@@ -3,6 +3,9 @@
33#if __GLASGOW_HASKELL__ >= 702
44{-# LANGUAGE Trustworthy #-}
55#endif
6+#if __GLASGOW_HASKELL__ >= 708
7+{-# LANGUAGE TypeFamilies #-}
8+#endif
69
710-- |
811-- Module : Data.Text
@@ -29,8 +32,8 @@
2932--
3033-- To use an extended and very rich family of functions for working
3134-- with Unicode text (including normalization, regular expressions,
32−-- non-standard encodings, text breaking, and locales), see the
33−-- @text-icu@ package: <http://hackage.haskell.org/package/text-icu>
35+-- non-standard encodings, text breaking, and locales), see
36+-- <http://hackage.haskell.org/package/text-icu the text-icu package >.
3437
3538module Data.Text
3639 (
@@ -126,6 +129,7 @@ module Data.Text
126129 , drop
127130 , dropEnd
128131 , takeWhile
132+ , takeWhileEnd
129133 , dropWhile
130134 , dropWhileEnd
131135 , dropAround
@@ -188,32 +192,40 @@ module Data.Text
188192
189193 -- * Low level operations
190194 , copy
195+ , unpackCString#
191196 ) where
192197
193198import Prelude (Char, Bool(..), Int, Maybe(..), String,
194199 Eq(..), Ord(..), Ordering(..), (++),
195− Read(..), Show(..),
196− (&&), (||), (+), (-), (.), ($), ($!), (>>), (*),
197− maxBound, not, return, otherwise, quot)
200+ Read(..),
201+ (&&), (||), (+), (-), (.), ($), ($!), (>>),
202+ not, return, otherwise, quot)
198203#if defined(HAVE_DEEPSEQ)
199−import Control.DeepSeq (NFData)
204+import Control.DeepSeq (NFData(rnf))
200205#endif
201206#if defined(ASSERTS)
202207import Control.Exception (assert)
203208#endif
204209import Data.Char (isSpace)
205−import Data.Data (Data(gfoldl, toConstr, gunfold, dataTypeOf))
206−import Data.Data (mkNoRepType)
210+import Data.Data (Data(gfoldl, toConstr, gunfold, dataTypeOf), constrIndex,
211+ Constr, mkConstr, DataType, mkDataType, Fixity(Prefix))
207212import Control.Monad (foldM)
213+import Control.Monad.ST (ST)
208214import qualified Data.Text.Array as A
209215import qualified Data.List as L
216+import Data.Binary (Binary(get, put))
210217import Data.Monoid (Monoid(..))
218+#if MIN_VERSION_base(4,9,0)
219+import Data.Semigroup (Semigroup(..))
220+#endif
211221import Data.String (IsString(..))
212222import qualified Data.Text.Internal.Fusion as S
213223import qualified Data.Text.Internal.Fusion.Common as S
224+import Data.Text.Encoding (decodeUtf8', encodeUtf8)
214225import Data.Text.Internal.Fusion (stream, reverseStream, unstream)
215226import Data.Text.Internal.Private (span_)
216−import Data.Text.Internal (Text(..), empty, firstf, safe, text, textP)
227+import Data.Text.Internal (Text(..), empty, firstf, mul, safe, text)
228+import Data.Text.Show (singleton, unpack, unpackCString#)
217229import qualified Prelude as P
218230import Data.Text.Unsafe (Iter(..), iter, iter_, lengthWord16, reverseIter,
219231 reverseIter_, unsafeHead, unsafeTail)
@@ -226,12 +238,13 @@ import Data.ByteString (ByteString)
226238import qualified Data.Text.Lazy as L
227239import Data.Int (Int64)
228240#endif
229−#if __GLASGOW_HASKELL__ >= 702
230−import qualified GHC.CString as GHC
231−#else
232−import qualified GHC.Base as GHC
241+import GHC.Base (eqInt, neInt, gtInt, geInt, ltInt, leInt)
242+#if __GLASGOW_HASKELL__ >= 708
243+import qualified GHC.Exts as Exts
244+#endif
245+#if MIN_VERSION_base(4,7,0)
246+import Text.Printf (PrintfArg, formatArg, formatString)
233247#endif
234−import GHC.Prim (Addr#)
235248
236249-- $strict
237250--
@@ -241,29 +254,32 @@ import GHC.Prim (Addr#)
241254-- lazy @Text@ type consists of a list of strict chunks.
242255--
243256-- The strict 'Text' type requires that an entire string fit into
244−-- memory at once. The lazy @Text@ type is capable of streaming
245−-- strings that are larger than memory using a small memory footprint.
246−-- In many cases, the overhead of chunked streaming makes the lazy
247−-- @Text@ type slower than its strict counterpart, but this is not
248−-- always the case. Sometimes, the time complexity of a function in
249−-- one module may be different from the other, due to their differing
250−-- internal structures.
257+-- memory at once. The lazy 'Data.Text.Lazy.Text' type is capable of
258+-- streaming strings that are larger than memory using a small memory
259+-- footprint. In many cases, the overhead of chunked streaming makes
260+-- the lazy 'Data.Text.Lazy.Text' type slower than its strict
261+-- counterpart, but this is not always the case. Sometimes, the time
262+-- complexity of a function in one module may be different from the
263+-- other, due to their differing internal structures.
251264--
252265-- Each module provides an almost identical API, with the main
253266-- difference being that the strict module uses 'Int' values for
254−-- lengths and counts, while the lazy module uses 'Int64' lengths.
267+-- lengths and counts, while the lazy module uses 'Data.Int.Int64'
268+-- lengths.
255269
256270-- $replacement
257271--
258272-- A 'Text' value is a sequence of Unicode scalar values, as defined
259−-- in §3.9, definition D76 of the Unicode 5.2 standard:
260−-- <http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf#page=35>. As
261−-- such, a 'Text' cannot contain values in the range U+D800 to U+DFFF
262−-- inclusive. Haskell implementations admit all Unicode code points
263−-- (§3.4, definition D10) as 'Char' values, including code points
264−-- from this invalid range. This means that there are some 'Char'
265−-- values that are not valid Unicode scalar values, and the functions
266−-- in this module must handle those cases.
273+-- in
274+-- <http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf#page=35 §3.9, definition D76 of the Unicode 5.2 standard >.
275+-- As such, a 'Text' cannot contain values in the range U+D800 to
276+-- U+DFFF inclusive. Haskell implementations admit all Unicode code
277+-- points
278+-- (<http://www.unicode.org/versions/Unicode5.2.0/ch03.pdf#page=13 §3.4, definition D10 >)
279+-- as 'Char' values, including code points from this invalid range.
280+-- This means that there are some 'Char' values that are not valid
281+-- Unicode scalar values, and the functions in this module must handle
282+-- those cases.
267283--
268284-- Within this module, many functions construct a 'Text' from one or
269285-- more 'Char' values. Those functions will substitute 'Char' values
@@ -277,8 +293,8 @@ import GHC.Prim (Addr#)
277293-- range U+D800 through U+DFFF are used by UTF-16 to denote surrogate
278294-- code points, and so cannot be represented. The functions replace
279295-- invalid scalar values, instead of dropping them, as a security
280−-- measure. For details, see Unicode Technical Report 36, §3.5:
281−-- <http://unicode.org/reports/tr36/#Deletion_of_Noncharacters>)
296+-- measure. For details, see
297+-- <http://unicode.org/reports/tr36/#Deletion_of_Noncharacters Unicode Technical Report 36, §3.5 >.)
282298
283299-- $fusion
284300--
@@ -296,11 +312,11 @@ import GHC.Prim (Addr#)
296312-- > countChars = T.length . T.toUpper . E.decodeUtf8
297313--
298314-- From the type signatures involved, this looks like it should
299−-- allocate one 'ByteString' value, and two 'Text' values. However,
300−-- when a module is compiled with optimisation enabled under GHC, the
301−-- two intermediate 'Text' values will be optimised away, and the
302−-- function will be compiled down to a single loop over the source
303−-- 'ByteString'.
315+-- allocate one 'Data.ByteString.ByteString' value, and two 'Text'
316+-- values. However, when a module is compiled with optimisation
317+-- enabled under GHC, the two intermediate 'Text' values will be
318+-- optimised away, and the function will be compiled down to a single
319+-- loop over the source 'Data.ByteString.ByteString'.
304320--
305321-- Functions that can be fused by the compiler are documented with the
306322-- phrase \"Subject to fusion\".
@@ -314,41 +330,82 @@ instance Eq Text where
314330instance Ord Text where
315331 compare = compareText
316332
317−instance Show Text where
318− showsPrec p ps r = showsPrec p (unpack ps) r
319−
320333instance Read Text where
321334 readsPrec p str = [(pack x,y) | (x,y) <- readsPrec p str]
322335
336+#if MIN_VERSION_base(4,9,0)
337+-- Semigroup orphan instances for older GHCs are provided by
338+-- 'semigroups` package
339+
340+instance Semigroup Text where
341+ (<>) = append
342+#endif
343+
323344instance Monoid Text where
324345 mempty = empty
346+#if MIN_VERSION_base(4,9,0)
347+ mappend = (<>) -- future-proof definition
348+#else
325349 mappend = append
350+#endif
326351 mconcat = concat
327352
328353instance IsString Text where
329354 fromString = pack
330355
356+#if __GLASGOW_HASKELL__ >= 708
357+instance Exts.IsList Text where
358+ type Item Text = Char
359+ fromList = pack
360+ toList = unpack
361+#endif
362+
331363#if defined(HAVE_DEEPSEQ)
332−instance NFData Text
364+instance NFData Text where rnf !_ = ()
333365#endif
334366
367+instance Binary Text where
368+ put t = put (encodeUtf8 t)
369+ get = do
370+ bs <- get
371+ case decodeUtf8' bs of
372+ P.Left exn -> P.fail (P.show exn)
373+ P.Right a -> P.return a
374+
335375-- | This instance preserves data abstraction at the cost of inefficiency.
336376-- We omit reflection services for the sake of data abstraction.
337377--
338−-- This instance was created by copying the behavior of Data.Set and
339−-- Data.Map. If you feel a mistake has been made, please feel free to
340−-- submit improvements.
378+-- This instance was created by copying the updated behavior of
379+-- @"Data.Set".@'Data.Set.Set' and @"Data.Map".@'Data.Map.Map'. If you
380+-- feel a mistake has been made, please feel free to submit
381+-- improvements.
341382--
342−-- Original discussion is archived here:
383+-- The original discussion is archived here:
384+-- <http://groups.google.com/group/haskell-cafe/browse_thread/thread/b5bbb1b28a7e525d/0639d46852575b93 could we get a Data instance for Data.Text.Text? >
343385--
344−-- "could we get a Data instance for Data.Text.Text?"
345−-- http://groups.google.com/group/haskell-cafe/browse_thread/thread/b5bbb1b28a7e525d/0639d46852575b93
386+-- The followup discussion that changed the behavior of 'Data.Set.Set'
387+-- and 'Data.Map.Map' is archived here:
388+-- <http://markmail.org/message/trovdc6zkphyi3cr#query:+page:1+mid:a46der3iacwjcf6n+state:results Proposal: Allow gunfold for Data.Map, ... >
346389
347390instance Data Text where
348391 gfoldl f z txt = z pack `f` (unpack txt)
349− toConstr _ = P.error "Data.Text.Text.toConstr"
350− gunfold _ _ = P.error "Data.Text.Text.gunfold"
351− dataTypeOf _ = mkNoRepType "Data.Text.Text"
392+ toConstr _ = packConstr
393+ gunfold k z c = case constrIndex c of
394+ 1 -> k (z pack)
395+ _ -> P.error "gunfold"
396+ dataTypeOf _ = textDataType
397+
398+#if MIN_VERSION_base(4,7,0)
399+-- | Only defined for @base-4.7.0.0@ and later
400+instance PrintfArg Text where
401+ formatArg txt = formatString $ unpack txt
402+#endif
403+
404+packConstr :: Constr
405+packConstr = mkConstr textDataType "pack" [] Prefix
406+
407+textDataType :: DataType
408+textDataType = mkDataType "Data.Text.Text" [packConstr]
352409
353410-- | /O(n)/ Compare two 'Text' values lexicographically.
354411compareText :: Text -> Text -> Ordering
@@ -373,30 +430,6 @@ pack :: String -> Text
373430pack = unstream . S.map safe . S.streamList
374431{-# INLINE [1] pack #-}
375432
376−-- | /O(n)/ Convert a 'Text' into a 'String'. Subject to fusion.
377−unpack :: Text -> String
378−unpack = S.unstreamList . stream
379−{-# INLINE [1] unpack #-}
380−
381−-- | /O(n)/ Convert a literal string into a Text. Subject to fusion.
382−unpackCString# :: Addr# -> Text
383−unpackCString# addr# = unstream (S.streamCString# addr#)
384−{-# NOINLINE unpackCString# #-}
385−
386−{-# RULES "TEXT literal" forall a.
387− unstream (S.map safe (S.streamList (GHC.unpackCString# a)))
388− = unpackCString# a #-}
389−
390−{-# RULES "TEXT literal UTF8" forall a.
391− unstream (S.map safe (S.streamList (GHC.unpackCStringUtf8# a)))
392− = unpackCString# a #-}
393−
394−-- | /O(1)/ Convert a character into a Text. Subject to fusion.
395−-- Performs replacement on invalid scalar values.
396−singleton :: Char -> Text
397−singleton = unstream . S.singleton . safe
398−{-# INLINE [1] singleton #-}
399−
400433-- -----------------------------------------------------------------------------
401434-- * Basic functions
402435
@@ -427,6 +460,7 @@ append a@(Text arr1 off1 len1) b@(Text arr2 off2 len2)
427460 | otherwise = overflowError "append"
428461 where
429462 len = len1+len2
463+ x :: ST s (A.MArray s)
430464 x = do
431465 arr <- A.new len
432466 A.copyI arr 0 arr1 off1 len1
@@ -452,8 +486,8 @@ head t = S.head (stream t)
452486uncons :: Text -> Maybe (Char, Text)
453487uncons t@(Text arr off len)
454488 | len <= 0 = Nothing
455− | otherwise = Just (c, textP arr (off+d) (len-d))
456− where Iter c d = iter t 0
489+ | otherwise = Just $ let !(Iter c d) = iter t 0
490+ in (c, text arr (off+d) (len-d))
457491{-# INLINE [1] uncons #-}
458492
459493-- | Lifted from Control.Arrow and specialized.
@@ -483,7 +517,7 @@ last (Text arr off len)
483517tail :: Text -> Text
484518tail t@(Text arr off len)
485519 | len <= 0 = emptyError "tail"
486− | otherwise = textP arr (off+d) (len-d)
520+ | otherwise = text arr (off+d) (len-d)
487521 where d = iter_ t 0
488522{-# INLINE [1] tail #-}
489523
@@ -498,8 +532,8 @@ tail t@(Text arr off len)
498532-- be non-empty. Subject to fusion.
499533init :: Text -> Text
500534init (Text arr off len) | len <= 0 = emptyError "init"
501− | n >= 0xDC00 && n <= 0xDFFF = textP arr off (len-2)
502− | otherwise = textP arr off (len-1)
535+ | n >= 0xDC00 && n <= 0xDFFF = text arr off (len-2)
536+ | otherwise = text arr off (len-1)
503537 where
504538 n = A.unsafeIndex arr (off+len-1)
505539{-# INLINE [1] init #-}
@@ -538,7 +572,9 @@ isSingleton = S.isSingleton . stream
538572-- Subject to fusion.
539573length :: Text -> Int
540574length t = S.length (stream t)
541−{-# INLINE length #-}
575+{-# INLINE [0] length #-}
576+-- length needs to be phased after the compareN/length rules otherwise
577+-- it may inline before the rules have an opportunity to fire.
542578
543579-- | /O(n)/ Compare the count of characters in a 'Text' to a number.
544580-- Subject to fusion.
@@ -557,32 +593,32 @@ compareLength t n = S.compareLengthI (stream t) n
557593
558594{-# RULES
559595"TEXT ==N/length -> compareLength/==EQ" [~1] forall t n.
560− (==) (length t) n = compareLength t n == EQ
596+ eqInt (length t) n = compareLength t n == EQ
561597 #-}
562598
563599{-# RULES
564600"TEXT /=N/length -> compareLength//=EQ" [~1] forall t n.
565− (/=) (length t) n = compareLength t n /= EQ
601+ neInt (length t) n = compareLength t n /= EQ
566602 #-}
567603
568604{-# RULES
569605"TEXT <N/length -> compareLength/==LT" [~1] forall t n.
570− (<) (length t) n = compareLength t n == LT
606+ ltInt (length t) n = compareLength t n == LT
571607 #-}
572608
573609{-# RULES
574610"TEXT <=N/length -> compareLength//=GT" [~1] forall t n.
575− (<=) (length t) n = compareLength t n /= GT
611+ leInt (length t) n = compareLength t n /= GT
576612 #-}
577613
578614{-# RULES
579615"TEXT >N/length -> compareLength/==GT" [~1] forall t n.
580− (>) (length t) n = compareLength t n == GT
616+ gtInt (length t) n = compareLength t n == GT
581617 #-}
582618
583619{-# RULES
584620"TEXT >=N/length -> compareLength//=LT" [~1] forall t n.
585− (>=) (length t) n = compareLength t n /= LT
621+ geInt (length t) n = compareLength t n /= LT
586622 #-}
587623
588624-- -----------------------------------------------------------------------------
@@ -613,22 +649,48 @@ reverse :: Text -> Text
613649reverse t = S.reverse (stream t)
614650{-# INLINE reverse #-}
615651
616−-- | /O(m+n)/ Replace every occurrence of one substring with another.
652+-- | /O(m+n)/ Replace every non-overlapping occurrence of @needle@ in
653+-- @haystack@ with @replacement@.
654+--
655+-- This function behaves as though it was defined as follows:
656+--
657+-- @
658+-- replace needle replacement haystack =
659+-- 'intercalate' replacement ('splitOn' needle haystack)
660+-- @
661+--
662+-- As this suggests, each occurrence is replaced exactly once. So if
663+-- @needle@ occurs in @replacement@, that occurrence will /not/ itself
664+-- be replaced recursively:
665+--
666+-- > replace "oo" "foo" "oo" == "foo"
667+--
668+-- In cases where several instances of @needle@ overlap, only the
669+-- first one will be replaced:
670+--
671+-- > replace "ofo" "bar" "ofofo" == "barfo"
617672--
618673-- In (unlikely) bad cases, this function's time complexity degrades
619674-- towards /O(n*m)/.
620−replace :: Text -> Text -> Text -> Text
675+replace :: Text
676+ -- ^ @needle@ to search for. If this string is empty, an
677+ -- error will occur.
678+ -> Text
679+ -- ^ @replacement@ to replace @needle@ with.
680+ -> Text
681+ -- ^ @haystack@ in which to search.
682+ -> Text
621683replace needle@(Text _ _ neeLen)
622684 (Text repArr repOff repLen)
623685 haystack@(Text hayArr hayOff hayLen)
624686 | neeLen == 0 = emptyError "replace"
625687 | L.null ixs = haystack
626688 | len > 0 = Text (A.run x) 0 len
627− | len < 0 = overflowError "replace"
628689 | otherwise = empty
629690 where
630691 ixs = indices needle haystack
631− len = hayLen - (neeLen - repLen) * L.length ixs
692+ len = hayLen - (neeLen - repLen) `mul` L.length ixs
693+ x :: ST s (A.MArray s)
632694 x = do
633695 marr <- A.new len
634696 let loop (i:is) o d = do
@@ -658,9 +720,8 @@ replace needle@(Text _ _ neeLen)
658720-- /Note/: In some languages, case conversion is a locale- and
659721-- context-dependent operation. The case conversion functions in this
660722-- module are /not/ locale sensitive. Programs that require locale
661−-- sensitivity should use appropriate versions of the case mapping
662−-- functions from the @text-icu@ package:
663−-- <http://hackage.haskell.org/package/text-icu>
723+-- sensitivity should use appropriate versions of the
724+-- <http://hackage.haskell.org/package/text-icu-0.6.3.7/docs/Data-Text-ICU.html#g:4 case mapping functions from the text-icu package >.
664725
665726-- | /O(n)/ Convert a string to folded case. Subject to fusion.
666727--
@@ -680,7 +741,7 @@ replace needle@(Text _ _ neeLen)
680741-- instead of itself.
681742toCaseFold :: Text -> Text
682743toCaseFold t = unstream (S.toCaseFold (stream t))
683−{-# INLINE [0] toCaseFold #-}
744+{-# INLINE toCaseFold #-}
684745
685746-- | /O(n)/ Convert a string to lower case, using simple case
686747-- conversion. Subject to fusion.
@@ -840,6 +901,7 @@ concat ts = case ts' of
840901 where
841902 ts' = L.filter (not . null) ts
842903 len = sumP "concat" $ L.map lengthWord16 ts'
904+ go :: ST s (A.MArray s)
843905 go = do
844906 arr <- A.new len
845907 let step i (Text a o l) =
@@ -853,13 +915,13 @@ concatMap f = concat . foldr ((:) . f) []
853915{-# INLINE concatMap #-}
854916
855917-- | /O(n)/ 'any' @p@ @t@ determines whether any character in the
856−-- 'Text' @t@ satisifes the predicate @p@. Subject to fusion.
918+-- 'Text' @t@ satisfies the predicate @p@. Subject to fusion.
857919any :: (Char -> Bool) -> Text -> Bool
858920any p t = S.any p (stream t)
859921{-# INLINE any #-}
860922
861923-- | /O(n)/ 'all' @p@ @t@ determines whether all characters in the
862−-- 'Text' @t@ satisify the predicate @p@. Subject to fusion.
924+-- 'Text' @t@ satisfy the predicate @p@. Subject to fusion.
863925all :: (Char -> Bool) -> Text -> Bool
864926all p t = S.all p (stream t)
865927{-# INLINE all #-}
@@ -950,10 +1012,10 @@ replicate n t@(Text a o l)
9501012 | n <= 0 || l <= 0 = empty
9511013 | n == 1 = t
9521014 | isSingleton t = replicateChar n (unsafeHead t)
953− | n <= maxBound `quot` l = Text (A.run x) 0 len
954− | otherwise = overflowError "replicate"
1015+ | otherwise = Text (A.run x) 0 len
9551016 where
956− len = l * n
1017+ len = l `mul` n
1018+ x :: ST s (A.MArray s)
9571019 x = do
9581020 arr <- A.new len
9591021 let loop !d !i | i >= n = return arr
@@ -1079,7 +1141,7 @@ takeWhile :: (Char -> Bool) -> Text -> Text
10791141takeWhile p t@(Text arr off len) = loop 0
10801142 where loop !i | i >= len = t
10811143 | p c = loop (i+d)
1082− | otherwise = textP arr off i
1144+ | otherwise = text arr off i
10831145 where Iter c d = iter t i
10841146{-# INLINE [1] takeWhile #-}
10851147
@@ -1090,6 +1152,27 @@ takeWhile p t@(Text arr off len) = loop 0
10901152 unstream (S.takeWhile p (stream t)) = takeWhile p t
10911153 #-}
10921154
1155+-- | /O(n)/ 'takeWhileEnd', applied to a predicate @p@ and a 'Text',
1156+-- returns the longest suffix (possibly empty) of elements that
1157+-- satisfy @p@. Subject to fusion.
1158+-- Examples:
1159+--
1160+-- > takeWhileEnd (=='o') "foo" == "oo"
1161+takeWhileEnd :: (Char -> Bool) -> Text -> Text
1162+takeWhileEnd p t@(Text arr off len) = loop (len-1) len
1163+ where loop !i !l | l <= 0 = t
1164+ | p c = loop (i+d) (l+d)
1165+ | otherwise = text arr (off+l) (len-l)
1166+ where (c,d) = reverseIter t i
1167+{-# INLINE [1] takeWhileEnd #-}
1168+
1169+{-# RULES
1170+"TEXT takeWhileEnd -> fused" [~1] forall p t.
1171+ takeWhileEnd p t = S.reverse (S.takeWhile p (S.reverseStream t))
1172+"TEXT takeWhileEnd -> unfused" [1] forall p t.
1173+ S.reverse (S.takeWhile p (S.reverseStream t)) = takeWhileEnd p t
1174+ #-}
1175+
10931176-- | /O(n)/ 'dropWhile' @p@ @t@ returns the suffix remaining after
10941177-- 'takeWhile' @p@ @t@. Subject to fusion.
10951178dropWhile :: (Char -> Bool) -> Text -> Text
@@ -1108,8 +1191,9 @@ dropWhile p t@(Text arr off len) = loop 0 0
11081191 #-}
11091192
11101193-- | /O(n)/ 'dropWhileEnd' @p@ @t@ returns the prefix remaining after
1111−-- dropping characters that fail the predicate @p@ from the end of
1194+-- dropping characters that satisfy the predicate @p@ from the end of
11121195-- @t@. Subject to fusion.
1196+--
11131197-- Examples:
11141198--
11151199-- > dropWhileEnd (=='.') "foo..." == "foo"
@@ -1129,7 +1213,7 @@ dropWhileEnd p t@(Text arr off len) = loop (len-1) len
11291213 #-}
11301214
11311215-- | /O(n)/ 'dropAround' @p@ @t@ returns the substring remaining after
1132−-- dropping characters that fail the predicate @p@ from both the
1216+-- dropping characters that satisfy the predicate @p@ from both the
11331217-- beginning and end of @t@. Subject to fusion.
11341218dropAround :: (Char -> Bool) -> Text -> Text
11351219dropAround p = dropWhile p . dropWhileEnd p
@@ -1164,8 +1248,8 @@ splitAt :: Int -> Text -> (Text, Text)
11641248splitAt n t@(Text arr off len)
11651249 | n <= 0 = (empty, t)
11661250 | n >= len = (t, empty)
1167− | otherwise = (text arr off k, text arr (off+k) (len-k))
1168− where k = iterN n t
1251+ | otherwise = let k = iterN n t
1252+ in (text arr off k, text arr (off+k) (len-k))
11691253
11701254-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns
11711255-- a pair whose first element is the longest prefix (possibly empty)
@@ -1224,9 +1308,9 @@ tails t | null t = [empty]
12241308-- copies to create substrings; they just construct new 'Text's that
12251309-- are slices of the original.
12261310
1227−-- | /O(m+n)/ Break a 'Text' into pieces separated by the first
1228−-- 'Text' argument, consuming the delimiter. An empty delimiter is
1229−-- invalid, and will cause an error to be raised.
1311+-- | /O(m+n)/ Break a 'Text' into pieces separated by the first 'Text'
1312+-- argument (which cannot be empty), consuming the delimiter. An empty
1313+-- delimiter is invalid, and will cause an error to be raised.
12301314--
12311315-- Examples:
12321316--
@@ -1239,16 +1323,23 @@ tails t | null t = [empty]
12391323-- > intercalate s . splitOn s == id
12401324-- > splitOn (singleton c) == split (==c)
12411325--
1326+-- (Note: the string @s@ to split on above cannot be empty.)
1327+--
12421328-- In (unlikely) bad cases, this function's time complexity degrades
12431329-- towards /O(n*m)/.
1244−splitOn :: Text -> Text -> [Text]
1330+splitOn :: Text
1331+ -- ^ String to split on. If this string is empty, an error
1332+ -- will occur.
1333+ -> Text
1334+ -- ^ Input text.
1335+ -> [Text]
12451336splitOn pat@(Text _ _ l) src@(Text arr off len)
12461337 | l <= 0 = emptyError "splitOn"
12471338 | isSingleton pat = split (== unsafeHead pat) src
12481339 | otherwise = go 0 (indices pat src)
12491340 where
1250− go !s (x:xs) = textP arr (s+off) (x-s) : go (x+l) xs
1251− go s _ = [textP arr (s+off) (len-s)]
1341+ go !s (x:xs) = text arr (s+off) (x-s) : go (x+l) xs
1342+ go s _ = [text arr (s+off) (len-s)]
12521343{-# INLINE [1] splitOn #-}
12531344
12541345{-# RULES
@@ -1340,7 +1431,7 @@ breakOn pat src@(Text arr off len)
13401431 | null pat = emptyError "breakOn"
13411432 | otherwise = case indices pat src of
13421433 [] -> (src, empty)
1343− (x:_) -> (textP arr off x, textP arr (off+x) (len-x))
1434+ (x:_) -> (text arr off x, text arr (off+x) (len-x))
13441435{-# INLINE breakOn #-}
13451436
13461437-- | /O(n+m)/ Similar to 'breakOn', but searches from the end of the
@@ -1382,7 +1473,7 @@ breakOnAll pat src@(Text arr off slen)
13821473 | otherwise = L.map step (indices pat src)
13831474 where
13841475 step x = (chunk 0 x, chunk x (slen-x))
1385− chunk !n !l = textP arr (n+off) l
1476+ chunk !n !l = text arr (n+off) l
13861477{-# INLINE breakOnAll #-}
13871478
13881479-------------------------------------------------------------------------------
@@ -1449,7 +1540,7 @@ countChar c t = S.countChar c (stream t)
14491540-- equivalent to a pair of 'unpack' operations.
14501541zip :: Text -> Text -> [(Char,Char)]
14511542zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
1452−{-# INLINE [0] zip #-}
1543+{-# INLINE zip #-}
14531544
14541545-- | /O(n)/ 'zipWith' generalises 'zip' by zipping with the function
14551546-- given as the first argument, instead of a tupling function.
@@ -1457,7 +1548,7 @@ zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
14571548zipWith :: (Char -> Char -> Char) -> Text -> Text -> Text
14581549zipWith f t1 t2 = unstream (S.zipWith g (stream t1) (stream t2))
14591550 where g a b = safe (f a b)
1460−{-# INLINE [0] zipWith #-}
1551+{-# INLINE zipWith #-}
14611552
14621553-- | /O(n)/ Breaks a 'Text' up into a list of words, delimited by 'Char's
14631554-- representing white space.
@@ -1582,7 +1673,7 @@ isInfixOf needle haystack
15821673-- > fnordLength _ = -1
15831674stripPrefix :: Text -> Text -> Maybe Text
15841675stripPrefix p@(Text _arr _off plen) t@(Text arr off len)
1585− | p `isPrefixOf` t = Just $! textP arr (off+plen) (len-plen)
1676+ | p `isPrefixOf` t = Just $! text arr (off+plen) (len-plen)
15861677 | otherwise = Nothing
15871678
15881679-- | /O(n)/ Find the longest non-empty common prefix of two strings
@@ -1602,8 +1693,8 @@ commonPrefixes t0@(Text arr0 off0 len0) t1@(Text arr1 off1 len1) = go 0 0
16021693 where
16031694 go !i !j | i < len0 && j < len1 && a == b = go (i+d0) (j+d1)
16041695 | i > 0 = Just (Text arr0 off0 i,
1605− textP arr0 (off0+i) (len0-i),
1606− textP arr1 (off1+j) (len1-j))
1696+ text arr0 (off0+i) (len0-i),
1697+ text arr1 (off1+j) (len1-j))
16071698 | otherwise = Nothing
16081699 where Iter a d0 = iter t0 i
16091700 Iter b d1 = iter t1 j
@@ -1628,7 +1719,7 @@ commonPrefixes t0@(Text arr0 off0 len0) t1@(Text arr1 off1 len1) = go 0 0
16281719-- > quuxLength _ = -1
16291720stripSuffix :: Text -> Text -> Maybe Text
16301721stripSuffix p@(Text _arr _off plen) t@(Text arr off len)
1631− | p `isSuffixOf` t = Just $! textP arr off (len-plen)
1722+ | p `isSuffixOf` t = Just $! text arr off (len-plen)
16321723 | otherwise = Nothing
16331724
16341725-- | Add a list of non-negative numbers. Errors out on overflow.
@@ -1657,6 +1748,7 @@ overflowError fun = P.error $ "Data.Text." ++ fun ++ ": size overflow"
16571748copy :: Text -> Text
16581749copy (Text arr off len) = Text (A.run go) 0 len
16591750 where
1751+ go :: ST s (A.MArray s)
16601752 go = do
16611753 marr <- A.new len
16621754 A.copyI marr 0 arr off len