $
+    (if not sus
+       then "OK"
+       else "Possibly problematic") ++
+    " patch with " ++ show (length rej :: Int) ++ " rejected hunks"
+  when (not (null rej)) $
+    if dry
+      then note $ "Would write rejected hunks to " ++ rejFile
+      else do
+        note $ "Writing rejected hunks to " ++ rejFile
+        BB.writeFile rejFile (pprHunks rej)
+  when (sus && not (null backupFile)) $
+    if dry
+      then note $ "Would write backup to " ++ backupFile
+      else do
+        note $ "Writing backup to " ++ backupFile
+        B.readFile (patchInput o) >>= B.writeFile backupFile
+  let doWrite output =
+        if outputStdout
+          then if dry
+                 then note "Would write output to stdout"
+                 else BB.hPutBuilder stdout output
+          else if dry
+                 then note $ "Would write output to " ++ outFile
+                 else do
+                   note $ "Writing output to " ++ outFile
+                   BB.writeFile outFile output
+  doWrite (mconcat . map (BB.byteString . snd) . V.toList $ toks)
+  when (dry && not (null rej)) $ do
+    note "Rejected hunks:"
+    BB.hPutBuilder stdout (pprHunks rej)
+  when sus $ exitWith (ExitFailure 1)
+
+main' :: IO ()
+main' =
   let opts :: ParserInfo ADiffOptions
       opts =
         info
           (adiffOptions <**> versionOption "adiff" <**> helperOption)
           (fullDesc <>
            progDesc
-             "Compare, patch and merge files on arbitrarily-tokenized sequences." <>
+             "Compare, patch and merge files on arbitrarily tokenized sequences." <>
            header "adiff: arbitrary-token diff utilities")
-   in do ADiffOptions {adiffRedfaOpt = ropt, adiffCmdOpts = copt} <-
+   in do ADiffOptions {adiffTokOpts = topt, adiffCmdOpts = copt} <-
-           execParser opts
+           customExecParser (prefs $ helpShowGlobals <> subparserInline) opts
-         redfa <- redfaPrepareRules ropt
+         (case copt of
-         case copt of
+            CmdDiff {} -> doDiff
-           CmdDiff {diffFile1 = f1, diffFile2 = f2, context = ctxt} -> do
+            CmdDiff3 {} -> doDiff3
-             [toks1, toks2] <- traverse (loadToks redfa) [f1, f2]
+            CmdPatch {} -> doPatch)
-             BB.hPutBuilder stdout $
+           topt
-               pprHunks $ hunks (max 0 ctxt) $ diffToks toks1 toks2
+           copt
-           CmdPatch {} -> putStrLn "not supported yet"
+
-           CmdDiff3 ctxt f1 f2 f3 mo -> do
+main :: IO ()
-             [toksMine, toksOld, toksYour] <-
+main =
-               traverse (loadToks redfa) [f1, f2, f3]
+  main' `catch`
-             let d3 = diff3Toks mo toksMine toksOld toksYour
+  (\e -> do
-             BB.hPutBuilder stdout $
+     let err = show (e :: IOException)
-               if mergeDoMerge mo
+     note err
-                 then fmtMerged mo d3
+     exitWith $ ExitFailure 2)
-                 else pprHunks $ hunks (max 0 ctxt) d3

src/Merge.hs

@@ -1,12 +1,13 @@
 {-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE TupleSections #-}
 
 module Merge
   ( MergeOpts(..)
   , mergeOption
   , fmtMerged
+  , merge
   ) where
 
-import qualified Data.ByteString as B
 import qualified Data.ByteString.Builder as BB
 import Data.String
 import Options.Applicative
@@ -15,6 +16,7 @@ import Types
 data MergeOpts =
   MergeOpts
     { mergeDoMerge :: Bool
+    , mergeIgnoreWhitespace :: Bool
     , mergeForceWhitespace :: Bool
     , mergeKeepWhitespace :: Bool
     , mergeCStartStr :: BS
@@ -24,17 +26,17 @@ data MergeOpts =
     }
   deriving (Show)
 
-marker = fromString . replicate 7
+mergeOption :: Bool -> Parser MergeOpts
-
 mergeOption forPatch =
   addLBR <$>
   ((,) <$>
    switch
      (short 'a' <>
       long "add-linebreak" <>
-      help "Automatically add a line break after conflict markers") <*>
+      help "Automatically add a line break after conflict markers. Useful with `lines' lexer.") <*>
    mo)
   where
+    marker = fromString . replicate 7
     mo =
       MergeOpts <$>
       switch
@@ -42,27 +44,36 @@ mergeOption forPatch =
          long "merge" <>
          help
            (if forPatch
-              then "Merge using conflict markers instead of printing the rejected hunks"
+              then "Instead of printing the rejected thunks, merge using conflict markers as if the INPUT was `MYFILE' and the patch would produced `YOURFILE' from the original."
               else "Output the merged file instead of the patch")) <*>
       switch
-        (short 'w' <>
+        (short 'I' <>
-         long "whitespace" <>
+         long "ignore-whitespace" <>
+         help
+           ("Ignore " ++
+            (if forPatch
+               then "hunks"
+               else "chunks") ++
+            " that change only whitespace")) <*>
+      switch
+        (short 'F' <>
+         long "force-whitespace" <>
          help
            ((if forPatch
-               then "Force rejecting a thunk"
+               then "Force rejecting a hunk"
                else "Force a merge conflict") ++
-            " on whitespace mismatch")) <*>
+            " on whitespace mismatch (overrides `ignore-whitespace')")) <*>
       switch
-        (short 'k' <>
+        (short 'K' <>
          long "keep-whitespace" <>
          help
-           ("On whitespace mismatch, default to the version from " ++
+           ("On whitespace mismatch, output the version from " ++
             (if forPatch
-               then "original file"
+               then "the original file"
                else "MYFILE") ++
             " instead of the one from " ++
             (if forPatch
-               then "patch"
+               then "the context in patch"
                else "YOURFILE"))) <*>
       strOption
         (long "merge-start" <>
@@ -89,24 +100,27 @@ mergeOption forPatch =
         , mergeCEndStr = mergeCEndStr x <> "\n"
         }
 
-{- This kinda relies on reasonable ordering within the conflicts in the Diff -}
+{- This kinda relies on reasonable ordering
+ - within the conflicts in the Diff -}
 fmtMerged :: MergeOpts -> Diff -> BB.Builder
 fmtMerged mo = go Keep
   where
     go op []
       | conflictOp op = bb $ mergeCEndStr mo
       | otherwise = mempty
-    go last l@((op, (_, tok)):xs)
+    go prev l@((op, (_, tok)):xs)
-      | conflictOp last && not (conflictOp op) =
+      | conflictOp prev && not (conflictOp op) =
         bb (mergeCEndStr mo) <> go Keep l
-      | not (conflictOp last) && conflictOp op =
+      | not (conflictOp prev) && conflictOp op =
         bb (mergeCStartStr mo) <> go MineChanged l
-      | last /= op && conflictOp op =
+      | prev /= op && conflictOp op =
         (case op of
            MineChanged -> bb $ mergeCStartStr mo
            Original -> bb $ mergeMineSepStr mo
-           YourChanged -> bb $ mergeYourSepStr mo) <>
+           YourChanged -> bb $ mergeYourSepStr mo
+           _ -> error "Internal conflict handling failure") <>
         go op l
+      | op == Remove = go op xs
       | otherwise = bb tok <> go op xs
     conflictOp o =
       case o of
@@ -115,3 +129,27 @@ fmtMerged mo = go Keep
         Remove -> False
         _ -> True
     bb = BB.byteString
+
+merge :: MergeOpts -> [(Origin, (Op, Tok))] -> Diff
+merge mo cs = go
+  where
+    mys@[diffMine, diffYour] =
+      map (\a -> map snd $ filter ((a ==) . fst) cs) [Mine, Your]
+    [tokOrigsMine, tokOrigsYour] = map (map snd . filter ((Add /=) . fst)) mys
+    [tokMine, tokYour] = map (map snd . filter ((Remove /=) . fst)) mys
+    conflict =
+      map (MineChanged, ) tokMine ++
+      map (Original, ) tokOrigsMine ++ map (YourChanged, ) tokYour
+    noTokens = all (not . fst . snd) (diffMine ++ diffYour)
+    go
+      | tokOrigsMine /= tokOrigsYour =
+        error "Internal failure: merge origins differ"
+      | mergeIgnoreWhitespace mo && noTokens = map (Keep, ) tokOrigsMine
+      | all ((Keep ==) . fst) diffYour = diffMine -- only mine changed
+      | all ((Keep ==) . fst) diffMine = diffYour -- only your changed
+      | diffMine == diffYour = diffMine -- false conflict
+      | not (mergeForceWhitespace mo) && noTokens =
+        if mergeKeepWhitespace mo
+          then diffMine
+          else diffYour -- conflict happened, but not on significant tokens
+      | otherwise = conflict -- true conflict

src/Patch.hs

@@ -1,4 +1,210 @@
-module Patch where
+module Patch
+  ( patchToks
+  , pprPatchWarn
+  ) where
 
-patchToks :: a
+import Data.Maybe (catMaybes)
-patchToks = undefined
+import qualified Data.Vector as V
+import Merge
+import Types
+
+data PatchWarn
+  = HunkMatched Hunk (Int, Int)
+  | HunkFailed Hunk (Int, Int)
+  deriving (Show)
+
+data PatchState =
+  PatchState
+    { input :: TV
+    , output :: [TV]
+    , warns :: [PatchWarn]
+    , inOff :: Int -- where we are in `input`
+    , patchInOff :: Int -- to what position does that correspond in patch input (gets adjusted on fuzzy matches)
+    , outOff :: Int -- where we are in output tokens (informative)
+    , patchOutOff :: Int -- to what position does that correspond in patch output (gets adjusted from processed hunks)
+    }
+
+pprPatchWarn :: PatchWarn -> String
+pprPatchWarn (HunkMatched (offs, _) poffs) =
+  "hunk (" ++ pprOffs offs ++ ") succeeded at " ++ pprOffs poffs
+pprPatchWarn (HunkFailed (offs, _) poffs) =
+  "hunk (" ++ pprOffs offs ++ ") FAILED, expected at " ++ pprOffs poffs
+
+pprOffs :: (Int, Int) -> String
+pprOffs (o, n) = "-" ++ show o ++ " +" ++ show n
+
+patchToks ::
+     TV
+  -> [Hunk]
+  -> Bool
+  -> Int
+  -> Int
+  -> MergeOpts
+  -> (TV, [Hunk], [PatchWarn])
+patchToks toks hunks' revPatch scan ctxt mopt =
+  go hunks $ PatchState toks [] [] 0 0 0 0
+  where
+    hunks
+      | revPatch = map revHunk hunks'
+      | otherwise = hunks'
+    revHunk ((o, n), diff) = ((n, o), map revDiff diff)
+    revDiff (Add, t) = (Remove, t)
+    revDiff (Remove, t) = (Add, t)
+    revDiff (Keep, t) = (Keep, t)
+    revDiff _ = error "cannot reverse conflict diff"
+    go [] ps =
+      ( V.concat (output ps ++ [V.drop (inOff ps) (input ps)])
+      , [rej | HunkFailed rej _ <- warns ps]
+      , warns ps)
+    go (h:hs) ps = go hs ps'
+      where
+        ((fromPos, toPos), diff) = h
+        advance = fromPos - patchInOff ps
+        noMatch =
+          ps
+            { warns =
+                warns ps ++
+                [HunkFailed h (advance + inOff ps, advance + outOff ps)]
+            , patchOutOff = patchOutOff ps - diffOffChange diff
+            }
+        cleanMatch :: Maybe PatchState
+        cleanMatch = patchHunkClean ps h mopt
+        isContext :: (Op, Tok) -> Bool
+        isContext (op, _) = op == Keep
+        discardedContextDiffs :: [(Int, Diff)]
+        discardedContextDiffs =
+          let (fwdCtxt, d') = span isContext diff
+              (revCtxt, revMid) = span isContext (reverse d')
+              mid = reverse revMid
+              discards n c@(_:r) = (n, c) : discards (n + 1) r
+              discards n [] = (n, []) : discards n []
+           in zipWith
+                (\(dfwd, fwd) (drev, rev) ->
+                   (max dfwd drev, fwd ++ mid ++ reverse rev))
+                (discards 0 fwdCtxt)
+                (discards 0 revCtxt)
+        fuzzyHunks :: [Hunk]
+        fuzzyHunks = do
+          (discarded, ddiff) <- take (ctxt + 1) discardedContextDiffs
+          off <- 0 : concatMap (\x -> [-x, x]) [1 .. scan]
+          pure ((fromPos + discarded + off, toPos + discarded), ddiff)
+        fuzzyMatches =
+          [ (\x -> x {warns = warns x ++ [HunkMatched h fPos]}) <$>
+          patchHunkClean ps fh mopt
+          | fh@(fPos, _) <- tail fuzzyHunks -- tail omits the "clean" 0,0 one
+          ]
+        ps' = head $ catMaybes (cleanMatch : fuzzyMatches) ++ [noMatch]
+
+patchHunkClean :: PatchState -> Hunk -> MergeOpts -> Maybe PatchState
+patchHunkClean ps ((fromPos, toPos), diff) mopts
+  | expInOff < 0 || expOutOff < 0 = Nothing
+  | mergeIgnoreWhitespace mopts && whitespaceOnly diff = Just ps
+  | Just repl <- matchDiff mopts (V.toList origPart) diff =
+    Just
+      ps
+        { output = output ps ++ [skipped, V.fromList repl]
+        , inOff = expInOff + matchLen
+        , patchInOff = fromPos + matchLen
+        , outOff = expOutOff + replLen
+        , patchOutOff = toPos + replLen
+        }
+  | otherwise = Nothing
+  where
+    matchLen = diffMatchLen diff
+    replLen = diffReplLen diff
+    advance = fromPos - patchInOff ps
+    expInOff = advance + inOff ps
+    expOutOff = advance + outOff ps
+    skipped = V.take expInOff $ input ps
+    origPart = V.take matchLen $ V.drop expInOff $ input ps
+
+whitespaceOnly :: Diff -> Bool
+whitespaceOnly = all wsOnly
+  where
+    wsOnly (Keep, _) = True
+    wsOnly (Original, _) = True
+    wsOnly (_, (False, _)) = True
+    wsOnly _ = False
+
+diffMatchLen :: Diff -> Int
+diffMatchLen = sum . map (off . fst)
+  where
+    off Keep = 1
+    off Remove = 1
+    off Original = 1
+    off _ = 0
+
+diffReplLen :: Diff -> Int
+diffReplLen = sum . map (off . fst)
+  where
+    off Keep = 1
+    off Add = 1
+    off Original = 1
+    off _ = 0 -- tricky: the conflicts do not actually add to the diff counters
+
+diffOffChange :: Diff -> Int
+diffOffChange = sum . map (off . fst)
+  where
+    off Add = 1
+    off Remove = -1
+    off _ = 0
+
+markToks :: MergeOpts -> Op -> Op -> [Tok]
+markToks mopts x' y' = map (\s -> (True, s)) $ go x' y'
+  where
+    unmarked :: Op -> Bool
+    unmarked Keep = True
+    unmarked Add = True
+    unmarked Remove = True
+    unmarked _ = False
+    go :: Op -> Op -> [BS]
+    go x y
+      | x == y = []
+      | unmarked x = goUnmarked y
+      | x == MineChanged
+      , y /= Original = mergeYourSepStr mopts : go YourChanged y
+      | x == MineChanged = mergeMineSepStr mopts : go Original y
+      | x == Original = mergeYourSepStr mopts : go YourChanged y
+      | x == YourChanged = mergeCEndStr mopts : goUnmarked y
+      | otherwise = error "internal error in markToks"
+    goUnmarked :: Op -> [BS]
+    goUnmarked y
+      | unmarked y = []
+      | otherwise = mergeCStartStr mopts : go MineChanged y
+
+matchDiff :: MergeOpts -> [Tok] -> Diff -> Maybe [Tok]
+matchDiff mopt = go Keep
+  where
+    withMark :: Op -> Op -> [Tok] -> ([Tok] -> [Tok])
+    withMark prev op toks = (++) (markToks mopt prev op ++ toks)
+    go :: Op -> [Tok] -> Diff -> Maybe [Tok]
+    go prev ts ds
+      | null ts
+      , null ds = return $ markToks mopt prev Keep
+      | ((op, tok):ds') <- ds
+      , op == Add || op == MineChanged || op == YourChanged =
+        withMark prev op [tok] <$> go op ts ds'
+      | (intok:ts') <- ts
+      , ((op, tok):ds') <- ds
+      , op == Keep || op == Original
+      , tokCmp' mopt intok tok =
+        withMark
+          prev
+          op
+          [ if mergeKeepWhitespace mopt && not (fst intok)
+              then intok
+              else tok
+          ] <$>
+        go op ts' ds'
+      | (intok:ts') <- ts
+      , ((op, tok):ds') <- ds
+      , op == Remove
+      , tokCmp' mopt intok tok = withMark prev op [] <$> go op ts' ds'
+      | otherwise = Nothing
+
+tokCmp' :: MergeOpts -> Tok -> Tok -> Bool
+tokCmp' MergeOpts {mergeForceWhitespace = x} = tokCmp x
+
+tokCmp :: Bool -> Tok -> Tok -> Bool
+tokCmp False (False, _) (False, _) = True -- do not force rejecting on whitespace change
+tokCmp _ a b = a == b -- otherwise just compare

src/Redfa.hs

@@ -1,185 +0,0 @@
-{-# LANGUAGE FlexibleContexts #-}
-
-module Redfa where
-
-import Control.Monad
-import qualified Data.ByteString as B
-import qualified Data.ByteString.Builder as BB
-import qualified Data.ByteString.Char8 as B8
-import qualified Data.ByteString.Internal as BI
-import qualified Data.ByteString.Lazy as BL
-import Data.ByteString.UTF8 (fromString, toString)
-import Data.List
-import Data.Maybe
-import Data.Monoid
-import qualified Data.Vector as V
-import Options.Applicative
-import Text.Regex.TDFA
-import Types
-import Substr
-
-data RedfaOption
-  = RedfaOptionRules [BS]
-  | RedfaOptionFile String
-  deriving (Show)
-
-data RedfaSpec =
-  RedfaSpec
-    { redfaStart :: Int
-    , redfaNames :: V.Vector BS
-    , redfaRules :: V.Vector [RedfaRule] --start, mid, jump, token
-    }
-
-data RedfaRule =
-  RedfaRule
-    { rrRegexStart :: Regex
-    , rrRegexMid :: Regex
-    , rrJump :: Int
-    , rrIsToken :: Bool
-    }
-
-redfaOption :: Parser RedfaOption
-redfaOption =
-  RedfaOptionRules <$>
-  some
-    (strOption $
-     metavar "RULE" <>
-     short 'L' <>
-     long "lex" <> help "Lexing rule (specify repeatedly for more rules)") <|>
-  RedfaOptionFile <$>
-  strOption
-    (metavar "FILE" <>
-     short 'l' <>
-     long "lexer" <> help "File from where to load the lexing rules")
-
-redfaOptionToRuleStrings :: RedfaOption -> IO [BS]
-redfaOptionToRuleStrings (RedfaOptionRules x) = return x
-redfaOptionToRuleStrings (RedfaOptionFile fn) =
-  B8.lines <$> B.readFile fn -- TODO improve
-
-splitFirst :: Char -> BS -> (BS, BS)
-splitFirst c s = B.splitAt (fromMaybe (B.length s) $ B8.elemIndex c s) s
-
-redfaRuleStringToRuleStr :: BS -> Maybe (BS, BS, BS, Bool)
-redfaRuleStringToRuleStr s =
-  let (spec, regex) = splitFirst ':' s
-      (from, to) = splitFirst '>' spec
-      sf = B8.strip from
-      (cleanFrom, isToken)
-        | B.null sf = (sf, True)
-        | B.head sf == fromIntegral (fromEnum '_') = (B.tail sf, False)
-        | otherwise = (sf, True)
-      go
-        | B.null s = Nothing
-        | B.head s == 35 = Nothing -- # comment
-        | B.null regex = Just (B.empty, B.empty, spec, isToken)
-        | B.null to = Just (cleanFrom, cleanFrom, B.tail regex, isToken)
-        | otherwise =
-          Just (cleanFrom, B8.strip $ B.tail to, B.tail regex, isToken)
-   in go
-
-unescapeRegex :: MonadFail m => BS -> m BS
-unescapeRegex s = BL.toStrict . BB.toLazyByteString <$> unescape' s
-  where
-    unescape' :: MonadFail m => BS -> m BB.Builder
-    unescape' s
-      | B.null s = return mempty
-      | B.head s == BI.c2w '\\' && B.null (B.tail s) =
-        fail "incomplete escape sequence"
-      | B.head s == BI.c2w '\\' =
-        let rest = B.tail s
-            cc = B.head rest
-            thechar =
-              BB.stringUtf8 $
-              case BI.w2c cc of
-                '`' -> "\\`"
-                '\'' -> "\\'"
-                'b' -> "\\b"
-                'B' -> "\\B"
-                '<' -> "\\<"
-                '>' -> "\\>"
-                'a' -> "\a"
-                'e' -> "\x1b"
-                'f' -> "\f"
-                'n' -> "\n"
-                'r' -> "\r"
-                't' -> "\t"
-                '\\' -> "\\\\"
-                a -> [a] --TODO add support for \x and \u
-         in (thechar <>) <$> unescape' (B.tail rest)
-      | otherwise = mappend (BB.word8 $ B.head s) <$> unescape' (B.tail s)
-
-redfaPrepareRules :: RedfaOption -> IO RedfaSpec
-redfaPrepareRules opt = do
-  (states, jumps, regexes, isToken) <-
-    unzip4 . mapMaybe redfaRuleStringToRuleStr <$>
-    redfaOptionToRuleStrings opt
-  uRegexes <- traverse unescapeRegex regexes
-  startREs <- traverse (makeRegexM . (fromString "\\`" <>)) uRegexes
-  midREs <- traverse (makeRegexM . (fromString "\\`(.|\n)" <>)) uRegexes
-  let ids = nub . sort $ states ++ jumps
-      index' = fromJust . flip elemIndex ids
-      statesIds = map index' states
-      jumpsIds = map index' jumps
-      start = index' $ head states
-      rules :: [[RedfaRule]]
-      rules = do
-        stateId <- [0 .. length ids - 1]
-        return $ do
-          (a, b, (rs, rm), t) <-
-            zip4 statesIds jumpsIds (zip startREs midREs) isToken
-          guard $ a == stateId
-          return $ RedfaRule rs rm b t
-  return $ RedfaSpec start (V.fromList ids) (V.fromList rules)
-
-redfaTokenize :: MonadFail m => RedfaSpec -> BS -> m [Tok]
-redfaTokenize spec s = redfaTokenize' spec s (redfaStart spec) 0 []
-
-redfaTokenize' ::
-     MonadFail m
-  => RedfaSpec
-  -> BS
-  -> Int
-  -> Int
-  -> [Int]
-  -> m [Tok]
-redfaTokenize' spec s state off visited
-  | off >= B.length s = pure []
-  | otherwise =
-    let (ooff, reg) =
-          if off == 0
-            then (0, rrRegexStart)
-            else (1, rrRegexMid)
-        matchString = B.drop (off - ooff) s
-        matches :: [(RedfaRule, (MatchOffset, MatchLength))]
-        matches =
-          filter contOK $
-          zip rules $ map (\x -> match (reg x) matchString) rules
-        contOK (RedfaRule {rrJump = j}, (off', len))
-          | off' /= 0 = False
-          | len > ooff = True
-          | otherwise = j `notElem` visited
-     in case matches of
-          [] ->
-            fail $
-            "Tokenization could not continue from " <>
-            stateStrPretty <> " at offset " <> show (off + ooff)
-          ((rule, (_, len)):_) ->
-            let matchLen = len - ooff
-             in (if matchLen > 0
-                   then (:) (rrIsToken rule, substr off matchLen s)
-                   else id) <$>
-                redfaTokenize'
-                  spec
-                  s
-                  (rrJump rule)
-                  (off + matchLen)
-                  (if matchLen > 0
-                     then []
-                     else state : visited)
-  where
-    rules = redfaRules spec V.! state
-    stateStr = redfaNames spec V.! state
-    stateStrPretty
-      | B.null stateStr = "anonymous state"
-      | otherwise = "state `" <> toString stateStr <> "'"

src/Substr.hs

@@ -1,7 +0,0 @@
-module Substr where 
-
-import Types
-import qualified Data.ByteString as B
-
-substr :: Int -> Int -> BS -> BS
-substr b l = B.take l . B.drop b

src/Tokenizers.hs

@@ -0,0 +1,114 @@
+{-# LANGUAGE Rank2Types #-}
+{-# LANGUAGE TupleSections #-}
+
+module Tokenizers (TokOpts, tokOptions, tokenize) where
+
+import qualified Data.ByteString as B
+import qualified Data.ByteString.UTF8 as BU8
+import qualified Data.Word8 as W8
+import Options.Applicative
+import Types
+import qualified Unicode.Char.General as UC
+
+data TokOpts =
+  TokOpts
+    { optTokenizerName :: String
+    }
+  deriving (Show)
+
+tokOptions :: Parser TokOpts
+tokOptions =
+  TokOpts <$>
+  strOption
+    (metavar "LEXER" <>
+     short 'l' <>
+     long "lexer" <>
+     value "text" <>
+     help "Lexer name (use --help-lexers to list available ones)")
+
+--TODO this should later allow choosing the lexer by suffix
+tokenize :: MonadFail m => TokOpts -> FilePath -> BS -> m [Tok]
+tokenize topt _ =
+  case filter ((== optTokenizerName topt) . tkName) tokenizers of
+    [t] -> runTokenizer t
+    _ -> const $ fail "Lexer not found"
+
+data Tokenizer =
+  Tokenizer
+    { tkName :: String
+    , tkDescription :: String
+    , runTokenizer :: forall m. MonadFail m =>
+                                  BS -> m [Tok]
+    }
+
+tokenizers :: [Tokenizer]
+tokenizers =
+  [ Tokenizer "lines" "works like the traditional diff" tokenizeLines
+  , Tokenizer "asciiwords" "separate by ASCII whitespace" tokenizeWords8
+  , Tokenizer "words" "separate on any UTF8 spacing" tokenizeWords
+  , Tokenizer
+      "text"
+      "separate groups of similar-class UTF8 characters"
+      tokenizeUnicode
+  ]
+
+tokenizeLines :: MonadFail m => BS -> m [Tok]
+tokenizeLines = pure . map (True, ) . BU8.lines'
+
+tokenizeWords8 :: MonadFail m => BS -> m [Tok]
+tokenizeWords8 =
+  pure . map (\x -> (not $ W8.isSpace $ B.head x, x)) . bsGroupOn W8.isSpace
+  where
+    bsGroupOn f = B.groupBy (\l r -> f l == f r)
+
+tokenizeWords :: MonadFail m => BS -> m [Tok]
+tokenizeWords = pure . map makeTokenBU8 . groupOnBU8 UC.isWhiteSpace
+
+makeTokenBU8 :: BS -> Tok
+makeTokenBU8 s = if
+  maybe False (not . UC.isWhiteSpace . fst) (BU8.decode s)
+  then (True, s)
+  else (False, s)
+
+{- needed?
+foldri :: (Char -> Int -> a -> a) -> a -> BS -> a
+foldri cons nil cs =
+  case decode cs of
+    Just (a, l) -> cons a l (foldrS cons nil $ drop l cs)
+    Nothing -> nil
+-}
+
+spanBU8 :: (Char -> Bool) -> BS -> (BS, BS)
+spanBU8 f orig = B.splitAt (go 0 orig) orig
+  where
+    go :: Int -> BS -> Int
+    go n bs =
+      case BU8.decode bs of
+        Just (a, l) ->
+          if f a
+            then go (n + l) (B.drop l bs)
+            else n
+        Nothing -> n
+
+groupByBU8 :: (Char -> Char -> Bool) -> BS -> [BS]
+groupByBU8 f s =
+  case BU8.decode s of
+    Just (a, _) ->
+      let (g, s') = spanBU8 (f a) s
+       in g : groupByBU8 f s'
+    Nothing -> []
+
+groupOnBU8 :: Eq a => (Char -> a) -> BS -> [BS]
+groupOnBU8 f = groupByBU8 (\l r -> f l == f r)
+
+tokenizeUnicode :: MonadFail m => BS -> m [Tok]
+tokenizeUnicode = pure . map makeTokenBU8 . groupOnBU8 simpleUnicodeCategory
+
+simpleUnicodeCategory :: Char -> Int
+simpleUnicodeCategory c
+  | UC.isWhiteSpace c = 1
+  | UC.isAlphaNum c = 2
+  | UC.isSymbol c = 3
+  | UC.isPunctuation c = 4
+  | UC.isSeparator c = 5
+  | otherwise = 0

src/Types.hs

@@ -1,3 +1,5 @@
+--{-# LANGUAGE BangPatterns #-}
+
 module Types where
 
 import Data.ByteString
@@ -5,6 +7,7 @@ import Data.Vector
 
 type BS = ByteString
 
+{- TODO: all this needs to get unboxed -}
 type Tok = (Bool, BS)
 
 type TV = Vector Tok
@@ -21,3 +24,9 @@ data Op
   | Original
   | YourChanged
   deriving (Show, Eq)
+
+data Origin
+  = Stable
+  | Mine
+  | Your
+  deriving (Show, Eq)

src/Version.hs

@@ -7,6 +7,7 @@ import Options.Applicative
 adiffVersion :: String
 adiffVersion = VERSION_adiff
 
+versionOption :: String -> Parser (a -> a)
 versionOption prog =
   infoOption
     (prog <>