Pure-haskell "memchr"

src/Data/Text.hs

@@ -231,6 +231,7 @@ import Data.Binary (Binary(get, put))
 import Data.Monoid (Monoid(..))
 import Data.Semigroup (Semigroup(..))
 import Data.String (IsString(..))
+import Data.Text.Internal.ArrayUtils (memchr)
 import Data.Text.Internal.IsAscii (isAscii)
 import Data.Text.Internal.Reverse (reverse)
 import Data.Text.Internal.Measure (measure_off)
@@ -254,7 +255,7 @@ import qualified Data.Text.Lazy as L
 #endif
 import Data.Word (Word8)
 import Foreign.C.Types
-import GHC.Base (eqInt, neInt, gtInt, geInt, ltInt, leInt, ByteArray#)
+import GHC.Base (eqInt, neInt, gtInt, geInt, ltInt, leInt)
 import qualified GHC.Exts as Exts
 import GHC.Int (Int8)
 import GHC.Stack (HasCallStack)
@@ -1864,13 +1865,9 @@ lines (Text arr@(A.ByteArray arr#) off len) = go off
       | delta < 0 = [Text arr n (len + off - n)]
       | otherwise = Text arr n delta : go (n + delta + 1)
       where
-        delta = cSsizeToInt $
-          memchr arr# (intToCSize n) (intToCSize (len + off - n)) 0x0A
+        delta = memchr arr# n (len + off - n) 0x0A
 {-# INLINE lines #-}
 
-foreign import ccall unsafe "_hs_text_memchr" memchr
-    :: ByteArray# -> CSize -> CSize -> Word8 -> CSsize
-
 -- | /O(n)/ Joins lines, after appending a terminating newline to
 -- each.
 unlines :: [Text] -> Text

src/Data/Text/Internal/ArrayUtils.hs

@@ -0,0 +1,33 @@
+{-# LANGUAGE MagicHash #-}
+{-# LANGUAGE UnliftedFFITypes #-}
+{-# LANGUAGE CPP #-}
+
+module Data.Text.Internal.ArrayUtils (memchr) where
+
+#if defined(PURE_HASKELL)
+import qualified Data.Text.Array as A
+import Data.List (elemIndex)
+#else
+import Foreign.C.Types
+import System.Posix.Types (CSsize(..))
+#endif
+import GHC.Exts (ByteArray#)
+import Data.Word (Word8)
+
+memchr :: ByteArray# -> Int -> Int -> Word8 -> Int
+#if defined(PURE_HASKELL)
+memchr arr# off len w =
+    let tempBa = A.ByteArray arr#
+    in case elemIndex w (A.toList tempBa off len) of
+        Nothing -> -1
+        Just i -> i
+#else
+memchr arr# off len w = fromIntegral $ c_memchr arr# (intToCSize off) (intToCSize len) w
+
+intToCSize :: Int -> CSize
+intToCSize = fromIntegral
+
+
+foreign import ccall unsafe "_hs_text_memchr" c_memchr
+    :: ByteArray# -> CSize -> CSize -> Word8 -> CSsize
+#endif

src/Data/Text/Internal/Lazy/Search.hs

@@ -24,18 +24,15 @@ module Data.Text.Internal.Lazy.Search
       indices
     ) where
 
-import Data.Bits (unsafeShiftL)
+import Data.Bits (unsafeShiftL, (.|.), (.&.))
 import qualified Data.Text.Array as A
 import Data.Int (Int64)
 import Data.Word (Word8, Word64)
 import qualified Data.Text.Internal as T
 import qualified Data.Text as T (concat, isPrefixOf)
+import Data.Text.Internal.ArrayUtils (memchr)
 import Data.Text.Internal.Fusion.Types (PairS(..))
 import Data.Text.Internal.Lazy (Text(..), foldrChunks)
-import Data.Bits ((.|.), (.&.))
-import Foreign.C.Types
-import GHC.Exts (ByteArray#)
-import System.Posix.Types (CSsize(..))
 
 -- | /O(n+m)/ Find the offsets of all non-overlapping indices of
 -- @needle@ within @haystack@.
@@ -66,9 +63,9 @@ indices needle
          delta | nextInPattern = nlen + 1
                | c == z        = skip + 1
                | l >= i + nlen = case
-                  memchr xarr# (intToCSize (xoff + i + nlen)) (intToCSize (l - i - nlen)) z of
+                  memchr xarr# (xoff + i + nlen) (l - i - nlen) z of
                     -1 -> max 1 (l - i - nlen)
-                    s  -> cSsizeToInt s + 1
+                    s  -> s + 1
                 | otherwise = 1
          nextInPattern         = mask .&. swizzle (index xxs (i + nlen)) == 0
 
@@ -133,12 +130,3 @@ intToInt64 = fromIntegral
 
 word8ToInt :: Word8 -> Int
 word8ToInt = fromIntegral
-
-intToCSize :: Int -> CSize
-intToCSize = fromIntegral
-
-cSsizeToInt :: CSsize -> Int
-cSsizeToInt = fromIntegral
-
-foreign import ccall unsafe "_hs_text_memchr" memchr
-    :: ByteArray# -> CSize -> CSize -> Word8 -> CSsize

src/Data/Text/Internal/Search.hs

@@ -37,9 +37,7 @@ import qualified Data.Text.Array as A
 import Data.Word (Word64, Word8)
 import Data.Text.Internal (Text(..))
 import Data.Bits ((.|.), (.&.), unsafeShiftL)
-import Foreign.C.Types
-import GHC.Exts (ByteArray#)
-import System.Posix.Types (CSsize(..))
+import Data.Text.Internal.ArrayUtils (memchr)
 
 data T = {-# UNPACK #-} !Word64 :* {-# UNPACK #-} !Int
 
@@ -87,9 +85,9 @@ indices' (Text narr noff nlen) (Text harr@(A.ByteArray harr#) hoff hlen) = loop
       | mask .&. swizzle (A.unsafeIndex harr i) == 0
       = loop (i + nlen + 1)
       | otherwise
-      = case memchr harr# (intToCSize i) (intToCSize (hlen + hoff - i)) z of
+      = case memchr harr# i (hlen + hoff - i) z of
         -1 -> []
-        x  -> loop (i + cSsizeToInt x + 1)
+        x  -> loop (i + x + 1)
 {-# INLINE indices' #-}
 
 scanOne :: Word8 -> Text -> [Int]
@@ -103,12 +101,3 @@ scanOne c (Text harr hoff hlen) = loop 0
 
 word8ToInt :: Word8 -> Int
 word8ToInt = fromIntegral
-
-intToCSize :: Int -> CSize
-intToCSize = fromIntegral
-
-cSsizeToInt :: CSsize -> Int
-cSsizeToInt = fromIntegral
-
-foreign import ccall unsafe "_hs_text_memchr" memchr
-    :: ByteArray# -> CSize -> CSize -> Word8 -> CSsize

text.cabal

@@ -96,7 +96,6 @@ flag pure-haskell
 library
   if arch(javascript) || flag(pure-haskell)
     cpp-options: -DPURE_HASKELL
-    c-sources:  cbits/utils.c
   else
     c-sources:  cbits/is_ascii.c
                 cbits/measure_off.c
@@ -162,6 +161,7 @@ library
     Data.Text.IO
     Data.Text.IO.Utf8
     Data.Text.Internal
+    Data.Text.Internal.ArrayUtils
     Data.Text.Internal.Builder
     Data.Text.Internal.Builder.Functions
     Data.Text.Internal.Builder.Int.Digits