@@ -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 : rtharper@aftereternity.co.uk, bos@serpentine.com,
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
@@ -74,19 +54,6 @@ module Data.Text
7454 , intersperse
7555 , transpose
7656 , reverse
77− , replace
78−
79− -- ** Case conversion
80− -- $case
81− , toCaseFold
82− , toLower
83− , toUpper
84− , toTitle
85−
86− -- ** Justification
87− , justifyLeft
88− , justifyRight
89− , center
9057
9158 -- * Folds
9259 , foldl
@@ -109,12 +76,12 @@ module Data.Text
10976 -- ** Scans
11077 , scanl
11178 , scanl1
112− , scanr
113− , scanr1
79+ -- , scanr
80+ -- , scanr1
11481
11582 -- ** Accumulating maps
11683 , mapAccumL
117− , mapAccumR
84+ -- , mapAccumR
11885
11986 -- ** Generation and unfolding
12087 , replicate
@@ -125,36 +92,24 @@ module Data.Text
12592
12693 -- ** Breaking strings
12794 , take
128− , takeEnd
12995 , drop
130− , dropEnd
13196 , takeWhile
132− , takeWhileEnd
13397 , dropWhile
134− , dropWhileEnd
135− , dropAround
136− , strip
137− , stripStart
138− , stripEnd
13998 , splitAt
140− , breakOn
141− , breakOnEnd
142− , break
14399 , span
100+ , break
144101 , group
145102 , groupBy
146103 , inits
147104 , tails
148105
149106 -- ** Breaking into many substrings
150− -- $split
151− , splitOn
152107 , split
153− , chunksOf
108+ , splitWith
109+ , breakSubstring
154110
155111 -- ** Breaking into lines and words
156112 , lines
157− --, lines'
158113 , words
159114 , unlines
160115 , unwords
@@ -164,309 +119,146 @@ module Data.Text
164119 , isSuffixOf
165120 , isInfixOf
166121
167− -- ** View patterns
168− , stripPrefix
169− , stripSuffix
170− , commonPrefixes
171−
172122 -- * Searching
123+ , elem
173124 , filter
174− , breakOnAll
175125 , find
176126 , partition
177127
178128 -- , findSubstring
179−
129+
180130 -- * Indexing
181− -- $index
182131 , index
183132 , findIndex
184− , count
133+ , findIndices
134+ , elemIndex
135+ -- , elemIndices
136+ -- , count
185137
186− -- * Zipping
187− , zip
138+ -- * Zipping and unzipping
188139 , zipWith
189140
190− -- -* Ordered text
191− -- , sort
192−
193− -- * Low level operations
194− , copy
195− , unpackCString#
141+ -- * Ordered ByteStrings
142+ , -- sort
196143 ) where
197144
198−import Prelude (Char, Bool(..), Int, Maybe(..), String,
199− Eq(..), Ord(..), Ordering(..), (++),
200− Read(..),
201− (&&), (||), (+), (-), (.), ($), ($!), (>>),
202− not, return, otherwise, quot)
203−#if defined(HAVE_DEEPSEQ)
204−import Control.DeepSeq (NFData(rnf))
205−#endif
206−#if defined(ASSERTS)
145+import Prelude (Char, Bool(..), Functor(..), Int, Maybe(..), String,
146+ Eq, (==), (++), error,
147+ Read(..), Show(..),
148+ (&&), (||), (+), (-), (<), (>), (<=), (>=), (.), ($),
149+ not, return, otherwise)
207150import Control.Exception (assert)
208−#endif
209151import 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)
213152import Control.Monad.ST (ST)
214153import qualified Data.Text.Array as A
215154import qualified Data.List as L
216−import Data.Binary (Binary(get, put))
217155import Data.Monoid (Monoid(..))
218−#if MIN_VERSION_base(4,9,0)
219−import Data.Semigroup (Semigroup(..))
220−#endif
156+import Data.Word (Word16)
221157import 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.
269158
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 >.)
159+import qualified Data.Text.Fusion as S
160+import Data.Text.Fusion (Stream(..), Step(..), stream, unstream)
161+import Data.Text.Internal (Text(..), empty, text)
162+import qualified Prelude as P
163+import Data.Text.Unsafe (iter, iter_)
164+import Data.Text.UnsafeChar (unsafeChr)
165+import qualified Data.Text.Utf16 as U16
298166
299167-- $fusion
300168--
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\".
169+-- Most of the functions in this module are subject to /array fusion/,
170+-- meaning that a pipeline of functions will usually allocate at most
171+-- one 'Text' value.
323172
324173instance 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
328− {-# INLINE (==) #-}
174+ t1 == t2 = (stream t1) `S.eq` (stream t2)
329175
330−instance Ord Text where
331− compare = compareText
176+instance Show Text where
177+ showsPrec p ps r = showsPrec p (unpack ps) r
332178
333179instance Read Text where
334180 readsPrec p str = [(pack x,y) | (x,y) <- readsPrec p str]
335181
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−
344182instance Monoid Text where
345183 mempty = empty
346−#if MIN_VERSION_base(4,9,0)
347− mappend = (<>) -- future-proof definition
348−#else
349184 mappend = append
350−#endif
351185 mconcat = concat
352186
353187instance IsString Text where
354188 fromString = pack
355189
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−
424190-- -----------------------------------------------------------------------------
425191-- * Conversion to/from 'Text'
426192
427−-- | /O(n)/ Convert a 'String' into a 'Text'. Subject to
428−-- fusion. Performs replacement on invalid scalar values.
193+-- | /O(n)/ Convert a 'String' into a 'Text'.
194+--
195+-- This function is subject to array fusion.
429196pack :: String -> Text
430−pack = unstream . S.map safe . S.streamList
197+pack str = (unstream (stream_list str))
198+ where
199+ stream_list s0 = S.Stream next s0 (P.length s0) -- total guess
200+ where
201+ next [] = S.Done
202+ next (x:xs) = S.Yield x xs
431203{-# INLINE [1] pack #-}
204+-- TODO: Has to do validation! -- No, it doesn't, the
205+
206+-- | /O(n)/ Convert a Text into a String.
207+-- Subject to array fusion.
208+unpack :: Text -> String
209+unpack txt = (unstream_list (stream txt))
210+ where
211+ unstream_list (S.Stream next s0 _len) = unfold s0
212+ where
213+ unfold !s = case next s of
214+ S.Done -> []
215+ S.Skip s' -> unfold s'
216+ S.Yield x s' -> x : unfold s'
217+{-# INLINE [1] unpack #-}
218+
219+-- | /O(1)/ Convert a character into a Text.
220+-- Subject to array fusion.
221+singleton :: Char -> Text
222+singleton c = unstream (Stream next (c:[]) 1)
223+ where
224+ {-# INLINE next #-}
225+ next (k:ks) = Yield k ks
226+ next [] = Done
227+{-# INLINE [1] singleton #-}
432228
433229-- -----------------------------------------------------------------------------
434230-- * Basic functions
435231
436232-- | /O(n)/ Adds a character to the front of a 'Text'. This function
437233-- 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.
234+-- copying a new array. Subject to array fusion.
440235cons :: Char -> Text -> Text
441−cons c t = unstream (S.cons (safe c) (stream t))
236+cons c t = unstream (S.cons c (stream t))
442237{-# INLINE cons #-}
443238
444−infixr 5 `cons`
445−
446239-- | /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.
240+-- entire array in the process. Subject to array fusion.
449241snoc :: Text -> Char -> Text
450−snoc t c = unstream (S.snoc (stream t) (safe c))
242+snoc t c = unstream (S.snoc (stream t) c)
451243{-# INLINE snoc #-}
452244
453245-- | /O(n)/ Appends one 'Text' to the other by copying both of them
454−-- into a new 'Text'. Subject to fusion.
246+-- into a new 'Text'. Subject to array fusion.
455247append :: 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"
248+append (Text arr1 off1 len1) (Text arr2 off2 len2) = Text (A.run x) 0 len
461249 where
462250 len = len1+len2
463− x :: ST s (A.MArray s)
464251 x = do
465− arr <- A.new len
466− A.copyI arr 0 arr1 off1 len1
467− A.copyI arr len1 arr2 off2 len
252+ arr <- A.unsafeNew len :: ST s (A.MArray s Word16)
253+ copy arr1 off1 (len1+off1) arr 0
254+ copy arr2 off2 (len2+off2) arr len1
468255 return arr
469−{-# NOINLINE append #-}
256+ where
257+ copy arr i max arr' j
258+ | i >= max = return ()
259+ | otherwise = do A.unsafeWrite arr' j (arr `A.unsafeIndex` i)
260+ copy arr (i+1) max arr' (j+1)
261+{-# INLINE append #-}
470262
471263{-# RULES
472264"TEXT append -> fused" [~1] forall t1 t2.
@@ -476,29 +268,35 @@ append a@(Text arr1 off1 len1) b@(Text arr2 off2 len2)
476268 #-}
477269
478270-- | /O(1)/ Returns the first character of a 'Text', which must be
479−-- non-empty. Subject to fusion.
271+-- non-empty. Subject to array fusion.
480272head :: Text -> Char
481273head t = S.head (stream t)
482274{-# INLINE head #-}
483275
484276-- | /O(1)/ Returns the first character and rest of a 'Text', or
485−-- 'Nothing' if empty. Subject to fusion.
277+-- 'Nothing' if empty. Subject to array fusion.
486278uncons :: Text -> Maybe (Char, Text)
487279uncons t@(Text arr off len)
488280 | 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 #-}
281+ | otherwise = Just (c, Text arr (off+d) (len-d))
282+ where (c,d) = iter t 0
283+{-# INLINE uncons #-}
492284
493−-- | Lifted from Control.Arrow and specialized.
494285second :: (b -> c) -> (a,b) -> (a,c)
495286second f (a, b) = (a, f b)
496287
288+{-# RULES
289+"TEXT uncons -> fused" [~1] forall t.
290+ uncons t = fmap (second unstream) (S.uncons (stream t))
291+"TEXT uncons -> unfused" [1] forall t.
292+ fmap (second unstream) (S.uncons (stream t)) = uncons t
293+ #-}
294+
497295-- | /O(1)/ Returns the last character of a 'Text', which must be
498−-- non-empty. Subject to fusion.
296+-- non-empty. Subject to array fusion.
499297last :: Text -> Char
500298last (Text arr off len)
501− | len <= 0 = emptyError "last"
299+ | len <= 0 = errorEmptyList "last"
502300 | n < 0xDC00 || n > 0xDFFF = unsafeChr n
503301 | otherwise = U16.chr2 n0 n
504302 where n = A.unsafeIndex arr (off+len-1)
@@ -512,12 +310,19 @@ last (Text arr off len)
512310 S.last (stream t) = last t
513311 #-}
514312
313+-- | Construct a 'Text' without invisibly pinning its byte array in
314+-- memory if its length has dwindled to zero.
315+textP :: A.Array Word16 -> Int -> Int -> Text
316+textP arr off len | len == 0 = empty
317+ | otherwise = text arr off len
318+{-# INLINE textP #-}
319+
515320-- | /O(1)/ Returns all characters after the head of a 'Text', which
516−-- must be non-empty. Subject to fusion.
321+-- must be non-empty. Subject to array fusion.
517322tail :: Text -> Text
518323tail t@(Text arr off len)
519− | len <= 0 = emptyError "tail"
520− | otherwise = text arr (off+d) (len-d)
324+ | len <= 0 = errorEmptyList "tail"
325+ | otherwise = textP arr (off+d) (len-d)
521326 where d = iter_ t 0
522327{-# INLINE [1] tail #-}
523328
@@ -529,11 +334,11 @@ tail t@(Text arr off len)
529334 #-}
530335
531336-- | /O(1)/ Returns all but the last character of a 'Text', which must
532−-- be non-empty. Subject to fusion.
337+-- be non-empty. Subject to array fusion.
533338init :: Text -> Text
534−init (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)
339+init (Text arr off len) | len <= 0 = errorEmptyList "init"
340+ | n >= 0xDC00 && n <= 0xDFFF = textP arr off (len-2)
341+ | otherwise = textP arr off (len-1)
537342 where
538343 n = A.unsafeIndex arr (off+len-1)
539344{-# INLINE [1] init #-}
@@ -545,14 +350,10 @@ init (Text arr off len) | len <= 0 = emptyError "init"
545350 unstream (S.init (stream t)) = init t
546351 #-}
547352
548−-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to
353+-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to array
549354-- fusion.
550355null :: Text -> Bool
551−null (Text _arr _off len) =
552−#if defined(ASSERTS)
553− assert (len >= 0) $
554−#endif
555− len <= 0
356+null (Text _arr _off len) = assert (len >= 0) $ len <= 0
556357{-# INLINE [1] null #-}
557358
558359{-# RULES
@@ -562,287 +363,41 @@ null (Text _arr _off len) =
562363 S.null (stream t) = null t
563364 #-}
564365
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−
571366-- | /O(n)/ Returns the number of characters in a 'Text'.
572−-- Subject to fusion.
367+-- Subject to array fusion.
573368length :: Text -> Int
574369length 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− #-}
370+{-# INLINE length #-}
623371
624372-- -----------------------------------------------------------------------------
625373-- * Transformations
626−-- | /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.
374+-- | /O(n)/ 'map' @f @xs is the 'Text' obtained by applying @f@ to
375+-- each element of @xs@. Subject to array fusion.
629376map :: (Char -> Char) -> Text -> Text
630−map f t = unstream (S.map (safe . f) (stream t))
377+map f t = unstream (S.map f (stream t))
631378{-# INLINE [1] map #-}
632379
633380-- | /O(n)/ The 'intercalate' function takes a 'Text' and a list of
634381-- 'Text's and concatenates the list after interspersing the first
635382-- argument between each element of the list.
636383intercalate :: Text -> [Text] -> Text
637−intercalate t = concat . (F.intersperse t)
384+intercalate t ts = unstream (S.intercalate (stream t) (L.map stream ts))
638385{-# INLINE intercalate #-}
639386
640387-- | /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.
388+-- between the characters of a 'Text'. Subject to array fusion.
643389intersperse :: Char -> Text -> Text
644−intersperse c t = unstream (S.intersperse (safe c) (stream t))
390+intersperse c t = unstream (S.intersperse c (stream t))
645391{-# INLINE intersperse #-}
646392
647−-- | /O(n)/ Reverse the characters of a string. Subject to fusion.
393+-- | /O(n)/ Reverse the characters of a string. Subject to array fusion.
648394reverse :: Text -> Text
649395reverse t = S.reverse (stream t)
650396{-# INLINE reverse #-}
651397
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.
680− -> 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
705−
706−-- ----------------------------------------------------------------------------
707−-- ** Case conversions (folds)
708−
709−-- $case
710−--
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 >.
725−
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.
730−--
731−-- A string @x@ is a caseless match for a string @y@ if and only if:
732−--
733−-- @toCaseFold x == toCaseFold y@
734−--
735−-- The result string may be longer than the input string, and may
736−-- 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.
742−toCaseFold :: Text -> Text
743−toCaseFold t = unstream (S.toCaseFold (stream t))
744−{-# INLINE toCaseFold #-}
745−
746−-- | /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).
753−toLower :: Text -> Text
754−toLower t = unstream (S.toLower (stream t))
755−{-# INLINE toLower #-}
756−
757−-- | /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\".
763−toUpper :: Text -> Text
764−toUpper t = unstream (S.toUpper (stream t))
765−{-# INLINE toUpper #-}
766−
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−
843398-- | /O(n)/ The 'transpose' function transposes the rows and columns
844399-- of its 'Text' argument. Note that this function uses 'pack',
845−-- 'unpack', and the list version of transpose, and is thus not very
400+-- 'unpack', and the 'List' version of transpose and is thus not very
846401-- efficient.
847402transpose :: [Text] -> [Text]
848403transpose ts = P.map pack (L.transpose (P.map unpack ts))
@@ -850,40 +405,43 @@ transpose ts = P.map pack (L.transpose (P.map unpack ts))
850405-- -----------------------------------------------------------------------------
851406-- * Reducing 'Text's (folds)
852407
853−-- | /O(n)/ 'foldl', applied to a binary operator, a starting value
408+-- | 'foldl', applied to a binary operator, a starting value
854409-- (typically the left-identity of the operator), and a 'Text',
855410-- reduces the 'Text' using the binary operator, from left to right.
856−-- Subject to fusion.
857−foldl :: (a -> Char -> a) -> a -> Text -> a
411+-- Subject to array fusion.
412+foldl :: (b -> Char -> b) -> b -> Text -> b
858413foldl f z t = S.foldl f z (stream t)
859414{-# INLINE foldl #-}
860415
861−-- | /O(n)/ A strict version of 'foldl'. Subject to fusion.
862−foldl' :: (a -> Char -> a) -> a -> Text -> a
416+-- | A strict version of 'foldl'.
417+-- Subject to array fusion.
418+foldl' :: (b -> Char -> b) -> b -> Text -> b
863419foldl' f z t = S.foldl' f z (stream t)
864420{-# INLINE foldl' #-}
865421
866−-- | /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.
422+-- | A variant of 'foldl' that has no starting value argument, and
423+-- thus must be applied to a non-empty 'Text'. Subject to array
424+-- fusion.
868425foldl1 :: (Char -> Char -> Char) -> Text -> Char
869426foldl1 f t = S.foldl1 f (stream t)
870427{-# INLINE foldl1 #-}
871428
872−-- | /O(n)/ A strict version of 'foldl1'. Subject to fusion.
429+-- | A strict version of 'foldl1'.
430+-- Subject to array fusion.
873431foldl1' :: (Char -> Char -> Char) -> Text -> Char
874432foldl1' f t = S.foldl1' f (stream t)
875433{-# INLINE foldl1' #-}
876434
877−-- | /O(n)/ 'foldr', applied to a binary operator, a starting value
435+-- | 'foldr', applied to a binary operator, a starting value
878436-- (typically the right-identity of the operator), and a 'Text',
879437-- reduces the 'Text' using the binary operator, from right to left.
880−-- Subject to fusion.
881−foldr :: (Char -> a -> a) -> a -> Text -> a
438+-- Subject to array fusion.
439+foldr :: (Char -> b -> b) -> b -> Text -> b
882440foldr f z t = S.foldr f z (stream t)
883441{-# INLINE foldr #-}
884442
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
443+-- | A variant of 'foldr' that has no starting value argument, and
444+-- thust must be applied to a non-empty 'Text'. Subject to array
887445-- fusion.
888446foldr1 :: (Char -> Char -> Char) -> Text -> Char
889447foldr1 f t = S.foldr1 f (stream t)
@@ -892,48 +450,40 @@ foldr1 f t = S.foldr1 f (stream t)
892450-- -----------------------------------------------------------------------------
893451-- ** Special folds
894452
895−-- | /O(n)/ Concatenate a list of 'Text's.
453+-- | /O(n)/ Concatenate a list of 'Text's. Subject to array fusion.
896454concat :: [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
910−
911−-- | /O(n)/ Map a function over a 'Text' that results in a 'Text', and
912−-- concatenate the results.
455+concat ts = unstream (S.concat (L.map stream ts))
456+{-# INLINE concat #-}
457+
458+-- | Map a function over a 'Text' that results in a 'Text', and concatenate the
459+-- results. This function is subject to array fusion.
460+--
461+-- Note: if in 'concatMap' @f @xs, @f@ is defined in terms of fusible
462+-- functions, it will also be fusible.
913463concatMap :: (Char -> Text) -> Text -> Text
914−concatMap f = concat . foldr ((:) . f) []
464+concatMap f t = unstream (S.concatMap (stream . f) (stream t))
915465{-# INLINE concatMap #-}
916466
917−-- | /O(n)/ 'any' @p@ @t@ determines whether any character in the
918−-- 'Text' @t@ satisfies the predicate @p@. Subject to fusion.
467+-- | 'any' @p @xs determines whether any character in the 'Text' @xs@
468+-- satisifes the predicate @p@. Subject to array fusion.
919469any :: (Char -> Bool) -> Text -> Bool
920470any p t = S.any p (stream t)
921471{-# INLINE any #-}
922472
923−-- | /O(n)/ 'all' @p@ @t@ determines whether all characters in the
924−-- 'Text' @t@ satisfy the predicate @p@. Subject to fusion.
473+-- | 'all' @p @xs determines whether all characters in the 'Text' @xs@
474+-- satisify the predicate @p@. Subject to array fusion.
925475all :: (Char -> Bool) -> Text -> Bool
926476all p t = S.all p (stream t)
927477{-# INLINE all #-}
928478
929479-- | /O(n)/ 'maximum' returns the maximum value from a 'Text', which
930−-- must be non-empty. Subject to fusion.
480+-- must be non-empty. Subject to array fusion.
931481maximum :: Text -> Char
932482maximum t = S.maximum (stream t)
933483{-# INLINE maximum #-}
934484
935485-- | /O(n)/ 'minimum' returns the minimum value from a 'Text', which
936−-- must be non-empty. Subject to fusion.
486+-- must be non-empty. Subject to array fusion.
937487minimum :: Text -> Char
938488minimum t = S.minimum (stream t)
939489{-# INLINE minimum #-}
@@ -942,8 +492,8 @@ minimum t = S.minimum (stream t)
942492-- * Building 'Text's
943493
944494-- | /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.
495+-- successive reduced values from the left. This function is subject
496+-- to array fusion.
947497--
948498-- > scanl f z [x1, x2, ...] == [z, z `f` x1, (z `f` x1) `f` x2, ...]
949499--
@@ -951,131 +501,72 @@ minimum t = S.minimum (stream t)
951501--
952502-- > last (scanl f z xs) == foldl f z xs.
953503scanl :: (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)
504+scanl f z t = unstream (S.scanl f z (stream t))
956505{-# INLINE scanl #-}
957506
958507-- | /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.
508+-- value argument. This function is subject to array fusion.
961509--
962510-- > scanl1 f [x1, x2, ...] == [x1, x1 `f` x2, ...]
963511scanl1 :: (Char -> Char -> Char) -> Text -> Text
964512scanl1 f t | null t = empty
965− | otherwise = scanl f (unsafeHead t) (unsafeTail t)
513+ | otherwise = scanl f (head t) (tail t)
966514{-# INLINE scanl1 #-}
967515
968−-- | /O(n)/ 'scanr' is the right-to-left dual of 'scanl'. Performs
969−-- replacement on invalid scalar values.
970−--
971−-- > scanr f v == reverse . scanl (flip f) v . reverse
972−scanr :: (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)
975−{-# INLINE scanr #-}
976−
977−-- | /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.
980−scanr1 :: (Char -> Char -> Char) -> Text -> Text
981−scanr1 f t | null t = empty
982− | otherwise = scanr f (last t) (init t)
983−{-# INLINE scanr1 #-}
984−
985−-- | /O(n)/ Like a combination of 'map' and 'foldl''. Applies a
516+-- | /O(n)/ Like a combination of 'map' and 'foldl'. Applies a
986517-- 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.
989−mapAccumL :: (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)
518+-- parameter from left to right, and returns a final 'Text'.
519+--
520+-- /Note/: Unlike the version over lists, this function does not return a
521+-- final value for the accumulator.
522+mapAccumL :: (a -> Char -> (a,Char)) -> a -> Text -> Text
523+mapAccumL f z t = unstream (S.mapAccumL f z (stream t))
992524{-# INLINE mapAccumL #-}
993525
994−-- | 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.
1000−mapAccumR :: (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 #-}
1004−
1005526-- -----------------------------------------------------------------------------
1006527-- ** Generating and unfolding 'Text's
1007528
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
1033−-- 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 #-}
529+replicate :: Int -> Char -> Text
530+replicate n c = unstream (S.replicate n c)
531+{-# INLINE replicate #-}
1037532
1038533-- | /O(n)/, where @n@ is the length of the result. The 'unfoldr'
1039534-- function is analogous to the List 'L.unfoldr'. 'unfoldr' builds a
1040535-- 'Text' from a seed value. The function takes the element and
1041536-- returns 'Nothing' if it is done producing the 'Text', otherwise
1042537-- 'Just' @(a,b)@. In this case, @a@ is the next 'Char' in the
1043−-- string, and @b@ is the seed value for further production. Subject
1044−-- to fusion. Performs replacement on invalid scalar values.
538+-- string, and @b@ is the seed value for further production.
1045539unfoldr :: (a -> Maybe (Char,a)) -> a -> Text
1046−unfoldr f s = unstream (S.unfoldr (firstf safe . f) s)
540+unfoldr f s = unstream (S.unfoldr f s)
1047541{-# INLINE unfoldr #-}
1048542
1049543-- | /O(n)/ Like 'unfoldr', 'unfoldrN' builds a 'Text' from a seed
1050544-- value. However, the length of the result should be limited by the
1051545-- first argument to 'unfoldrN'. This function is more efficient than
1052546-- 'unfoldr' when the maximum length of the result is known and
1053−-- correct, otherwise its performance is similar to 'unfoldr'. Subject
1054−-- to fusion. Performs replacement on invalid scalar values.
547+-- correct, otherwise its performance is similar to 'unfoldr'.
1055548unfoldrN :: Int -> (a -> Maybe (Char,a)) -> a -> Text
1056−unfoldrN n f s = unstream (S.unfoldrN n (firstf safe . f) s)
549+unfoldrN n f s = unstream (S.unfoldrN n f s)
1057550{-# INLINE unfoldrN #-}
1058551
1059552-- -----------------------------------------------------------------------------
1060553-- * Substrings
1061554
1062−-- | /O(n)/ 'take' @n@, applied to a 'Text', returns the prefix of the
555+-- /O(n) 'take' @n@, applied to a 'Text', returns the prefix of the
1063556-- 'Text' of length @n@, or the 'Text' itself if @n@ is greater than
1064−-- the length of the Text. Subject to fusion.
557+-- the length of the Text.
1065558take :: Int -> Text -> Text
1066559take n t@(Text arr off len)
1067560 | n <= 0 = empty
1068561 | n >= len = t
1069− | otherwise = text arr off (iterN n t)
562+ | otherwise = Text arr off (loop 0 0)
563+ where
564+ loop !i !count
565+ | i >= len || count >= n = i
566+ | otherwise = loop (i+d) (count+1)
567+ where d = iter_ t i
1070568{-# INLINE [1] take #-}
1071569
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−
1079570{-# RULES
1080571"TEXT take -> fused" [~1] forall n t.
1081572 take n t = unstream (S.take n (stream t))
@@ -1083,36 +574,19 @@ iterN n t@(Text _arr _off len) = loop 0 0
1083574 unstream (S.take n (stream t)) = take n t
1084575 #-}
1085576
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−
1107577-- | /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.
578+-- 'Text' of length @n@, or the empty 'Text' if @n@ is greater than the
579+-- length of the 'Text'.
1110580drop :: Int -> Text -> Text
1111581drop n t@(Text arr off len)
1112582 | n <= 0 = t
1113583 | n >= len = empty
1114− | otherwise = text arr (off+i) (len-i)
1115− where i = iterN n t
584+ | otherwise = loop 0 0
585+ where end = off + len
586+ loop !i !count
587+ | i >= end || count >= n = Text arr (off+i) (len-i)
588+ | otherwise = loop (i+d) (count+1)
589+ where d = iter_ t i
1116590{-# INLINE [1] drop #-}
1117591
1118592{-# RULES
@@ -1122,27 +596,15 @@ drop n t@(Text arr off len)
1122596 unstream (S.drop n (stream t)) = drop n t
1123597 #-}
1124598
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.
599+-- | /O(n)/ 'takeWhile', applied to a predicate @p@ and a 'Text', returns
600+-- the longest prefix (possibly empty) of elements that satisfy @p@.
601+-- This function is subject to array fusion.
1140602takeWhile :: (Char -> Bool) -> Text -> Text
1141603takeWhile p t@(Text arr off len) = loop 0
1142604 where loop !i | i >= len = t
1143605 | p c = loop (i+d)
1144− | otherwise = text arr off i
1145− where Iter c d = iter t i
606+ | otherwise = textP arr off i
607+ where (c,d) = iter t i
1146608{-# INLINE [1] takeWhile #-}
1147609
1148610{-# RULES
@@ -1152,35 +614,14 @@ takeWhile p t@(Text arr off len) = loop 0
1152614 unstream (S.takeWhile p (stream t)) = takeWhile p t
1153615 #-}
1154616
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−
1176−-- | /O(n)/ 'dropWhile' @p@ @t@ returns the suffix remaining after
1177−-- 'takeWhile' @p@ @t@. Subject to fusion.
617+-- | /O(n)/ 'dropWhile' @p@ @xs@ returns the suffix remaining after
618+-- 'takeWhile' @p@ @xs@. This function is subject to array fusion.
1178619dropWhile :: (Char -> Bool) -> Text -> Text
1179620dropWhile p t@(Text arr off len) = loop 0 0
1180621 where loop !i !l | l >= len = empty
1181622 | p c = loop (i+d) (l+d)
1182623 | otherwise = Text arr (off+i) (len-l)
1183− where Iter c d = iter t i
624+ where (c,d) = iter t i
1184625{-# INLINE [1] dropWhile #-}
1185626
1186627{-# RULES
@@ -1190,57 +631,6 @@ dropWhile p t@(Text arr off len) = loop 0 0
1190631 unstream (S.dropWhile p (stream t)) = dropWhile p t
1191632 #-}
1192633
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.
1196−--
1197−-- Examples:
1198−--
1199−-- > dropWhileEnd (=='.') "foo..." == "foo"
1200−dropWhileEnd :: (Char -> Bool) -> Text -> Text
1201−dropWhileEnd p t@(Text arr off len) = loop (len-1) len
1202− where loop !i !l | l <= 0 = empty
1203− | p c = loop (i+d) (l+d)
1204− | otherwise = Text arr off l
1205− where (c,d) = reverseIter t i
1206−{-# INLINE [1] dropWhileEnd #-}
1207−
1208−{-# 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
1213− #-}
1214−
1215−-- | /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:
1230−--
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:
1238−--
1239−-- > dropAround isSpace
1240−strip :: Text -> Text
1241−strip = dropAround isSpace
1242−{-# INLINE [1] strip #-}
1243−
1244634-- | /O(n)/ 'splitAt' @n t@ returns a pair whose first element is a
1245635-- prefix of @t@ of length @n@, and whose second is the remainder of
1246636-- the string. It is equivalent to @('take' n t, 'drop' n t)@.
@@ -1248,20 +638,28 @@ splitAt :: Int -> Text -> (Text, Text)
1248638splitAt n t@(Text arr off len)
1249639 | n <= 0 = (empty, t)
1250640 | n >= len = (t, empty)
1251− | otherwise = let k = iterN n t
1252− in (text arr off k, text arr (off+k) (len-k))
1253−
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.
641+ | otherwise = (Text arr off k, Text arr (off+k) (len-k))
642+ where k = loop 0 0
643+ loop !i !count
644+ | i >= len || count >= n = i
645+ | otherwise = loop (i+d) (count+1)
646+ where d = iter_ t i
647+{-# INLINE splitAt #-}
648+
649+-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns a
650+-- pair whose first element is the longest prefix (possibly empty) of
651+-- @t@ of elements that satisfy @p@, and whose second is the remainder
652+-- of the list.
1258653span :: (Char -> Bool) -> Text -> (Text, Text)
1259−span p t = case span_ p t of
1260− (# hd,tl #) -> (hd,tl)
654+span p t@(Text arr off len) = (textP arr off k, textP arr (off+k) (len-k))
655+ where k = loop 0
656+ loop !i | i >= len || not (p c) = i
657+ | otherwise = loop (i+d)
658+ where (c,d) = iter t i
1261659{-# INLINE span #-}
1262660
1263−-- | /O(n)/ 'break' is like 'span', but the prefix returned is
1264−-- over elements that fail the predicate @p@.
661+-- | /O(n)/ 'break' is like 'span', but the prefix returned is over
662+-- elements that fail the predicate @p@.
1265663break :: (Char -> Bool) -> Text -> (Text, Text)
1266664break p = span (not . p)
1267665{-# INLINE break #-}
@@ -1273,7 +671,7 @@ groupBy p = loop
1273671 loop t@(Text arr off len)
1274672 | null t = []
1275673 | otherwise = text arr off n : loop (text arr (off+n) (len-n))
1276− where Iter c d = iter t 0
674+ where (c,d) = iter t 0
1277675 n = d + findAIndexOrEnd (not . p c) (Text arr (off+d) (len-d))
1278676
1279677-- | Returns the /array/ index (in units of 'Word16') at which a
@@ -1283,8 +681,8 @@ findAIndexOrEnd :: (Char -> Bool) -> Text -> Int
1283681findAIndexOrEnd q t@(Text _arr _off len) = go 0
1284682 where go !i | i >= len || q c = i
1285683 | otherwise = go (i+d)
1286− where Iter c d = iter t i
1287−
684+ where (c,d) = iter t i
685+
1288686-- | /O(n)/ Group characters in a string by equality.
1289687group :: Text -> [Text]
1290688group = groupBy (==)
@@ -1300,93 +698,64 @@ inits t@(Text arr off len) = loop 0
1300698-- first.
1301699tails :: Text -> [Text]
1302700tails t | null t = [empty]
1303− | otherwise = t : tails (unsafeTail t)
701+ | otherwise = t : tails (tail t)
1304702
1305−-- $split
1306−--
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.
1310−
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.
1314−--
1315−-- Examples:
1316−--
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" == ["",""]
703+-- | /O(n)/ Break a 'Text' into pieces separated by the 'Char'
704+-- argument, consuming the delimiter. I.e.
1320705--
706+-- > split '\n' "a\nb\nd\ne" == ["a","b","d","e"]
707+-- > split 'a' "aXaXaXa" == ["","X","X","X",""]
708+-- > split 'x' "x" == ["",""]
709+--
1321710-- and
1322711--
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)
1340− 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− #-}
712+-- > intercalate (singleton c) . split c == id
713+-- > split == splitWith . (==)
714+--
715+-- As for all splitting functions in this library, this function does
716+-- not copy the substrings, it just constructs new 'Text's that are
717+-- slices of the original.
718+split :: Char -> Text -> [Text]
719+split c = splitWith (==c)
720+{-# INLINE split #-}
1349721
1350722-- | /O(n)/ Splits a 'Text' into components delimited by separators,
1351723-- where the predicate returns True for a separator element. The
1352724-- resulting components do not contain the separators. Two adjacent
1353725-- separators result in an empty component in the output. eg.
1354726--
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
1360− where loop s | null s' = [l]
1361− | 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 #-}
727+-- > splitWith (=='a') "aabbaca" == ["","","bb","c",""]
728+-- > splitWith (=='a') [] == []
729+splitWith :: (Char -> Bool) -> Text -> [Text]
730+splitWith p = loop
731+ where loop t@(Text arr off len)
732+ | null t = [empty]
733+ | otherwise = text arr off n : rest
734+ where n = findAIndexOrEnd p t
735+ m = n + iter_ t n
736+ rest | n >= len = []
737+ | otherwise = loop (text arr (off+m) (len-m))
738+{-# INLINE splitWith #-}
1378739
1379740-- ----------------------------------------------------------------------------
1380741-- * Searching
1381742
743+-------------------------------------------------------------------------------
744+-- ** Searching by equality
745+
746+-- | /O(n)/ 'elem' is the 'Text' membership predicate.
747+elem :: Char -> Text -> Bool
748+elem c t = S.elem c (stream t)
749+{-# INLINE elem #-}
750+
1382751-------------------------------------------------------------------------------
1383752-- ** Searching with a predicate
1384753
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.
754+-- | /O(n)/ The 'find' function takes a predicate and a 'Text',
755+-- and returns the first element in matching the predicate, or 'Nothing'
756+-- if there is no such element.
1388757find :: (Char -> Bool) -> Text -> Maybe Char
1389−find p t = S.findBy p (stream t)
758+find p t = S.find p (stream t)
1390759{-# INLINE find #-}
1391760
1392761-- | /O(n)/ The 'partition' function takes a predicate and a 'Text',
@@ -1398,157 +767,90 @@ partition :: (Char -> Bool) -> Text -> (Text, Text)
1398767partition p t = (filter p t, filter (not . p) t)
1399768{-# INLINE partition #-}
1400769
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:
770+-- | Break a string on a substring, returning a pair of the part of the
771+-- string prior to the match, and the rest of the string.
1419772--
1420−-- > append prefix match == haystack
1421−-- > where (prefix, match) = breakOn needle haystack
773+-- The following relationships hold:
1422774--
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.
775+-- > break (== c) l == breakSubstring (singleton c) l
1426776--
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.
777+-- and:
1439778--
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.
779+-- > findSubstring s l ==
780+-- > if null s then Just 0
781+-- > else case breakSubstring s l of
782+-- > (x,y) | null y -> Nothing
783+-- > | otherwise -> Just (length x)
1443784--
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:
785+-- For example, to tokenise a string, dropping delimiters:
1452786--
1453−-- * The entire string prior to the /k/th match (i.e. the prefix)
787+-- > tokenise x y = h : if null t then [] else tokenise x (drop (length x) t)
788+-- > where (h,t) = breakSubstring x y
1454789--
1455−-- * The /k/th match, followed by the remainder of the string
790+-- To skip to the first occurence of a string:
1456791--
1457−-- Examples:
792+-- > snd (breakSubstring x y)
1458793--
1459−-- > breakOnAll "::" ""
1460−-- > ==> []
1461−-- > breakOnAll "/" "a/b/c/"
1462−-- > ==> [("a", "/b/c/"), ("a/b", "/c/"), ("a/b/c", "/")]
794+-- To take the parts of a string before a delimiter:
1463795--
1464−-- In (unlikely) bad cases, this function's time complexity degrades
1465−-- towards /O(n*m)/.
796+-- > fst (breakSubstring x y)
1466797--
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)
798+breakSubstring :: Text -- ^ String to search for
799+ -> Text -- ^ String to search in
800+ -> (Text,Text) -- ^ Head and tail of string broken at substring
801+
802+breakSubstring pat src = search 0 src
1474803 where
1475− step x = (chunk 0 x, chunk x (slen-x))
1476− chunk !n !l = text arr (n+off) l
1477−{-# INLINE breakOnAll #-}
804+ search !n !s
805+ | null s = (src,empty) -- not found
806+ | pat `isPrefixOf` s = (take n src,s)
807+ | otherwise = search (n+1) (tail s)
808+{-# INLINE breakSubstring #-}
809+
810+-- | /O(n)/ 'filter', applied to a predicate and a 'Text',
811+-- returns a 'Text' containing those characters that satisfy the
812+-- predicate.
813+filter :: (Char -> Bool) -> Text -> Text
814+filter p t = unstream (S.filter p (stream t))
815+{-# INLINE filter #-}
816+
1478817
1479818-------------------------------------------------------------------------------
1480819-- ** Indexing 'Text's
1481820
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−
1498−-- | /O(n)/ 'Text' index (subscript) operator, starting from 0.
821+-- | /O(1)/ 'Text' index (subscript) operator, starting from 0.
1499822index :: Text -> Int -> Char
1500823index t n = S.index (stream t) n
1501824{-# INLINE index #-}
1502825
1503826-- | /O(n)/ The 'findIndex' function takes a predicate and a 'Text'
1504827-- and returns the index of the first element in the 'Text' satisfying
1505−-- the predicate. Subject to fusion.
828+-- the predicate. This function is subject to fusion.
1506829findIndex :: (Char -> Bool) -> Text -> Maybe Int
1507830findIndex p t = S.findIndex p (stream t)
1508831{-# INLINE findIndex #-}
1509832
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 #-}
833+-- | The 'findIndices' function extends 'findIndex', by returning the
834+-- indices of all elements satisfying the predicate, in ascending
835+-- order. This function is subject to fusion.
836+findIndices :: (Char -> Bool) -> Text -> [Int]
837+findIndices p t = S.findIndices p (stream t)
838+{-# INLINE findIndices #-}
1522839
1523−{-# RULES
1524−"TEXT count/singleton -> countChar" [~1] forall c t.
1525− count (singleton c) t = countChar c t
1526− #-}
1527−
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 #-}
840+-- | /O(n)/ The 'elemIndex' function returns the index of the first
841+-- element in the given 'Text' which is equal to the query element, or
842+-- 'Nothing' if there is no such element. This function is subject to
843+-- fusion.
844+elemIndex :: Char -> Text -> Maybe Int
845+elemIndex c t = S.elemIndex c (stream t)
1533846
1534847-------------------------------------------------------------------------------
1535848-- * Zipping
1536849
1537−-- | /O(n)/ 'zip' takes two 'Text's and returns a list of
1538−-- corresponding pairs of bytes. If one input 'Text' is short,
1539−-- excess elements of the longer 'Text' are discarded. This is
1540−-- equivalent to a pair of 'unpack' operations.
1541−zip :: Text -> Text -> [(Char,Char)]
1542−zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
1543−{-# INLINE zip #-}
1544−
1545850-- | /O(n)/ 'zipWith' generalises 'zip' by zipping with the function
1546851-- given as the first argument, instead of a tupling function.
1547−-- Performs replacement on invalid scalar values.
1548852zipWith :: (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 #-}
853+zipWith f t1 t2 = unstream (S.zipWith f (stream t1) (stream t2))
1552854
1553855-- | /O(n)/ Breaks a 'Text' up into a list of words, delimited by 'Char's
1554856-- representing white space.
@@ -1564,41 +866,21 @@ words t@(Text arr off len) = loop 0 0
1564866 then loop (start+1) (start+1)
1565867 else Text arr (start+off) (n-start) : loop (n+d) (n+d)
1566868 | otherwise = loop start (n+d)
1567− where Iter c d = iter t n
869+ where (c,d) = iter t n
1568870{-# INLINE words #-}
1569871
1570872-- | /O(n)/ Breaks a 'Text' up into a list of 'Text's at
1571873-- newline 'Char's. The resulting strings do not contain newlines.
874+--
1572875lines :: Text -> [Text]
1573−lines ps | null ps = []
1574− | otherwise = h : if null t
1575− then []
1576− else lines (unsafeTail t)
1577− where (# h,t #) = span_ (/= '\n') ps
876+lines ps
877+ | null ps = []
878+ | otherwise = case search ps of
879+ Nothing -> [ps]
880+ Just n -> take n ps : lines (drop (n+1) ps)
881+ where search = elemIndex '\n'
1578882{-# INLINE lines #-}
1579883
1580−{-
1581−-- | /O(n)/ Portably breaks a 'Text' up into a list of 'Text's at line
1582−-- boundaries.
1583−--
1584−-- A line boundary is considered to be either a line feed, a carriage
1585−-- return immediately followed by a line feed, or a carriage return.
1586−-- This accounts for both Unix and Windows line ending conventions,
1587−-- and for the old convention used on Mac OS 9 and earlier.
1588−lines' :: Text -> [Text]
1589−lines' ps | null ps = []
1590− | otherwise = h : case uncons t of
1591− Nothing -> []
1592− Just (c,t')
1593− | c == '\n' -> lines t'
1594− | c == '\r' -> case uncons t' of
1595− Just ('\n',t'') -> lines t''
1596− _ -> lines t'
1597− where (h,t) = span notEOL ps
1598− notEOL c = c /= '\n' && c /= '\r'
1599−{-# INLINE lines' #-}
1600−-}
1601−
1602884-- | /O(n)/ Joins lines, after appending a terminating newline to
1603885-- each.
1604886unlines :: [Text] -> Text
@@ -1611,7 +893,8 @@ unwords = intercalate (singleton ' ')
1611893{-# INLINE unwords #-}
1612894
1613895-- | /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.
896+-- 'True' iff the first is a prefix of the second. This function is
897+-- subject to fusion.
1615898isPrefixOf :: Text -> Text -> Bool
1616899isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
1617900 alen <= blen && S.isPrefixOf (stream a) (stream b)
@@ -1620,6 +903,8 @@ isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
1620903{-# RULES
1621904"TEXT isPrefixOf -> fused" [~1] forall s t.
1622905 isPrefixOf s t = S.isPrefixOf (stream s) (stream t)
906+"TEXT isPrefixOf -> unfused" [1] forall s t.
907+ S.isPrefixOf (stream s) (stream t) = isPrefixOf s t
1623908 #-}
1624909
1625910-- | /O(n)/ The 'isSuffixOf' function takes two 'Text's and returns
@@ -1632,124 +917,12 @@ isSuffixOf a@(Text _aarr _aoff alen) b@(Text barr boff blen) =
1632917 | otherwise = Text barr (boff+d) alen
1633918{-# INLINE isSuffixOf #-}
1634919
1635−-- | /O(n+m)/ The 'isInfixOf' function takes two 'Text's and returns
920+-- | /O(n)/ The 'isInfixOf' function takes two 'Text's and returns
1636921-- 'True' iff the first is contained, wholly and intact, anywhere
1637922-- within the second.
1638−--
1639−-- In (unlikely) bad cases, this function's time complexity degrades
1640−-- towards /O(n*m)/.
1641923isInfixOf :: 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 #-}
924+isInfixOf needle haystack = L.any (isPrefixOf needle) (tails haystack)
925+{-# INLINE isInfixOf #-}
1647926
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
1733−
1734−emptyError :: 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
927+errorEmptyList :: String -> a
928+errorEmptyList fun = error ("Data.Text." ++ fun ++ ": empty list")