@@ -1,48 +1,29 @@
1−{-# LANGUAGE BangPatterns, CPP, MagicHash, Rank2Types, UnboxedTuples #-}
1+{-# LANGUAGE BangPatterns #-}
22{-# OPTIONS_GHC -fno-warn-orphans #-}
3−#if __GLASGOW_HASKELL__ >= 702
4−{-# LANGUAGE Trustworthy #-}
5−#endif
6−#if __GLASGOW_HASKELL__ >= 708
7−{-# LANGUAGE TypeFamilies #-}
8−#endif
93
104-- |
115-- Module : Data.Text
12−-- Copyright : (c) 2009, 2010, 2011, 2012 Bryan O'Sullivan,
13−-- (c) 2009 Duncan Coutts,
14−-- (c) 2008, 2009 Tom Harper
6+-- Copyright : (c) Tom Harper 2008-2009,
7+-- (c) Bryan O'Sullivan 2009,
8+-- (c) Duncan Coutts 2009
159--
1610-- License : BSD-style
17−-- Maintainer : bos@serpentine.com
11+-- Maintainer : bos@serpentine.com, rtharper@aftereternity.co.uk,
12+-- duncan@haskell.org
1813-- Stability : experimental
1914-- Portability : GHC
2015--
21−-- A time and space-efficient implementation of Unicode text.
22−-- Suitable for performance critical use, both in terms of large data
23−-- quantities and high speed.
24−--
25−-- /Note/: Read below the synopsis for important notes on the use of
26−-- this module.
16+-- A time and space-efficient implementation of Unicode text using
17+-- packed Word16 arrays. Suitable for performance critical use, both
18+-- in terms of large data quantities and high speed.
2719--
2820-- This module is intended to be imported @qualified@, to avoid name
2921-- clashes with "Prelude" functions, e.g.
3022--
3123-- > import qualified Data.Text as T
32−--
33−-- To use an extended and very rich family of functions for working
34−-- with Unicode text (including normalization, regular expressions,
35−-- non-standard encodings, text breaking, and locales), see
36−-- <http://hackage.haskell.org/package/text-icu the text-icu package >.
3724
3825module Data.Text
3926 (
40− -- * Strict vs lazy types
41− -- $strict
42−
43− -- * Acceptable data
44− -- $replacement
45−
4627 -- * Fusion
4728 -- $fusion
4829
@@ -66,7 +47,6 @@ module Data.Text
6647 , init
6748 , null
6849 , length
69− , compareLength
7050
7151 -- * Transformations
7252 , map
@@ -76,17 +56,11 @@ module Data.Text
7656 , reverse
7757 , replace
7858
79− -- ** Case conversion
59+ -- * Case conversion
8060 -- $case
8161 , toCaseFold
8262 , toLower
8363 , toUpper
84− , toTitle
85−
86− -- ** Justification
87− , justifyLeft
88− , justifyRight
89− , center
9064
9165 -- * Folds
9266 , foldl
@@ -125,32 +99,25 @@ module Data.Text
12599
126100 -- ** Breaking strings
127101 , take
128− , takeEnd
129102 , drop
130− , dropEnd
131103 , takeWhile
132− , takeWhileEnd
133104 , dropWhile
134− , dropWhileEnd
105+ , dropAfter
135106 , dropAround
136− , strip
137− , stripStart
138− , stripEnd
139107 , splitAt
140− , breakOn
141− , breakOnEnd
142− , break
143108 , span
109+ , break
144110 , group
145111 , groupBy
146112 , inits
147113 , tails
148114
149115 -- ** Breaking into many substrings
150− -- $split
151− , splitOn
152116 , split
153− , chunksOf
117+ , splitTimes
118+ , splitChar
119+ , splitWith
120+ , breakSubstring
154121
155122 -- ** Breaking into lines and words
156123 , lines
@@ -164,309 +131,137 @@ module Data.Text
164131 , isSuffixOf
165132 , isInfixOf
166133
167− -- ** View patterns
168− , stripPrefix
169− , stripSuffix
170− , commonPrefixes
171−
172134 -- * Searching
135+ , elem
173136 , filter
174− , breakOnAll
175137 , find
176138 , partition
177139
178140 -- , findSubstring
179−
141+
180142 -- * Indexing
181− -- $index
182143 , index
183144 , findIndex
145+ , findIndices
146+ , elemIndex
147+ , elemIndices
184148 , count
185149
186− -- * Zipping
150+ -- * Zipping and unzipping
187151 , zip
188152 , zipWith
189153
190154 -- -* Ordered text
191155 -- , sort
192−
193− -- * Low level operations
194− , copy
195− , unpackCString#
196156 ) where
197157
198−import Prelude (Char, Bool(..), Int, Maybe(..), String,
199− Eq(..), Ord(..), Ordering(..), (++),
200− Read(..),
201− (&&), (||), (+), (-), (.), ($), ($!), (>>),
202− not, return, otherwise, quot)
203−#if defined(HAVE_DEEPSEQ)
204−import Control.DeepSeq (NFData(rnf))
205−#endif
206−#if defined(ASSERTS)
158+import Prelude (Char, Bool(..), Functor(..), Int, Maybe(..), String,
159+ Eq(..), Ord(..), (++),
160+ Read(..), Show(..),
161+ (&&), (||), (+), (-), (.), ($),
162+ not, return, otherwise)
207163import Control.Exception (assert)
208−#endif
209164import Data.Char (isSpace)
210−import Data.Data (Data(gfoldl, toConstr, gunfold, dataTypeOf), constrIndex,
211− Constr, mkConstr, DataType, mkDataType, Fixity(Prefix))
212−import Control.Monad (foldM)
213165import Control.Monad.ST (ST)
214166import qualified Data.Text.Array as A
215167import qualified Data.List as L
216−import Data.Binary (Binary(get, put))
217168import Data.Monoid (Monoid(..))
218−#if MIN_VERSION_base(4,9,0)
219−import Data.Semigroup (Semigroup(..))
220−#endif
169+import Data.Word (Word16)
221170import Data.String (IsString(..))
222−import qualified Data.Text.Internal.Fusion as S
223−import qualified Data.Text.Internal.Fusion.Common as S
224−import Data.Text.Encoding (decodeUtf8', encodeUtf8)
225−import Data.Text.Internal.Fusion (stream, reverseStream, unstream)
226−import Data.Text.Internal.Private (span_)
227−import Data.Text.Internal (Text(..), empty, firstf, mul, safe, text)
228−import Data.Text.Show (singleton, unpack, unpackCString#)
229−import qualified Prelude as P
230−import Data.Text.Unsafe (Iter(..), iter, iter_, lengthWord16, reverseIter,
231− reverseIter_, unsafeHead, unsafeTail)
232−import Data.Text.Internal.Unsafe.Char (unsafeChr)
233−import qualified Data.Text.Internal.Functions as F
234−import qualified Data.Text.Internal.Encoding.Utf16 as U16
235−import Data.Text.Internal.Search (indices)
236−#if defined(__HADDOCK__)
237−import Data.ByteString (ByteString)
238−import qualified Data.Text.Lazy as L
239−import Data.Int (Int64)
240−#endif
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)
247−#endif
248−
249−-- $strict
250−--
251−-- This package provides both strict and lazy 'Text' types. The
252−-- strict type is provided by the "Data.Text" module, while the lazy
253−-- type is provided by the "Data.Text.Lazy" module. Internally, the
254−-- lazy @Text@ type consists of a list of strict chunks.
255−--
256−-- The strict 'Text' type requires that an entire string fit into
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.
264−--
265−-- Each module provides an almost identical API, with the main
266−-- difference being that the strict module uses 'Int' values for
267−-- lengths and counts, while the lazy module uses 'Data.Int.Int64'
268−-- lengths.
269171
270−-- $replacement
271−--
272−-- A 'Text' value is a sequence of Unicode scalar values, as defined
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.
283−--
284−-- Within this module, many functions construct a 'Text' from one or
285−-- more 'Char' values. Those functions will substitute 'Char' values
286−-- that are not valid Unicode scalar values with the replacement
287−-- character \"�\" (U+FFFD). Functions that perform this
288−-- inspection and replacement are documented with the phrase
289−-- \"Performs replacement on invalid scalar values\".
290−--
291−-- (One reason for this policy of replacement is that internally, a
292−-- 'Text' value is represented as packed UTF-16 data. Values in the
293−-- range U+D800 through U+DFFF are used by UTF-16 to denote surrogate
294−-- code points, and so cannot be represented. The functions replace
295−-- invalid scalar values, instead of dropping them, as a security
296−-- measure. For details, see
297−-- <http://unicode.org/reports/tr36/#Deletion_of_Noncharacters Unicode Technical Report 36, §3.5 >.)
172+import qualified Data.Text.Fusion as S
173+import qualified Data.Text.Fusion.Common as S
174+import Data.Text.Fusion (stream, reverseStream, unstream)
175+
176+import Data.Text.Internal (Text(..), empty, text, textP)
177+import qualified Prelude as P
178+import Data.Text.Unsafe (iter, iter_, reverseIter, unsafeHead, unsafeTail)
179+import Data.Text.UnsafeChar (unsafeChr)
180+import qualified Data.Text.Encoding.Utf16 as U16
298181
299182-- $fusion
300183--
301−-- Most of the functions in this module are subject to /fusion/,
302−-- meaning that a pipeline of such functions will usually allocate at
303−-- most one 'Text' value.
304−--
305−-- As an example, consider the following pipeline:
306−--
307−-- > import Data.Text as T
308−-- > import Data.Text.Encoding as E
309−-- > import Data.ByteString (ByteString)
310−-- >
311−-- > countChars :: ByteString -> Int
312−-- > countChars = T.length . T.toUpper . E.decodeUtf8
313−--
314−-- From the type signatures involved, this looks like it should
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'.
320−--
321−-- Functions that can be fused by the compiler are documented with the
322−-- phrase \"Subject to fusion\".
184+-- Most of the functions in this module are subject to /array fusion/,
185+-- meaning that a pipeline of functions will usually allocate at most
186+-- one 'Text' value.
323187
324188instance Eq Text where
325− Text arrA offA lenA == Text arrB offB lenB
326− | lenA == lenB = A.equal arrA offA arrB offB lenA
327− | otherwise = False
189+ t1 == t2 = stream t1 == stream t2
328190 {-# INLINE (==) #-}
329191
330192instance Ord Text where
331− compare = compareText
193+ compare t1 t2 = compare (stream t1) (stream t2)
194+ {-# INLINE compare #-}
195+
196+instance Show Text where
197+ showsPrec p ps r = showsPrec p (unpack ps) r
332198
333199instance Read Text where
334200 readsPrec p str = [(pack x,y) | (x,y) <- readsPrec p str]
335201
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−
344202instance Monoid Text where
345203 mempty = empty
346−#if MIN_VERSION_base(4,9,0)
347− mappend = (<>) -- future-proof definition
348−#else
349204 mappend = append
350−#endif
351205 mconcat = concat
352206
353207instance IsString Text where
354208 fromString = pack
355209
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−
363−#if defined(HAVE_DEEPSEQ)
364−instance NFData Text where rnf !_ = ()
365−#endif
366−
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−
375−-- | This instance preserves data abstraction at the cost of inefficiency.
376−-- We omit reflection services for the sake of data abstraction.
377−--
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.
382−--
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? >
385−--
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, ... >
389−
390−instance Data Text where
391− gfoldl f z txt = z pack `f` (unpack txt)
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]
409−
410−-- | /O(n)/ Compare two 'Text' values lexicographically.
411−compareText :: Text -> Text -> Ordering
412−compareText ta@(Text _arrA _offA lenA) tb@(Text _arrB _offB lenB)
413− | lenA == 0 && lenB == 0 = EQ
414− | otherwise = go 0 0
415− where
416− go !i !j
417− | i >= lenA || j >= lenB = compare lenA lenB
418− | a < b = LT
419− | a > b = GT
420− | otherwise = go (i+di) (j+dj)
421− where Iter a di = iter ta i
422− Iter b dj = iter tb j
423−
424210-- -----------------------------------------------------------------------------
425211-- * Conversion to/from 'Text'
426212
427−-- | /O(n)/ Convert a 'String' into a 'Text'. Subject to
428−-- fusion. Performs replacement on invalid scalar values.
213+-- | /O(n)/ Convert a 'String' into a 'Text'.
214+--
215+-- This function is subject to array fusion.
429216pack :: String -> Text
430−pack = unstream . S.map safe . S.streamList
217+pack = unstream . S.streamList
431218{-# INLINE [1] pack #-}
432219
220+-- | /O(n)/ Convert a Text into a String.
221+-- Subject to array fusion.
222+unpack :: Text -> String
223+unpack = S.unstreamList . stream
224+{-# INLINE [1] unpack #-}
225+
226+-- | /O(1)/ Convert a character into a Text.
227+-- Subject to array fusion.
228+singleton :: Char -> Text
229+singleton = unstream . S.singleton
230+{-# INLINE [1] singleton #-}
231+
433232-- -----------------------------------------------------------------------------
434233-- * Basic functions
435234
436235-- | /O(n)/ Adds a character to the front of a 'Text'. This function
437236-- is more costly than its 'List' counterpart because it requires
438−-- copying a new array. Subject to fusion. Performs replacement on
439−-- invalid scalar values.
237+-- copying a new array. Subject to array fusion.
440238cons :: Char -> Text -> Text
441−cons c t = unstream (S.cons (safe c) (stream t))
239+cons c t = unstream (S.cons c (stream t))
442240{-# INLINE cons #-}
443241
444−infixr 5 `cons`
445−
446242-- | /O(n)/ Adds a character to the end of a 'Text'. This copies the
447−-- entire array in the process, unless fused. Subject to fusion.
448−-- Performs replacement on invalid scalar values.
243+-- entire array in the process. Subject to array fusion.
449244snoc :: Text -> Char -> Text
450−snoc t c = unstream (S.snoc (stream t) (safe c))
245+snoc t c = unstream (S.snoc (stream t) c)
451246{-# INLINE snoc #-}
452247
453248-- | /O(n)/ Appends one 'Text' to the other by copying both of them
454−-- into a new 'Text'. Subject to fusion.
249+-- into a new 'Text'. Subject to array fusion.
455250append :: Text -> Text -> Text
456−append a@(Text arr1 off1 len1) b@(Text arr2 off2 len2)
457− | len1 == 0 = b
458− | len2 == 0 = a
459− | len > 0 = Text (A.run x) 0 len
460− | otherwise = overflowError "append"
251+append (Text arr1 off1 len1) (Text arr2 off2 len2) = Text (A.run x) 0 len
461252 where
462253 len = len1+len2
463− x :: ST s (A.MArray s)
464254 x = do
465− arr <- A.new len
466− A.copyI arr 0 arr1 off1 len1
467− A.copyI arr len1 arr2 off2 len
255+ arr <- A.unsafeNew len :: ST s (A.MArray s Word16)
256+ copy arr1 off1 (len1+off1) arr 0
257+ copy arr2 off2 (len2+off2) arr len1
468258 return arr
469−{-# NOINLINE append #-}
259+ where
260+ copy arr i top arr' j
261+ | i >= top = return ()
262+ | otherwise = do A.unsafeWrite arr' j (arr `A.unsafeIndex` i)
263+ copy arr (i+1) top arr' (j+1)
264+{-# INLINE append #-}
470265
471266{-# RULES
472267"TEXT append -> fused" [~1] forall t1 t2.
@@ -476,26 +271,33 @@ append a@(Text arr1 off1 len1) b@(Text arr2 off2 len2)
476271 #-}
477272
478273-- | /O(1)/ Returns the first character of a 'Text', which must be
479−-- non-empty. Subject to fusion.
274+-- non-empty. Subject to array fusion.
480275head :: Text -> Char
481276head t = S.head (stream t)
482277{-# INLINE head #-}
483278
484279-- | /O(1)/ Returns the first character and rest of a 'Text', or
485−-- 'Nothing' if empty. Subject to fusion.
280+-- 'Nothing' if empty. Subject to array fusion.
486281uncons :: Text -> Maybe (Char, Text)
487282uncons t@(Text arr off len)
488283 | len <= 0 = Nothing
489− | otherwise = Just $ let !(Iter c d) = iter t 0
490− in (c, text arr (off+d) (len-d))
491−{-# INLINE [1] uncons #-}
284+ | otherwise = Just (c, textP arr (off+d) (len-d))
285+ where (c,d) = iter t 0
286+{-# INLINE uncons #-}
492287
493288-- | Lifted from Control.Arrow and specialized.
494289second :: (b -> c) -> (a,b) -> (a,c)
495290second f (a, b) = (a, f b)
496291
292+{-# RULES
293+"TEXT uncons -> fused" [~1] forall t.
294+ uncons t = fmap (second unstream) (S.uncons (stream t))
295+"TEXT uncons -> unfused" [1] forall t.
296+ fmap (second unstream) (S.uncons (stream t)) = uncons t
297+ #-}
298+
497299-- | /O(1)/ Returns the last character of a 'Text', which must be
498−-- non-empty. Subject to fusion.
300+-- non-empty. Subject to array fusion.
499301last :: Text -> Char
500302last (Text arr off len)
501303 | len <= 0 = emptyError "last"
@@ -513,11 +315,11 @@ last (Text arr off len)
513315 #-}
514316
515317-- | /O(1)/ Returns all characters after the head of a 'Text', which
516−-- must be non-empty. Subject to fusion.
318+-- must be non-empty. Subject to array fusion.
517319tail :: Text -> Text
518320tail t@(Text arr off len)
519321 | len <= 0 = emptyError "tail"
520− | otherwise = text arr (off+d) (len-d)
322+ | otherwise = textP arr (off+d) (len-d)
521323 where d = iter_ t 0
522324{-# INLINE [1] tail #-}
523325
@@ -529,11 +331,11 @@ tail t@(Text arr off len)
529331 #-}
530332
531333-- | /O(1)/ Returns all but the last character of a 'Text', which must
532−-- be non-empty. Subject to fusion.
334+-- be non-empty. Subject to array fusion.
533335init :: Text -> Text
534336init (Text arr off len) | len <= 0 = emptyError "init"
535− | n >= 0xDC00 && n <= 0xDFFF = text arr off (len-2)
536− | otherwise = text arr off (len-1)
337+ | n >= 0xDC00 && n <= 0xDFFF = textP arr off (len-2)
338+ | otherwise = textP arr off (len-1)
537339 where
538340 n = A.unsafeIndex arr (off+len-1)
539341{-# INLINE [1] init #-}
@@ -545,14 +347,10 @@ init (Text arr off len) | len <= 0 = emptyError "init"
545347 unstream (S.init (stream t)) = init t
546348 #-}
547349
548−-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to
350+-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to array
549351-- fusion.
550352null :: Text -> Bool
551−null (Text _arr _off len) =
552−#if defined(ASSERTS)
553− assert (len >= 0) $
554−#endif
555− len <= 0
353+null (Text _arr _off len) = assert (len >= 0) $ len <= 0
556354{-# INLINE [1] null #-}
557355
558356{-# RULES
@@ -562,171 +360,61 @@ null (Text _arr _off len) =
562360 S.null (stream t) = null t
563361 #-}
564362
565−-- | /O(1)/ Tests whether a 'Text' contains exactly one character.
566−-- Subject to fusion.
567−isSingleton :: Text -> Bool
568−isSingleton = S.isSingleton . stream
569−{-# INLINE isSingleton #-}
570−
571363-- | /O(n)/ Returns the number of characters in a 'Text'.
572−-- Subject to fusion.
364+-- Subject to array fusion.
573365length :: Text -> Int
574366length t = S.length (stream t)
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.
578−
579−-- | /O(n)/ Compare the count of characters in a 'Text' to a number.
580−-- Subject to fusion.
581−--
582−-- This function gives the same answer as comparing against the result
583−-- of 'length', but can short circuit if the count of characters is
584−-- greater than the number, and hence be more efficient.
585−compareLength :: Text -> Int -> Ordering
586−compareLength t n = S.compareLengthI (stream t) n
587−{-# INLINE [1] compareLength #-}
588−
589−{-# RULES
590−"TEXT compareN/length -> compareLength" [~1] forall t n.
591− compare (length t) n = compareLength t n
592− #-}
593−
594−{-# RULES
595−"TEXT ==N/length -> compareLength/==EQ" [~1] forall t n.
596− eqInt (length t) n = compareLength t n == EQ
597− #-}
598−
599−{-# RULES
600−"TEXT /=N/length -> compareLength//=EQ" [~1] forall t n.
601− neInt (length t) n = compareLength t n /= EQ
602− #-}
603−
604−{-# RULES
605−"TEXT <N/length -> compareLength/==LT" [~1] forall t n.
606− ltInt (length t) n = compareLength t n == LT
607− #-}
608−
609−{-# RULES
610−"TEXT <=N/length -> compareLength//=GT" [~1] forall t n.
611− leInt (length t) n = compareLength t n /= GT
612− #-}
613−
614−{-# RULES
615−"TEXT >N/length -> compareLength/==GT" [~1] forall t n.
616− gtInt (length t) n = compareLength t n == GT
617− #-}
618−
619−{-# RULES
620−"TEXT >=N/length -> compareLength//=LT" [~1] forall t n.
621− geInt (length t) n = compareLength t n /= LT
622− #-}
367+{-# INLINE length #-}
623368
624369-- -----------------------------------------------------------------------------
625370-- * Transformations
626371-- | /O(n)/ 'map' @f@ @t@ is the 'Text' obtained by applying @f@ to
627−-- each element of @t@. Subject to fusion. Performs replacement on
628−-- invalid scalar values.
372+-- each element of @t@. Subject to array fusion.
629373map :: (Char -> Char) -> Text -> Text
630−map f t = unstream (S.map (safe . f) (stream t))
374+map f t = unstream (S.map f (stream t))
631375{-# INLINE [1] map #-}
632376
633377-- | /O(n)/ The 'intercalate' function takes a 'Text' and a list of
634378-- 'Text's and concatenates the list after interspersing the first
635379-- argument between each element of the list.
636380intercalate :: Text -> [Text] -> Text
637−intercalate t = concat . (F.intersperse t)
381+intercalate t ts = unstream (S.intercalate (stream t) (L.map stream ts))
638382{-# INLINE intercalate #-}
639383
640384-- | /O(n)/ The 'intersperse' function takes a character and places it
641−-- between the characters of a 'Text'. Subject to fusion. Performs
642−-- replacement on invalid scalar values.
385+-- between the characters of a 'Text'. Subject to array fusion.
643386intersperse :: Char -> Text -> Text
644−intersperse c t = unstream (S.intersperse (safe c) (stream t))
387+intersperse c t = unstream (S.intersperse c (stream t))
645388{-# INLINE intersperse #-}
646389
647−-- | /O(n)/ Reverse the characters of a string. Subject to fusion.
390+-- | /O(n)/ Reverse the characters of a string. Subject to array fusion.
648391reverse :: Text -> Text
649392reverse t = S.reverse (stream t)
650393{-# INLINE reverse #-}
651394
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"
672−--
673−-- In (unlikely) bad cases, this function's time complexity degrades
674−-- towards /O(n*m)/.
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.
395+-- | /O(m)*O(n)/ Replace every occurrence of one substring with another.
396+replace :: Text -- ^ Text to search for
397+ -> Text -- ^ Replacement text
398+ -> Text -- ^ Input text
680399 -> Text
681− -- ^ @haystack@ in which to search.
682− -> Text
683−replace needle@(Text _ _ neeLen)
684− (Text repArr repOff repLen)
685− haystack@(Text hayArr hayOff hayLen)
686− | neeLen == 0 = emptyError "replace"
687− | L.null ixs = haystack
688− | len > 0 = Text (A.run x) 0 len
689− | otherwise = empty
690− where
691− ixs = indices needle haystack
692− len = hayLen - (neeLen - repLen) `mul` L.length ixs
693− x :: ST s (A.MArray s)
694− x = do
695− marr <- A.new len
696− let loop (i:is) o d = do
697− let d0 = d + i - o
698− d1 = d0 + repLen
699− A.copyI marr d hayArr (hayOff+o) d0
700− A.copyI marr d0 repArr repOff d1
701− loop is (i + neeLen) d1
702− loop [] o d = A.copyI marr d hayArr (hayOff+o) len
703− loop ixs 0 0
704− return marr
400+replace s d = intercalate d . split s
401+{-# INLINE replace #-}
705402
706403-- ----------------------------------------------------------------------------
707404-- ** Case conversions (folds)
708405
709406-- $case
710407--
711−-- When case converting 'Text' values, do not use combinators like
712−-- @map toUpper@ to case convert each character of a string
713−-- individually, as this gives incorrect results according to the
714−-- rules of some writing systems. The whole-string case conversion
715−-- functions from this module, such as @toUpper@, obey the correct
716−-- case conversion rules. As a result, these functions may map one
717−-- input character to two or three output characters. For examples,
718−-- see the documentation of each function.
719−--
720−-- /Note/: In some languages, case conversion is a locale- and
721−-- context-dependent operation. The case conversion functions in this
722−-- module are /not/ locale sensitive. Programs that require locale
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 >.
408+-- With Unicode text, it is incorrect to use combinators like @map
409+-- toUpper@ to case convert each character of a string individually.
410+-- Instead, use the whole-string case conversion functions from this
411+-- module. For correctness in different writing systems, these
412+-- functions may map one input character to two or three output
413+-- characters.
725414
726−-- | /O(n)/ Convert a string to folded case. Subject to fusion.
727−--
728−-- This function is mainly useful for performing caseless (also known
729−-- as case insensitive) string comparisons.
415+-- | /O(n)/ Convert a string to folded case. This function is mainly
416+-- useful for performing caseless (or case insensitive) string
417+-- comparisons.
730418--
731419-- A string @x@ is a caseless match for a string @y@ if and only if:
732420--
@@ -734,112 +422,31 @@ replace needle@(Text _ _ neeLen)
734422--
735423-- The result string may be longer than the input string, and may
736424-- differ from applying 'toLower' to the input string. For instance,
737−-- the Armenian small ligature \"ﬓ\" (men now, U+FB13) is case
738−-- folded to the sequence \"մ\" (men, U+0574) followed by
739−-- \"ն\" (now, U+0576), while the Greek \"µ\" (micro sign,
740−-- U+00B5) is case folded to \"μ\" (small letter mu, U+03BC)
741−-- instead of itself.
425+-- the Armenian small ligature men now (U+FB13) is case folded to the
426+-- bigram men now (U+0574 U+0576), while the micro sign (U+00B5) is
427+-- case folded to the Greek small letter letter mu (U+03BC) instead of
428+-- itself.
742429toCaseFold :: Text -> Text
743430toCaseFold t = unstream (S.toCaseFold (stream t))
744−{-# INLINE toCaseFold #-}
431+{-# INLINE [0] toCaseFold #-}
745432
746433-- | /O(n)/ Convert a string to lower case, using simple case
747−-- conversion. Subject to fusion.
748−--
749−-- The result string may be longer than the input string. For
750−-- instance, \"İ\" (Latin capital letter I with dot above,
751−-- U+0130) maps to the sequence \"i\" (Latin small letter i, U+0069)
752−-- followed by \" ̇\" (combining dot above, U+0307).
434+-- conversion. The result string may be longer than the input string.
435+-- For instance, the Latin capital letter I with dot above (U+0130)
436+-- maps to the bigram Latin small letter i (U+0069) followed by
437+-- combining dot above (U+0307).
753438toLower :: Text -> Text
754439toLower t = unstream (S.toLower (stream t))
755440{-# INLINE toLower #-}
756441
757442-- | /O(n)/ Convert a string to upper case, using simple case
758−-- conversion. Subject to fusion.
759−--
760−-- The result string may be longer than the input string. For
761−-- instance, the German \"ß\" (eszett, U+00DF) maps to the
762−-- two-letter sequence \"SS\".
443+-- conversion. The result string may be longer than the input string.
444+-- For instance, the German eszett (U+00DF) maps to the two-letter
445+-- sequence SS.
763446toUpper :: Text -> Text
764447toUpper t = unstream (S.toUpper (stream t))
765448{-# INLINE toUpper #-}
766449
767−-- | /O(n)/ Convert a string to title case, using simple case
768−-- conversion. Subject to fusion.
769−--
770−-- The first letter of the input is converted to title case, as is
771−-- every subsequent letter that immediately follows a non-letter.
772−-- Every letter that immediately follows another letter is converted
773−-- to lower case.
774−--
775−-- The result string may be longer than the input string. For example,
776−-- the Latin small ligature fl (U+FB02) is converted to the
777−-- sequence Latin capital letter F (U+0046) followed by Latin small
778−-- letter l (U+006C).
779−--
780−-- /Note/: this function does not take language or culture specific
781−-- rules into account. For instance, in English, different style
782−-- guides disagree on whether the book name \"The Hill of the Red
783−-- Fox\" is correctly title cased—but this function will
784−-- capitalize /every/ word.
785−toTitle :: Text -> Text
786−toTitle t = unstream (S.toTitle (stream t))
787−{-# INLINE toTitle #-}
788−
789−-- | /O(n)/ Left-justify a string to the given length, using the
790−-- specified fill character on the right. Subject to fusion.
791−-- Performs replacement on invalid scalar values.
792−--
793−-- Examples:
794−--
795−-- > justifyLeft 7 'x' "foo" == "fooxxxx"
796−-- > justifyLeft 3 'x' "foobar" == "foobar"
797−justifyLeft :: Int -> Char -> Text -> Text
798−justifyLeft k c t
799− | len >= k = t
800− | otherwise = t `append` replicateChar (k-len) c
801− where len = length t
802−{-# INLINE [1] justifyLeft #-}
803−
804−{-# RULES
805−"TEXT justifyLeft -> fused" [~1] forall k c t.
806− justifyLeft k c t = unstream (S.justifyLeftI k c (stream t))
807−"TEXT justifyLeft -> unfused" [1] forall k c t.
808− unstream (S.justifyLeftI k c (stream t)) = justifyLeft k c t
809− #-}
810−
811−-- | /O(n)/ Right-justify a string to the given length, using the
812−-- specified fill character on the left. Performs replacement on
813−-- invalid scalar values.
814−--
815−-- Examples:
816−--
817−-- > justifyRight 7 'x' "bar" == "xxxxbar"
818−-- > justifyRight 3 'x' "foobar" == "foobar"
819−justifyRight :: Int -> Char -> Text -> Text
820−justifyRight k c t
821− | len >= k = t
822− | otherwise = replicateChar (k-len) c `append` t
823− where len = length t
824−{-# INLINE justifyRight #-}
825−
826−-- | /O(n)/ Center a string to the given length, using the specified
827−-- fill character on either side. Performs replacement on invalid
828−-- scalar values.
829−--
830−-- Examples:
831−--
832−-- > center 8 'x' "HS" = "xxxHSxxx"
833−center :: Int -> Char -> Text -> Text
834−center k c t
835− | len >= k = t
836− | otherwise = replicateChar l c `append` t `append` replicateChar r c
837− where len = length t
838− d = k - len
839− r = d `quot` 2
840− l = d - r
841−{-# INLINE center #-}
842−
843450-- | /O(n)/ The 'transpose' function transposes the rows and columns
844451-- of its 'Text' argument. Note that this function uses 'pack',
845452-- 'unpack', and the list version of transpose, and is thus not very
@@ -853,23 +460,26 @@ transpose ts = P.map pack (L.transpose (P.map unpack ts))
853460-- | /O(n)/ 'foldl', applied to a binary operator, a starting value
854461-- (typically the left-identity of the operator), and a 'Text',
855462-- reduces the 'Text' using the binary operator, from left to right.
856−-- Subject to fusion.
857−foldl :: (a -> Char -> a) -> a -> Text -> a
463+-- Subject to array fusion.
464+foldl :: (b -> Char -> b) -> b -> Text -> b
858465foldl f z t = S.foldl f z (stream t)
859466{-# INLINE foldl #-}
860467
861−-- | /O(n)/ A strict version of 'foldl'. Subject to fusion.
862−foldl' :: (a -> Char -> a) -> a -> Text -> a
468+-- | /O(n)/ A strict version of 'foldl'.
469+-- Subject to array fusion.
470+foldl' :: (b -> Char -> b) -> b -> Text -> b
863471foldl' f z t = S.foldl' f z (stream t)
864472{-# INLINE foldl' #-}
865473
866474-- | /O(n)/ A variant of 'foldl' that has no starting value argument,
867−-- and thus must be applied to a non-empty 'Text'. Subject to fusion.
475+-- and thus must be applied to a non-empty 'Text'. Subject to array
476+-- fusion.
868477foldl1 :: (Char -> Char -> Char) -> Text -> Char
869478foldl1 f t = S.foldl1 f (stream t)
870479{-# INLINE foldl1 #-}
871480
872−-- | /O(n)/ A strict version of 'foldl1'. Subject to fusion.
481+-- | /O(n)/ A strict version of 'foldl1'.
482+-- Subject to array fusion.
873483foldl1' :: (Char -> Char -> Char) -> Text -> Char
874484foldl1' f t = S.foldl1' f (stream t)
875485{-# INLINE foldl1' #-}
@@ -877,13 +487,13 @@ foldl1' f t = S.foldl1' f (stream t)
877487-- | /O(n)/ 'foldr', applied to a binary operator, a starting value
878488-- (typically the right-identity of the operator), and a 'Text',
879489-- reduces the 'Text' using the binary operator, from right to left.
880−-- Subject to fusion.
881−foldr :: (Char -> a -> a) -> a -> Text -> a
490+-- Subject to array fusion.
491+foldr :: (Char -> b -> b) -> b -> Text -> b
882492foldr f z t = S.foldr f z (stream t)
883493{-# INLINE foldr #-}
884494
885−-- | /O(n)/ A variant of 'foldr' that has no starting value argument,
886−-- and thus must be applied to a non-empty 'Text'. Subject to
495+-- | /O(n)/ A variant of 'foldr' that has no starting value argument, and
496+-- thust must be applied to a non-empty 'Text'. Subject to array
887497-- fusion.
888498foldr1 :: (Char -> Char -> Char) -> Text -> Char
889499foldr1 f t = S.foldr1 f (stream t)
@@ -892,48 +502,40 @@ foldr1 f t = S.foldr1 f (stream t)
892502-- -----------------------------------------------------------------------------
893503-- ** Special folds
894504
895−-- | /O(n)/ Concatenate a list of 'Text's.
505+-- | /O(n)/ Concatenate a list of 'Text's. Subject to array fusion.
896506concat :: [Text] -> Text
897−concat ts = case ts' of
898− [] -> empty
899− [t] -> t
900− _ -> Text (A.run go) 0 len
901− where
902− ts' = L.filter (not . null) ts
903− len = sumP "concat" $ L.map lengthWord16 ts'
904− go :: ST s (A.MArray s)
905− go = do
906− arr <- A.new len
907− let step i (Text a o l) =
908− let !j = i + l in A.copyI arr i a o j >> return j
909− foldM step 0 ts' >> return arr
507+concat ts = unstream (S.concat (L.map stream ts))
508+{-# INLINE concat #-}
910509
911510-- | /O(n)/ Map a function over a 'Text' that results in a 'Text', and
912−-- concatenate the results.
511+-- concatenate the results. This function is subject to array fusion.
512+--
513+-- Note: if in 'concatMap' @f@ @t@, @f@ is defined in terms of fusible
514+-- functions, it will also be fusible.
913515concatMap :: (Char -> Text) -> Text -> Text
914−concatMap f = concat . foldr ((:) . f) []
516+concatMap f t = unstream (S.concatMap (stream . f) (stream t))
915517{-# INLINE concatMap #-}
916518
917519-- | /O(n)/ 'any' @p@ @t@ determines whether any character in the
918−-- 'Text' @t@ satisfies the predicate @p@. Subject to fusion.
520+-- 'Text' @t@ satisifes the predicate @p@. Subject to array fusion.
919521any :: (Char -> Bool) -> Text -> Bool
920522any p t = S.any p (stream t)
921523{-# INLINE any #-}
922524
923525-- | /O(n)/ 'all' @p@ @t@ determines whether all characters in the
924−-- 'Text' @t@ satisfy the predicate @p@. Subject to fusion.
526+-- 'Text' @t@ satisify the predicate @p@. Subject to array fusion.
925527all :: (Char -> Bool) -> Text -> Bool
926528all p t = S.all p (stream t)
927529{-# INLINE all #-}
928530
929531-- | /O(n)/ 'maximum' returns the maximum value from a 'Text', which
930−-- must be non-empty. Subject to fusion.
532+-- must be non-empty. Subject to array fusion.
931533maximum :: Text -> Char
932534maximum t = S.maximum (stream t)
933535{-# INLINE maximum #-}
934536
935537-- | /O(n)/ 'minimum' returns the minimum value from a 'Text', which
936−-- must be non-empty. Subject to fusion.
538+-- must be non-empty. Subject to array fusion.
937539minimum :: Text -> Char
938540minimum t = S.minimum (stream t)
939541{-# INLINE minimum #-}
@@ -942,8 +544,8 @@ minimum t = S.minimum (stream t)
942544-- * Building 'Text's
943545
944546-- | /O(n)/ 'scanl' is similar to 'foldl', but returns a list of
945−-- successive reduced values from the left. Subject to fusion.
946−-- Performs replacement on invalid scalar values.
547+-- successive reduced values from the left. This function is subject
548+-- to array fusion.
947549--
948550-- > scanl f z [x1, x2, ...] == [z, z `f` x1, (z `f` x1) `f` x2, ...]
949551--
@@ -951,13 +553,11 @@ minimum t = S.minimum (stream t)
951553--
952554-- > last (scanl f z xs) == foldl f z xs.
953555scanl :: (Char -> Char -> Char) -> Char -> Text -> Text
954−scanl f z t = unstream (S.scanl g z (stream t))
955− where g a b = safe (f a b)
556+scanl f z t = unstream (S.scanl f z (stream t))
956557{-# INLINE scanl #-}
957558
958559-- | /O(n)/ 'scanl1' is a variant of 'scanl' that has no starting
959−-- value argument. Subject to fusion. Performs replacement on
960−-- invalid scalar values.
560+-- value argument. This function is subject to array fusion.
961561--
962562-- > scanl1 f [x1, x2, ...] == [x1, x1 `f` x2, ...]
963563scanl1 :: (Char -> Char -> Char) -> Text -> Text
@@ -965,75 +565,49 @@ scanl1 f t | null t = empty
965565 | otherwise = scanl f (unsafeHead t) (unsafeTail t)
966566{-# INLINE scanl1 #-}
967567
968−-- | /O(n)/ 'scanr' is the right-to-left dual of 'scanl'. Performs
969−-- replacement on invalid scalar values.
568+-- | /O(n)/ 'scanr' is the right-to-left dual of 'scanl'.
970569--
971570-- > scanr f v == reverse . scanl (flip f) v . reverse
972571scanr :: (Char -> Char -> Char) -> Char -> Text -> Text
973−scanr f z = S.reverse . S.reverseScanr g z . reverseStream
974− where g a b = safe (f a b)
572+scanr f z = S.reverse . S.reverseScanr f z . reverseStream
975573{-# INLINE scanr #-}
976574
977575-- | /O(n)/ 'scanr1' is a variant of 'scanr' that has no starting
978−-- value argument. Subject to fusion. Performs replacement on
979−-- invalid scalar values.
576+-- value argument. This function is subject to array fusion.
980577scanr1 :: (Char -> Char -> Char) -> Text -> Text
981578scanr1 f t | null t = empty
982579 | otherwise = scanr f (last t) (init t)
983580{-# INLINE scanr1 #-}
984581
985−-- | /O(n)/ Like a combination of 'map' and 'foldl''. Applies a
582+-- | /O(n)/ Like a combination of 'map' and 'foldl'. Applies a
986583-- function to each element of a 'Text', passing an accumulating
987−-- parameter from left to right, and returns a final 'Text'. Performs
988−-- replacement on invalid scalar values.
584+-- parameter from left to right, and returns a final 'Text'.
989585mapAccumL :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
990−mapAccumL f z0 = S.mapAccumL g z0 . stream
991− where g a b = second safe (f a b)
992−{-# INLINE mapAccumL #-}
586+mapAccumL f s t = case uncons t of
587+ Nothing -> (s, empty)
588+ Just (x, xs) -> (s'', cons y ys)
589+ where (s', y ) = f s x
590+ (s'',ys) = mapAccumL f s' xs
993591
994592-- | The 'mapAccumR' function behaves like a combination of 'map' and
995−-- a strict 'foldr'; it applies a function to each element of a
996−-- 'Text', passing an accumulating parameter from right to left, and
997−-- returning a final value of this accumulator together with the new
998−-- 'Text'.
999−-- Performs replacement on invalid scalar values.
593+-- 'foldr'; it applies a function to each element of a 'Text', passing
594+-- an accumulating parameter from right to left, and returning a final
595+-- value of this accumulator together with the new 'Text'.
1000596mapAccumR :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
1001−mapAccumR f z0 = second reverse . S.mapAccumL g z0 . reverseStream
1002− where g a b = second safe (f a b)
1003−{-# INLINE mapAccumR #-}
597+mapAccumR f s t = case uncons t of
598+ Nothing -> (s, empty)
599+ Just (x, xs) -> (s'', cons y ys)
600+ where (s'',y ) = f s' x
601+ (s', ys) = mapAccumR f s xs
1004602
1005603-- -----------------------------------------------------------------------------
1006604-- ** Generating and unfolding 'Text's
1007605
1008−-- | /O(n*m)/ 'replicate' @n@ @t@ is a 'Text' consisting of the input
1009−-- @t@ repeated @n@ times.
1010−replicate :: Int -> Text -> Text
1011−replicate n t@(Text a o l)
1012− | n <= 0 || l <= 0 = empty
1013− | n == 1 = t
1014− | isSingleton t = replicateChar n (unsafeHead t)
1015− | otherwise = Text (A.run x) 0 len
1016− where
1017− len = l `mul` n
1018− x :: ST s (A.MArray s)
1019− x = do
1020− arr <- A.new len
1021− let loop !d !i | i >= n = return arr
1022− | otherwise = let m = d + l
1023− in A.copyI arr d a o m >> loop m (i+1)
1024− loop 0 0
1025−{-# INLINE [1] replicate #-}
1026−
1027−{-# RULES
1028−"TEXT replicate/singleton -> replicateChar" [~1] forall n c.
1029− replicate n (singleton c) = replicateChar n c
1030− #-}
1031−
1032−-- | /O(n)/ 'replicateChar' @n@ @c@ is a 'Text' of length @n@ with @c@ the
606+-- | /O(n)/ 'replicate' @n@ @c@ is a 'Text' of length @n@ with @c@ the
1033607-- value of every element. Subject to fusion.
1034−replicateChar :: Int -> Char -> Text
1035−replicateChar n c = unstream (S.replicateCharI n (safe c))
1036−{-# INLINE replicateChar #-}
608+replicate :: Int -> Char -> Text
609+replicate n c = unstream (S.replicate n c)
610+{-# INLINE replicate #-}
1037611
1038612-- | /O(n)/, where @n@ is the length of the result. The 'unfoldr'
1039613-- function is analogous to the List 'L.unfoldr'. 'unfoldr' builds a
@@ -1041,9 +615,9 @@ replicateChar n c = unstream (S.replicateCharI n (safe c))
1041615-- returns 'Nothing' if it is done producing the 'Text', otherwise
1042616-- 'Just' @(a,b)@. In this case, @a@ is the next 'Char' in the
1043617-- string, and @b@ is the seed value for further production. Subject
1044−-- to fusion. Performs replacement on invalid scalar values.
618+-- to fusion.
1045619unfoldr :: (a -> Maybe (Char,a)) -> a -> Text
1046−unfoldr f s = unstream (S.unfoldr (firstf safe . f) s)
620+unfoldr f s = unstream (S.unfoldr f s)
1047621{-# INLINE unfoldr #-}
1048622
1049623-- | /O(n)/ Like 'unfoldr', 'unfoldrN' builds a 'Text' from a seed
@@ -1051,9 +625,9 @@ unfoldr f s = unstream (S.unfoldr (firstf safe . f) s)
1051625-- first argument to 'unfoldrN'. This function is more efficient than
1052626-- 'unfoldr' when the maximum length of the result is known and
1053627-- correct, otherwise its performance is similar to 'unfoldr'. Subject
1054−-- to fusion. Performs replacement on invalid scalar values.
628+-- to fusion.
1055629unfoldrN :: Int -> (a -> Maybe (Char,a)) -> a -> Text
1056−unfoldrN n f s = unstream (S.unfoldrN n (firstf safe . f) s)
630+unfoldrN n f s = unstream (S.unfoldrN n f s)
1057631{-# INLINE unfoldrN #-}
1058632
1059633-- -----------------------------------------------------------------------------
@@ -1066,16 +640,14 @@ take :: Int -> Text -> Text
1066640take n t@(Text arr off len)
1067641 | n <= 0 = empty
1068642 | n >= len = t
1069− | otherwise = text arr off (iterN n t)
643+ | otherwise = Text arr off (loop 0 0)
644+ where
645+ loop !i !cnt
646+ | i >= len || cnt >= n = i
647+ | otherwise = loop (i+d) (cnt+1)
648+ where d = iter_ t i
1070649{-# INLINE [1] take #-}
1071650
1072−iterN :: Int -> Text -> Int
1073−iterN n t@(Text _arr _off len) = loop 0 0
1074− where loop !i !cnt
1075− | i >= len || cnt >= n = i
1076− | otherwise = loop (i+d) (cnt+1)
1077− where d = iter_ t i
1078−
1079651{-# RULES
1080652"TEXT take -> fused" [~1] forall n t.
1081653 take n t = unstream (S.take n (stream t))
@@ -1083,36 +655,18 @@ iterN n t@(Text _arr _off len) = loop 0 0
1083655 unstream (S.take n (stream t)) = take n t
1084656 #-}
1085657
1086−-- | /O(n)/ 'takeEnd' @n@ @t@ returns the suffix remaining after
1087−-- taking @n@ characters from the end of @t@.
1088−--
1089−-- Examples:
1090−--
1091−-- > takeEnd 3 "foobar" == "bar"
1092−takeEnd :: Int -> Text -> Text
1093−takeEnd n t@(Text arr off len)
1094− | n <= 0 = empty
1095− | n >= len = t
1096− | otherwise = text arr (off+i) (len-i)
1097− where i = iterNEnd n t
1098−
1099−iterNEnd :: Int -> Text -> Int
1100−iterNEnd n t@(Text _arr _off len) = loop (len-1) n
1101− where loop i !m
1102− | i <= 0 = 0
1103− | m <= 1 = i
1104− | otherwise = loop (i+d) (m-1)
1105− where d = reverseIter_ t i
1106−
1107658-- | /O(n)/ 'drop' @n@, applied to a 'Text', returns the suffix of the
1108−-- 'Text' after the first @n@ characters, or the empty 'Text' if @n@
1109−-- is greater than the length of the 'Text'. Subject to fusion.
659+-- 'Text' of length @n@, or the empty 'Text' if @n@ is greater than the
660+-- length of the 'Text'. Subject to fusion.
1110661drop :: Int -> Text -> Text
1111662drop n t@(Text arr off len)
1112663 | n <= 0 = t
1113664 | n >= len = empty
1114− | otherwise = text arr (off+i) (len-i)
1115− where i = iterN n t
665+ | otherwise = loop 0 0
666+ where loop !i !cnt
667+ | i >= len || cnt >= n = Text arr (off+i) (len-i)
668+ | otherwise = loop (i+d) (cnt+1)
669+ where d = iter_ t i
1116670{-# INLINE [1] drop #-}
1117671
1118672{-# RULES
@@ -1122,27 +676,15 @@ drop n t@(Text arr off len)
1122676 unstream (S.drop n (stream t)) = drop n t
1123677 #-}
1124678
1125−-- | /O(n)/ 'dropEnd' @n@ @t@ returns the prefix remaining after
1126−-- dropping @n@ characters from the end of @t@.
1127−--
1128−-- Examples:
1129−--
1130−-- > dropEnd 3 "foobar" == "foo"
1131−dropEnd :: Int -> Text -> Text
1132−dropEnd n t@(Text arr off len)
1133− | n <= 0 = t
1134− | n >= len = empty
1135− | otherwise = text arr off (iterNEnd n t)
1136−
1137−-- | /O(n)/ 'takeWhile', applied to a predicate @p@ and a 'Text',
1138−-- returns the longest prefix (possibly empty) of elements that
1139−-- satisfy @p@. Subject to fusion.
679+-- | /O(n)/ 'takeWhile', applied to a predicate @p@ and a 'Text', returns
680+-- the longest prefix (possibly empty) of elements that satisfy @p@.
681+-- This function is subject to array fusion.
1140682takeWhile :: (Char -> Bool) -> Text -> Text
1141683takeWhile p t@(Text arr off len) = loop 0
1142684 where loop !i | i >= len = t
1143685 | p c = loop (i+d)
1144− | otherwise = text arr off i
1145− where Iter c d = iter t i
686+ | otherwise = textP arr off i
687+ where (c,d) = iter t i
1146688{-# INLINE [1] takeWhile #-}
1147689
1148690{-# RULES
@@ -1152,35 +694,14 @@ takeWhile p t@(Text arr off len) = loop 0
1152694 unstream (S.takeWhile p (stream t)) = takeWhile p t
1153695 #-}
1154696
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−
1176697-- | /O(n)/ 'dropWhile' @p@ @t@ returns the suffix remaining after
1177−-- 'takeWhile' @p@ @t@. Subject to fusion.
698+-- 'takeWhile' @p@ @t@. This function is subject to array fusion.
1178699dropWhile :: (Char -> Bool) -> Text -> Text
1179700dropWhile p t@(Text arr off len) = loop 0 0
1180701 where loop !i !l | l >= len = empty
1181702 | p c = loop (i+d) (l+d)
1182703 | otherwise = Text arr (off+i) (len-l)
1183− where Iter c d = iter t i
704+ where (c,d) = iter t i
1184705{-# INLINE [1] dropWhile #-}
1185706
1186707{-# RULES
@@ -1190,56 +711,39 @@ dropWhile p t@(Text arr off len) = loop 0 0
1190711 unstream (S.dropWhile p (stream t)) = dropWhile p t
1191712 #-}
1192713
1193−-- | /O(n)/ 'dropWhileEnd' @p@ @t@ returns the prefix remaining after
1194−-- dropping characters that satisfy the predicate @p@ from the end of
1195−-- @t@. Subject to fusion.
714+-- | /O(n)/ 'dropAfter' @p@ @t@ returns the prefix remaining after
715+-- dropping characters that fail the predicate @p@ from the end of
716+-- @t@. This function is subject to array fusion.
1196717--
1197−-- Examples:
718+-- For example, to drop trailing white space:
1198719--
1199−-- > dropWhileEnd (=='.') "foo..." == "foo"
1200−dropWhileEnd :: (Char -> Bool) -> Text -> Text
1201−dropWhileEnd p t@(Text arr off len) = loop (len-1) len
720+-- > dropAfter isSpace
721+dropAfter :: (Char -> Bool) -> Text -> Text
722+dropAfter p t@(Text arr off len) = loop (len-1) len
1202723 where loop !i !l | l <= 0 = empty
1203724 | p c = loop (i+d) (l+d)
1204725 | otherwise = Text arr off l
1205726 where (c,d) = reverseIter t i
1206−{-# INLINE [1] dropWhileEnd #-}
727+{-# INLINE [1] dropAfter #-}
1207728
1208729{-# RULES
1209−"TEXT dropWhileEnd -> fused" [~1] forall p t.
1210− dropWhileEnd p t = S.reverse (S.dropWhile p (S.reverseStream t))
1211−"TEXT dropWhileEnd -> unfused" [1] forall p t.
1212− S.reverse (S.dropWhile p (S.reverseStream t)) = dropWhileEnd p t
730+"TEXT dropAfter -> fused" [~1] forall p t.
731+ dropAfter p t = S.reverse (S.dropWhile p (S.reverseStream t))
732+"TEXT dropAfter -> unfused" [1] forall p t.
733+ S.reverse (S.dropWhile p (S.reverseStream t)) = dropAfter p t
1213734 #-}
1214735
1215736-- | /O(n)/ 'dropAround' @p@ @t@ returns the substring remaining after
1216−-- dropping characters that satisfy the predicate @p@ from both the
1217−-- beginning and end of @t@. Subject to fusion.
1218−dropAround :: (Char -> Bool) -> Text -> Text
1219−dropAround p = dropWhile p . dropWhileEnd p
1220−{-# INLINE [1] dropAround #-}
1221−
1222−-- | /O(n)/ Remove leading white space from a string. Equivalent to:
1223−--
1224−-- > dropWhile isSpace
1225−stripStart :: Text -> Text
1226−stripStart = dropWhile isSpace
1227−{-# INLINE [1] stripStart #-}
1228−
1229−-- | /O(n)/ Remove trailing white space from a string. Equivalent to:
737+-- dropping characters that fail the predicate @p@ from both the
738+-- beginning and end of @t@. This function is subject to array
739+-- fusion.
1230740--
1231−-- > dropWhileEnd isSpace
1232−stripEnd :: Text -> Text
1233−stripEnd = dropWhileEnd isSpace
1234−{-# INLINE [1] stripEnd #-}
1235−
1236−-- | /O(n)/ Remove leading and trailing white space from a string.
1237−-- Equivalent to:
741+-- For example, to drop leading and trailing white space:
1238742--
1239743-- > dropAround isSpace
1240−strip :: Text -> Text
1241−strip = dropAround isSpace
1242−{-# INLINE [1] strip #-}
744+dropAround :: (Char -> Bool) -> Text -> Text
745+dropAround p = dropWhile p . dropAfter p
746+{-# INLINE [1] dropAround #-}
1243747
1244748-- | /O(n)/ 'splitAt' @n t@ returns a pair whose first element is a
1245749-- prefix of @t@ of length @n@, and whose second is the remainder of
@@ -1248,20 +752,28 @@ splitAt :: Int -> Text -> (Text, Text)
1248752splitAt n t@(Text arr off len)
1249753 | n <= 0 = (empty, t)
1250754 | n >= len = (t, empty)
1251− | otherwise = let k = iterN n t
1252− in (text arr off k, text arr (off+k) (len-k))
755+ | otherwise = (Text arr off k, Text arr (off+k) (len-k))
756+ where k = loop 0 0
757+ loop !i !cnt
758+ | i >= len || cnt >= n = i
759+ | otherwise = loop (i+d) (cnt+1)
760+ where d = iter_ t i
761+{-# INLINE splitAt #-}
1253762
1254−-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns
1255−-- a pair whose first element is the longest prefix (possibly empty)
1256−-- of @t@ of elements that satisfy @p@, and whose second is the
1257−-- remainder of the list.
763+-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns a
764+-- pair whose first element is the longest prefix (possibly empty) of
765+-- @t@ of elements that satisfy @p@, and whose second is the remainder
766+-- of the list.
1258767span :: (Char -> Bool) -> Text -> (Text, Text)
1259−span p t = case span_ p t of
1260− (# hd,tl #) -> (hd,tl)
768+span p t@(Text arr off len) = (textP arr off k, textP arr (off+k) (len-k))
769+ where k = loop 0
770+ loop !i | i >= len || not (p c) = i
771+ | otherwise = loop (i+d)
772+ where (c,d) = iter t i
1261773{-# INLINE span #-}
1262774
1263−-- | /O(n)/ 'break' is like 'span', but the prefix returned is
1264−-- over elements that fail the predicate @p@.
775+-- | /O(n)/ 'break' is like 'span', but the prefix returned is over
776+-- elements that fail the predicate @p@.
1265777break :: (Char -> Bool) -> Text -> (Text, Text)
1266778break p = span (not . p)
1267779{-# INLINE break #-}
@@ -1273,7 +785,7 @@ groupBy p = loop
1273785 loop t@(Text arr off len)
1274786 | null t = []
1275787 | otherwise = text arr off n : loop (text arr (off+n) (len-n))
1276− where Iter c d = iter t 0
788+ where (c,d) = iter t 0
1277789 n = d + findAIndexOrEnd (not . p c) (Text arr (off+d) (len-d))
1278790
1279791-- | Returns the /array/ index (in units of 'Word16') at which a
@@ -1283,8 +795,8 @@ findAIndexOrEnd :: (Char -> Bool) -> Text -> Int
1283795findAIndexOrEnd q t@(Text _arr _off len) = go 0
1284796 where go !i | i >= len || q c = i
1285797 | otherwise = go (i+d)
1286− where Iter c d = iter t i
1287−
798+ where (c,d) = iter t i
799+
1288800-- | /O(n)/ Group characters in a string by equality.
1289801group :: Text -> [Text]
1290802group = groupBy (==)
@@ -1302,91 +814,120 @@ tails :: Text -> [Text]
1302814tails t | null t = [empty]
1303815 | otherwise = t : tails (unsafeTail t)
1304816
1305−-- $split
817+-- | /O(n)/ Break a 'Text' into pieces separated by the 'Char'
818+-- argument, consuming the delimiter. I.e.
819+--
820+-- > splitChar '\n' "a\nb\nd\ne" == ["a","b","d","e"]
821+-- > splitChar 'a' "aXaXaXa" == ["","X","X","X",""]
822+-- > splitChar 'x' "x" == ["",""]
823+--
824+-- and
1306825--
1307−-- Splitting functions in this library do not perform character-wise
1308−-- copies to create substrings; they just construct new 'Text's that
1309−-- are slices of the original.
826+-- > intercalate (singleton c) . splitChar c == id
827+-- > splitChar == splitWith . (==)
828+--
829+-- As with all splitting functions in this library, this function does
830+-- not copy the substrings, it just constructs new 'Text's that are
831+-- slices of the original.
832+splitChar :: Char -> Text -> [Text]
833+splitChar c = splitWith (==c)
834+{-# INLINE splitChar #-}
835+
836+-- | /O(m)*O(n)/ Break a 'Text' into pieces separated by the 'Text'
837+-- argument, consuming the delimiter. I.e.
838+--
839+-- > split "\r\n" "a\r\nb\r\nd\r\ne" == ["a","b","d","e"]
840+-- > split "aaa" "aaaXaaaXaaaXaaa" == ["","X","X","X",""]
841+-- > split "x" "x" == ["",""]
842+--
843+-- and
844+--
845+-- > intercalate s . split s == id
846+--
847+-- As with all splitting functions in this library, this function does
848+-- not copy the substrings, it just constructs new 'Text's that are
849+-- slices of the original.
850+split :: Text -> Text -> [Text]
851+split pat src0
852+ | l == 0 = [src0]
853+ | otherwise = go src0
854+ where
855+ l = length pat
856+ go src = search 0 src
857+ where
858+ search !n !s
859+ | null s = [src] -- not found
860+ | pat `isPrefixOf` s = take n src : go (drop l s)
861+ | otherwise = search (n+1) (unsafeTail s)
862+{-# INLINE split #-}
1310863
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.
864+-- | /O(m)*O(n)/ Break a 'Text' into pieces at most @k@ times,
865+-- treating the first 'Text' argument as the delimiter to break on,
866+-- and consuming the delimiter. The last element of the list contains
867+-- the remaining text after the number of times to split has been
868+-- reached. A value of zero or less for @k@ causes no splitting to
869+-- occur.
1314870--
1315871-- Examples:
1316872--
1317−-- > splitOn "\r\n" "a\r\nb\r\nd\r\ne" == ["a","b","d","e"]
1318−-- > splitOn "aaa" "aaaXaaaXaaaXaaa" == ["","X","X","X",""]
1319−-- > splitOn "x" "x" == ["",""]
873+-- > splitTimes 0 "//" "a//b//c" == ["a//b//c"]
874+-- > splitTimes 2 ":" "a:b:c:d:e" == ["a","b","c:d:e"]
875+-- > splitTimes 100 "???" "a????b" == ["a","?b"]
1320876--
1321877-- and
1322878--
1323−-- > intercalate s . splitOn s == id
1324−-- > splitOn (singleton c) == split (==c)
1325−--
1326−-- (Note: the string @s@ to split on above cannot be empty.)
1327−--
1328−-- In (unlikely) bad cases, this function's time complexity degrades
1329−-- towards /O(n*m)/.
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]
1336−splitOn pat@(Text _ _ l) src@(Text arr off len)
1337− | l <= 0 = emptyError "splitOn"
1338− | isSingleton pat = split (== unsafeHead pat) src
1339− | otherwise = go 0 (indices pat src)
879+-- > intercalate s . splitTimes k s == id
880+splitTimes :: Int -- ^ Maximum number of times to split
881+ -> Text -- ^ Text to split on
882+ -> Text -- ^ Input text
883+ -> [Text]
884+splitTimes k pat src0
885+ | k <= 0 || l == 0 = [src0]
886+ | otherwise = go k src0
1340887 where
1341− go !s (x:xs) = text arr (s+off) (x-s) : go (x+l) xs
1342− go s _ = [text arr (s+off) (len-s)]
1343−{-# INLINE [1] splitOn #-}
1344−
1345−{-# RULES
1346−"TEXT splitOn/singleton -> split/==" [~1] forall c t.
1347− splitOn (singleton c) t = split (==c) t
1348− #-}
888+ l = length pat
889+ go !i src = search 0 src
890+ where
891+ search !n !s
892+ | i == 0 || null s = [src] -- not found or limit reached
893+ | pat `isPrefixOf` s = take n src : go (i-1) (drop l s)
894+ | otherwise = search (n+1) (unsafeTail s)
895+{-# INLINE splitTimes #-}
1349896
1350897-- | /O(n)/ Splits a 'Text' into components delimited by separators,
1351898-- where the predicate returns True for a separator element. The
1352899-- resulting components do not contain the separators. Two adjacent
1353900-- separators result in an empty component in the output. eg.
1354901--
1355−-- > split (=='a') "aabbaca" == ["","","bb","c",""]
1356−-- > split (=='a') "" == [""]
1357−split :: (Char -> Bool) -> Text -> [Text]
1358−split _ t@(Text _off _arr 0) = [t]
1359−split p t = loop t
902+-- > splitWith (=='a') "aabbaca" == ["","","bb","c",""]
903+-- > splitWith (=='a') [] == []
904+splitWith :: (Char -> Bool) -> Text -> [Text]
905+splitWith _ (Text _off _arr 0) = []
906+splitWith p t = loop t
1360907 where loop s | null s' = [l]
1361908 | otherwise = l : loop (unsafeTail s')
1362− where (# l, s' #) = span_ (not . p) s
1363−{-# INLINE split #-}
1364−
1365−-- | /O(n)/ Splits a 'Text' into components of length @k@. The last
1366−-- element may be shorter than the other chunks, depending on the
1367−-- length of the input. Examples:
1368−--
1369−-- > chunksOf 3 "foobarbaz" == ["foo","bar","baz"]
1370−-- > chunksOf 4 "haskell.org" == ["hask","ell.","org"]
1371−chunksOf :: Int -> Text -> [Text]
1372−chunksOf k = go
1373− where
1374− go t = case splitAt k t of
1375− (a,b) | null a -> []
1376− | otherwise -> a : go b
1377−{-# INLINE chunksOf #-}
909+ where (l, s') = break p s
910+{-# INLINE splitWith #-}
1378911
1379912-- ----------------------------------------------------------------------------
1380913-- * Searching
1381914
915+-------------------------------------------------------------------------------
916+-- ** Searching by equality
917+
918+-- | /O(n)/ 'elem' is the 'Text' membership predicate.
919+elem :: Char -> Text -> Bool
920+elem c t = S.elem c (stream t)
921+{-# INLINE elem #-}
922+
1382923-------------------------------------------------------------------------------
1383924-- ** Searching with a predicate
1384925
1385−-- | /O(n)/ The 'find' function takes a predicate and a 'Text', and
1386−-- returns the first element matching the predicate, or 'Nothing' if
1387−-- there is no such element.
926+-- | /O(n)/ The 'find' function takes a predicate and a 'Text',
927+-- and returns the first element in matching the predicate, or 'Nothing'
928+-- if there is no such element.
1388929find :: (Char -> Bool) -> Text -> Maybe Char
1389−find p t = S.findBy p (stream t)
930+find p t = S.find p (stream t)
1390931{-# INLINE find #-}
1391932
1392933-- | /O(n)/ The 'partition' function takes a predicate and a 'Text',
@@ -1398,103 +939,49 @@ partition :: (Char -> Bool) -> Text -> (Text, Text)
1398939partition p t = (filter p t, filter (not . p) t)
1399940{-# INLINE partition #-}
1400941
1401−-- | /O(n)/ 'filter', applied to a predicate and a 'Text',
1402−-- returns a 'Text' containing those characters that satisfy the
1403−-- predicate.
1404−filter :: (Char -> Bool) -> Text -> Text
1405−filter p t = unstream (S.filter p (stream t))
1406−{-# INLINE filter #-}
1407−
1408−-- | /O(n+m)/ Find the first instance of @needle@ (which must be
1409−-- non-'null') in @haystack@. The first element of the returned tuple
1410−-- is the prefix of @haystack@ before @needle@ is matched. The second
1411−-- is the remainder of @haystack@, starting with the match.
1412−--
1413−-- Examples:
1414−--
1415−-- > breakOn "::" "a::b::c" ==> ("a", "::b::c")
1416−-- > breakOn "/" "foobar" ==> ("foobar", "")
1417−--
1418−-- Laws:
1419−--
1420−-- > append prefix match == haystack
1421−-- > where (prefix, match) = breakOn needle haystack
1422−--
1423−-- If you need to break a string by a substring repeatedly (e.g. you
1424−-- want to break on every instance of a substring), use 'breakOnAll'
1425−-- instead, as it has lower startup overhead.
942+-- | /O(n)/ Break a string on a substring, returning a pair of the
943+-- part of the string prior to the match, and the rest of the string.
1426944--
1427−-- In (unlikely) bad cases, this function's time complexity degrades
1428−-- towards /O(n*m)/.
1429−breakOn :: Text -> Text -> (Text, Text)
1430−breakOn pat src@(Text arr off len)
1431− | null pat = emptyError "breakOn"
1432− | otherwise = case indices pat src of
1433− [] -> (src, empty)
1434− (x:_) -> (text arr off x, text arr (off+x) (len-x))
1435−{-# INLINE breakOn #-}
1436−
1437−-- | /O(n+m)/ Similar to 'breakOn', but searches from the end of the
1438−-- string.
945+-- The following relationship holds:
1439946--
1440−-- The first element of the returned tuple is the prefix of @haystack@
1441−-- up to and including the last match of @needle@. The second is the
1442−-- remainder of @haystack@, following the match.
947+-- > break (==c) l == breakSubstring (singleton c) l
1443948--
1444−-- > breakOnEnd "::" "a::b::c" ==> ("a::b::", "c")
1445−breakOnEnd :: Text -> Text -> (Text, Text)
1446−breakOnEnd pat src = (reverse b, reverse a)
1447− where (a,b) = breakOn (reverse pat) (reverse src)
1448−{-# INLINE breakOnEnd #-}
1449−
1450−-- | /O(n+m)/ Find all non-overlapping instances of @needle@ in
1451−-- @haystack@. Each element of the returned list consists of a pair:
949+-- For example, to tokenise a string, dropping delimiters:
1452950--
1453−-- * The entire string prior to the /k/th match (i.e. the prefix)
951+-- > tokenise x y = h : if null t then [] else tokenise x (drop (length x) t)
952+-- > where (h,t) = breakSubstring x y
1454953--
1455−-- * The /k/th match, followed by the remainder of the string
954+-- To skip to the first occurence of a string:
1456955--
1457−-- Examples:
956+-- > snd (breakSubstring x y)
1458957--
1459−-- > breakOnAll "::" ""
1460−-- > ==> []
1461−-- > breakOnAll "/" "a/b/c/"
1462−-- > ==> [("a", "/b/c/"), ("a/b", "/c/"), ("a/b/c", "/")]
958+-- To take the parts of a string before a delimiter:
1463959--
1464−-- In (unlikely) bad cases, this function's time complexity degrades
1465−-- towards /O(n*m)/.
960+-- > fst (breakSubstring x y)
1466961--
1467−-- The @needle@ parameter may not be empty.
1468−breakOnAll :: Text -- ^ @needle@ to search for
1469− -> Text -- ^ @haystack@ in which to search
1470− -> [(Text, Text)]
1471−breakOnAll pat src@(Text arr off slen)
1472− | null pat = emptyError "breakOnAll"
1473− | otherwise = L.map step (indices pat src)
962+breakSubstring :: Text -- ^ String to search for
963+ -> Text -- ^ String to search in
964+ -> (Text,Text) -- ^ Head and tail of string broken at substring
965+
966+breakSubstring pat src = search 0 src
1474967 where
1475− step x = (chunk 0 x, chunk x (slen-x))
1476− chunk !n !l = text arr (n+off) l
1477−{-# INLINE breakOnAll #-}
968+ search !n !s
969+ | null s = (src,empty) -- not found
970+ | pat `isPrefixOf` s = (take n src,s)
971+ | otherwise = search (n+1) (unsafeTail s)
972+{-# INLINE breakSubstring #-}
973+
974+-- | /O(n)/ 'filter', applied to a predicate and a 'Text',
975+-- returns a 'Text' containing those characters that satisfy the
976+-- predicate.
977+filter :: (Char -> Bool) -> Text -> Text
978+filter p t = unstream (S.filter p (stream t))
979+{-# INLINE filter #-}
980+
1478981
1479982-------------------------------------------------------------------------------
1480983-- ** Indexing 'Text's
1481984
1482−-- $index
1483−--
1484−-- If you think of a 'Text' value as an array of 'Char' values (which
1485−-- it is not), you run the risk of writing inefficient code.
1486−--
1487−-- An idiom that is common in some languages is to find the numeric
1488−-- offset of a character or substring, then use that number to split
1489−-- or trim the searched string. With a 'Text' value, this approach
1490−-- would require two /O(n)/ operations: one to perform the search, and
1491−-- one to operate from wherever the search ended.
1492−--
1493−-- For example, suppose you have a string that you want to split on
1494−-- the substring @\"::\"@, such as @\"foo::bar::quux\"@. Instead of
1495−-- searching for the index of @\"::\"@ and taking the substrings
1496−-- before and after that index, you would instead use @breakOnAll \"::\"@.
1497−
1498985-- | /O(n)/ 'Text' index (subscript) operator, starting from 0.
1499986index :: Text -> Int -> Char
1500987index t n = S.index (stream t) n
@@ -1502,34 +989,39 @@ index t n = S.index (stream t) n
1502989
1503990-- | /O(n)/ The 'findIndex' function takes a predicate and a 'Text'
1504991-- and returns the index of the first element in the 'Text' satisfying
1505−-- the predicate. Subject to fusion.
992+-- the predicate. This function is subject to fusion.
1506993findIndex :: (Char -> Bool) -> Text -> Maybe Int
1507994findIndex p t = S.findIndex p (stream t)
1508995{-# INLINE findIndex #-}
1509996
1510−-- | /O(n+m)/ The 'count' function returns the number of times the
1511−-- query string appears in the given 'Text'. An empty query string is
1512−-- invalid, and will cause an error to be raised.
1513−--
1514−-- In (unlikely) bad cases, this function's time complexity degrades
1515−-- towards /O(n*m)/.
1516−count :: Text -> Text -> Int
1517−count pat src
1518− | null pat = emptyError "count"
1519− | isSingleton pat = countChar (unsafeHead pat) src
1520− | otherwise = L.length (indices pat src)
1521−{-# INLINE [1] count #-}
1522−
1523−{-# RULES
1524−"TEXT count/singleton -> countChar" [~1] forall c t.
1525− count (singleton c) t = countChar c t
1526− #-}
997+-- | The 'findIndices' function extends 'findIndex', by returning the
998+-- indices of all elements satisfying the predicate, in ascending
999+-- order. This function is subject to fusion.
1000+findIndices :: (Char -> Bool) -> Text -> [Int]
1001+findIndices p t = S.findIndices p (stream t)
1002+{-# INLINE findIndices #-}
15271003
1528−-- | /O(n)/ The 'countChar' function returns the number of times the
1529−-- query element appears in the given 'Text'. Subject to fusion.
1530−countChar :: Char -> Text -> Int
1531−countChar c t = S.countChar c (stream t)
1532−{-# INLINE countChar #-}
1004+-- | /O(n)/ The 'elemIndex' function returns the index of the first
1005+-- element in the given 'Text' which is equal to the query element, or
1006+-- 'Nothing' if there is no such element. This function is subject to
1007+-- fusion.
1008+elemIndex :: Char -> Text -> Maybe Int
1009+elemIndex c t = S.elemIndex c (stream t)
1010+{-# INLINE elemIndex #-}
1011+
1012+-- | /O(n)/ The 'elemIndices' function returns the index of every
1013+-- element in the given 'Text' which is equal to the query
1014+-- element. This function is subject to fusion.
1015+elemIndices :: Char -> Text -> [Int]
1016+elemIndices c t = S.elemIndices c (stream t)
1017+{-# INLINE elemIndices #-}
1018+
1019+-- | /O(n)/ The 'count' function returns the number of times the query
1020+-- element appears in the given 'Text'. This function is subject to
1021+-- fusion.
1022+count :: Char -> Text -> Int
1023+count c t = S.count c (stream t)
1024+{-# INLINE count #-}
15331025
15341026-------------------------------------------------------------------------------
15351027-- * Zipping
@@ -1540,15 +1032,13 @@ countChar c t = S.countChar c (stream t)
15401032-- equivalent to a pair of 'unpack' operations.
15411033zip :: Text -> Text -> [(Char,Char)]
15421034zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
1543−{-# INLINE zip #-}
1035+{-# INLINE [0] zip #-}
15441036
15451037-- | /O(n)/ 'zipWith' generalises 'zip' by zipping with the function
15461038-- given as the first argument, instead of a tupling function.
1547−-- Performs replacement on invalid scalar values.
15481039zipWith :: (Char -> Char -> Char) -> Text -> Text -> Text
1549−zipWith f t1 t2 = unstream (S.zipWith g (stream t1) (stream t2))
1550− where g a b = safe (f a b)
1551−{-# INLINE zipWith #-}
1040+zipWith f t1 t2 = unstream (S.zipWith f (stream t1) (stream t2))
1041+{-# INLINE [0] zipWith #-}
15521042
15531043-- | /O(n)/ Breaks a 'Text' up into a list of words, delimited by 'Char's
15541044-- representing white space.
@@ -1564,7 +1054,7 @@ words t@(Text arr off len) = loop 0 0
15641054 then loop (start+1) (start+1)
15651055 else Text arr (start+off) (n-start) : loop (n+d) (n+d)
15661056 | otherwise = loop start (n+d)
1567− where Iter c d = iter t n
1057+ where (c,d) = iter t n
15681058{-# INLINE words #-}
15691059
15701060-- | /O(n)/ Breaks a 'Text' up into a list of 'Text's at
@@ -1574,10 +1064,9 @@ lines ps | null ps = []
15741064 | otherwise = h : if null t
15751065 then []
15761066 else lines (unsafeTail t)
1577− where (# h,t #) = span_ (/= '\n') ps
1067+ where (h,t) = span (/= '\n') ps
15781068{-# INLINE lines #-}
15791069
1580−{-
15811070-- | /O(n)/ Portably breaks a 'Text' up into a list of 'Text's at line
15821071-- boundaries.
15831072--
@@ -1585,6 +1074,7 @@ lines ps | null ps = []
15851074-- return immediately followed by a line feed, or a carriage return.
15861075-- This accounts for both Unix and Windows line ending conventions,
15871076-- and for the old convention used on Mac OS 9 and earlier.
1077+{-
15881078lines' :: Text -> [Text]
15891079lines' ps | null ps = []
15901080 | otherwise = h : case uncons t of
@@ -1611,7 +1101,8 @@ unwords = intercalate (singleton ' ')
16111101{-# INLINE unwords #-}
16121102
16131103-- | /O(n)/ The 'isPrefixOf' function takes two 'Text's and returns
1614−-- 'True' iff the first is a prefix of the second. Subject to fusion.
1104+-- 'True' iff the first is a prefix of the second. This function is
1105+-- subject to fusion.
16151106isPrefixOf :: Text -> Text -> Bool
16161107isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
16171108 alen <= blen && S.isPrefixOf (stream a) (stream b)
@@ -1620,6 +1111,8 @@ isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
16201111{-# RULES
16211112"TEXT isPrefixOf -> fused" [~1] forall s t.
16221113 isPrefixOf s t = S.isPrefixOf (stream s) (stream t)
1114+"TEXT isPrefixOf -> unfused" [1] forall s t.
1115+ S.isPrefixOf (stream s) (stream t) = isPrefixOf s t
16231116 #-}
16241117
16251118-- | /O(n)/ The 'isSuffixOf' function takes two 'Text's and returns
@@ -1632,124 +1125,13 @@ isSuffixOf a@(Text _aarr _aoff alen) b@(Text barr boff blen) =
16321125 | otherwise = Text barr (boff+d) alen
16331126{-# INLINE isSuffixOf #-}
16341127
1635−-- | /O(n+m)/ The 'isInfixOf' function takes two 'Text's and returns
1128+-- | /O(n)/ The 'isInfixOf' function takes two 'Text's and returns
16361129-- 'True' iff the first is contained, wholly and intact, anywhere
16371130-- within the second.
1638−--
1639−-- In (unlikely) bad cases, this function's time complexity degrades
1640−-- towards /O(n*m)/.
16411131isInfixOf :: Text -> Text -> Bool
1642−isInfixOf needle haystack
1643− | null needle = True
1644− | isSingleton needle = S.elem (unsafeHead needle) . S.stream $ haystack
1645− | otherwise = not . L.null . indices needle $ haystack
1646−{-# INLINE [1] isInfixOf #-}
1647−
1648−{-# RULES
1649−"TEXT isInfixOf/singleton -> S.elem/S.stream" [~1] forall n h.
1650− isInfixOf (singleton n) h = S.elem n (S.stream h)
1651− #-}
1652−
1653−-------------------------------------------------------------------------------
1654−-- * View patterns
1655−
1656−-- | /O(n)/ Return the suffix of the second string if its prefix
1657−-- matches the entire first string.
1658−--
1659−-- Examples:
1660−--
1661−-- > stripPrefix "foo" "foobar" == Just "bar"
1662−-- > stripPrefix "" "baz" == Just "baz"
1663−-- > stripPrefix "foo" "quux" == Nothing
1664−--
1665−-- This is particularly useful with the @ViewPatterns@ extension to
1666−-- GHC, as follows:
1667−--
1668−-- > {-# LANGUAGE ViewPatterns #-}
1669−-- > import Data.Text as T
1670−-- >
1671−-- > fnordLength :: Text -> Int
1672−-- > fnordLength (stripPrefix "fnord" -> Just suf) = T.length suf
1673−-- > fnordLength _ = -1
1674−stripPrefix :: Text -> Text -> Maybe Text
1675−stripPrefix p@(Text _arr _off plen) t@(Text arr off len)
1676− | p `isPrefixOf` t = Just $! text arr (off+plen) (len-plen)
1677− | otherwise = Nothing
1678−
1679−-- | /O(n)/ Find the longest non-empty common prefix of two strings
1680−-- and return it, along with the suffixes of each string at which they
1681−-- no longer match.
1682−--
1683−-- If the strings do not have a common prefix or either one is empty,
1684−-- this function returns 'Nothing'.
1685−--
1686−-- Examples:
1687−--
1688−-- > commonPrefixes "foobar" "fooquux" == Just ("foo","bar","quux")
1689−-- > commonPrefixes "veeble" "fetzer" == Nothing
1690−-- > commonPrefixes "" "baz" == Nothing
1691−commonPrefixes :: Text -> Text -> Maybe (Text,Text,Text)
1692−commonPrefixes t0@(Text arr0 off0 len0) t1@(Text arr1 off1 len1) = go 0 0
1693− where
1694− go !i !j | i < len0 && j < len1 && a == b = go (i+d0) (j+d1)
1695− | i > 0 = Just (Text arr0 off0 i,
1696− text arr0 (off0+i) (len0-i),
1697− text arr1 (off1+j) (len1-j))
1698− | otherwise = Nothing
1699− where Iter a d0 = iter t0 i
1700− Iter b d1 = iter t1 j
1701−
1702−-- | /O(n)/ Return the prefix of the second string if its suffix
1703−-- matches the entire first string.
1704−--
1705−-- Examples:
1706−--
1707−-- > stripSuffix "bar" "foobar" == Just "foo"
1708−-- > stripSuffix "" "baz" == Just "baz"
1709−-- > stripSuffix "foo" "quux" == Nothing
1710−--
1711−-- This is particularly useful with the @ViewPatterns@ extension to
1712−-- GHC, as follows:
1713−--
1714−-- > {-# LANGUAGE ViewPatterns #-}
1715−-- > import Data.Text as T
1716−-- >
1717−-- > quuxLength :: Text -> Int
1718−-- > quuxLength (stripSuffix "quux" -> Just pre) = T.length pre
1719−-- > quuxLength _ = -1
1720−stripSuffix :: Text -> Text -> Maybe Text
1721−stripSuffix p@(Text _arr _off plen) t@(Text arr off len)
1722− | p `isSuffixOf` t = Just $! text arr off (len-plen)
1723− | otherwise = Nothing
1724−
1725−-- | Add a list of non-negative numbers. Errors out on overflow.
1726−sumP :: String -> [Int] -> Int
1727−sumP fun = go 0
1728− where go !a (x:xs)
1729− | ax >= 0 = go ax xs
1730− | otherwise = overflowError fun
1731− where ax = a + x
1732− go a _ = a
1132+isInfixOf needle haystack = L.any (isPrefixOf needle) (tails haystack)
1133+{-# INLINE isInfixOf #-}
1134+-- TODO: a better implementation
17331135
17341136emptyError :: String -> a
1735−emptyError fun = P.error $ "Data.Text." ++ fun ++ ": empty input"
1736−
1737−overflowError :: String -> a
1738−overflowError fun = P.error $ "Data.Text." ++ fun ++ ": size overflow"
1739−
1740−-- | /O(n)/ Make a distinct copy of the given string, sharing no
1741−-- storage with the original string.
1742−--
1743−-- As an example, suppose you read a large string, of which you need
1744−-- only a small portion. If you do not use 'copy', the entire original
1745−-- array will be kept alive in memory by the smaller string. Making a
1746−-- copy \"breaks the link\" to the original array, allowing it to be
1747−-- garbage collected if there are no other live references to it.
1748−copy :: Text -> Text
1749−copy (Text arr off len) = Text (A.run go) 0 len
1750− where
1751− go :: ST s (A.MArray s)
1752− go = do
1753− marr <- A.new len
1754− A.copyI marr 0 arr off len
1755− return marr
1137+emptyError fun = P.error ("Data.Text." ++ fun ++ ": empty input")