Make D.B.Lazy.zipWith properly lazy

As a bonus, the new code is easier to read and doesn't trigger
a spurious incomplete-pattern-match warning.  (Or finding the
bug can be seen as a bonus for cleaning up that messy code.)

Fixes #667.

Data/ByteString/Lazy.hs

@@ -1,4 +1,3 @@
-{-# OPTIONS_GHC -fno-warn-incomplete-patterns #-}
 {-# OPTIONS_HADDOCK prune #-}
 {-# LANGUAGE Trustworthy #-}
 
@@ -1095,6 +1094,7 @@ splitWith p (Chunk c0 cs0) = comb [] (S.splitWith p c0) cs0
         comb acc [s] Empty        = [revChunks (s:acc)]
         comb acc [s] (Chunk c cs) = comb (s:acc) (S.splitWith p c) cs
         comb acc (s:ss) cs        = revChunks (s:acc) : comb [] ss cs
+        comb _ [] _ = error "Strict splitWith returned [] for nonempty input"
 {-# INLINE splitWith #-}
 
 -- | /O(n)/ Break a 'ByteString' into pieces separated by the byte
@@ -1122,6 +1122,7 @@ split w (Chunk c0 cs0) = comb [] (S.split w c0) cs0
         comb acc [s] Empty        = [revChunks (s:acc)]
         comb acc [s] (Chunk c cs) = comb (s:acc) (S.split w c) cs
         comb acc (s:ss) cs        = revChunks (s:acc) : comb [] ss cs
+        comb _ [] _ = error "Strict split returned [] for nonempty input"
 {-# INLINE split #-}
 
 -- | The 'group' function takes a ByteString and returns a list of
@@ -1441,16 +1442,18 @@ zipWith _ Empty     _  = []
 zipWith _ _      Empty = []
 zipWith f (Chunk a as) (Chunk b bs) = go a as b bs
   where
-    go x xs y ys = f (S.unsafeHead x) (S.unsafeHead y)
-                 : to (S.unsafeTail x) xs (S.unsafeTail y) ys
-
-    to x Empty         _ _             | S.null x       = []
-    to _ _             y Empty         | S.null y       = []
-    to x xs            y ys            | not (S.null x)
-                                      && not (S.null y) = go x  xs y  ys
-    to x xs            _ (Chunk y' ys) | not (S.null x) = go x  xs y' ys
-    to _ (Chunk x' xs) y ys            | not (S.null y) = go x' xs y  ys
-    to _ (Chunk x' xs) _ (Chunk y' ys)                  = go x' xs y' ys
+    -- This loop is written in a slightly awkward way but ensures we
+    -- don't have to allocate any 'Chunk' objects to pass to a recursive
+    -- call.  We have in some sense performed SpecConstr manually.
+    go !x xs !y ys
+      = f (S.unsafeHead x) (S.unsafeHead y)
+      : to (S.unsafeTail x) xs (S.unsafeTail y) ys
+
+    to !x xs !y ys
+      | Chunk x' xs' <- chunk x xs
+      , Chunk y' ys' <- chunk y ys
+      = go x' xs' y' ys'
+      | otherwise = []
 
 -- | A specialised version of `zipWith` for the common case of a
 -- simultaneous map over two ByteStrings, to build a 3rd.

tests/Properties/ByteString.hs

@@ -4,6 +4,7 @@
 -- License     : BSD-style
 
 {-# LANGUAGE CPP #-}
+
 {-# LANGUAGE AllowAmbiguousTypes #-}
 {-# LANGUAGE ViewPatterns #-}
 {-# LANGUAGE ScopedTypeVariables #-}
@@ -603,6 +604,16 @@ tests =
     \f x y -> (B.zipWith f x y :: [Int]) === zipWith f (B.unpack x) (B.unpack y)
   , testProperty "packZipWith" $
     \f x y -> B.unpack (B.packZipWith ((toElem .) . f) x y) === zipWith ((toElem .) . f) (B.unpack x) (B.unpack y)
+# ifdef BYTESTRING_LAZY
+    -- Don't use (===) in these laziness tests:
+    -- We don't want printing the test case to fail!
+  , testProperty "zip is lazy" $ lazyZipTest $
+    \x y -> B.zip x y == zip (B.unpack x) (B.unpack y)
+  , testProperty "zipWith is lazy" $ \f -> lazyZipTest $
+    \x y -> (B.zipWith f x y :: [Int]) == zipWith f (B.unpack x) (B.unpack y)
+  , testProperty "packZipWith is lazy" $ \f -> lazyZipTest $
+    \x y -> B.unpack (B.packZipWith ((toElem .) . f) x y) == zipWith ((toElem .) . f) (B.unpack x) (B.unpack y)
+# endif
   , testProperty "unzip" $
     \(fmap (toElem *** toElem) -> xs) -> (B.unpack *** B.unpack) (B.unzip xs) === unzip xs
 #endif
@@ -796,3 +807,17 @@ readIntegerUnsigned xs = case readMaybe ys of
   where
     (ys, zs) = span isDigit xs
 #endif
+
+#ifdef BYTESTRING_LAZY
+lazyZipTest
+  :: Testable prop
+  => (BYTESTRING_TYPE -> BYTESTRING_TYPE -> prop)
+  ->  BYTESTRING_TYPE -> BYTESTRING_TYPE -> Property
+lazyZipTest fun = \x0 y0 -> let
+  msg = "Input chunks are: " ++ show (B.toChunks x0, B.toChunks y0)
+  (x, y) | B.length x0 <= B.length y0
+         = (x0, y0 <> error "too strict")
+         | otherwise
+         = (x0 <> error "too strict", y0)
+  in counterexample msg (fun x y)
+#endif

tests/QuickCheckUtils.hs

@@ -56,6 +56,8 @@ instance Arbitrary L.ByteString where
                                                      (sizedByteString
                                                           (n `div` numChunks))
 
+  shrink = map L.fromChunks . shrink . L.toChunks
+
 instance CoArbitrary L.ByteString where
   coarbitrary s = coarbitrary (L.unpack s)