@@ -1,29 +1,48 @@
1−{-# LANGUAGE BangPatterns, CPP, Rank2Types #-}
1+{-# LANGUAGE BangPatterns, CPP, MagicHash, Rank2Types, UnboxedTuples #-}
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
39
410-- |
511-- Module : Data.Text
6−-- Copyright : (c) Tom Harper 2008-2009,
7−-- (c) Bryan O'Sullivan 2009,
8−-- (c) Duncan Coutts 2009
12+-- Copyright : (c) 2009, 2010, 2011, 2012 Bryan O'Sullivan,
13+-- (c) 2009 Duncan Coutts,
14+-- (c) 2008, 2009 Tom Harper
915--
1016-- License : BSD-style
11−-- Maintainer : bos@serpentine.com, rtomharper@googlemail.com,
12−-- duncan@haskell.org
17+-- Maintainer : bos@serpentine.com
1318-- Stability : experimental
1419-- Portability : GHC
1520--
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.
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.
1927--
2028-- This module is intended to be imported @qualified@, to avoid name
2129-- clashes with "Prelude" functions, e.g.
2230--
2331-- > 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 >.
2437
2538module Data.Text
2639 (
40+ -- * Strict vs lazy types
41+ -- $strict
42+
43+ -- * Acceptable data
44+ -- $replacement
45+
2746 -- * Fusion
2847 -- $fusion
2948
@@ -47,6 +66,7 @@ module Data.Text
4766 , init
4867 , null
4968 , length
69+ , compareLength
5070
5171 -- * Transformations
5272 , map
@@ -61,6 +81,7 @@ module Data.Text
6181 , toCaseFold
6282 , toLower
6383 , toUpper
84+ , toTitle
6485
6586 -- ** Justification
6687 , justifyLeft
@@ -104,8 +125,11 @@ module Data.Text
104125
105126 -- ** Breaking strings
106127 , take
128+ , takeEnd
107129 , drop
130+ , dropEnd
108131 , takeWhile
132+ , takeWhileEnd
109133 , dropWhile
110134 , dropWhileEnd
111135 , dropAround
@@ -113,10 +137,10 @@ module Data.Text
113137 , stripStart
114138 , stripEnd
115139 , splitAt
116− , spanBy
140+ , breakOn
141+ , breakOnEnd
117142 , break
118− , breakEnd
119− , breakBy
143+ , span
120144 , group
121145 , groupBy
122146 , inits
@@ -124,8 +148,8 @@ module Data.Text
124148
125149 -- ** Breaking into many substrings
126150 -- $split
151+ , splitOn
127152 , split
128− , splitBy
129153 , chunksOf
130154
131155 -- ** Breaking into lines and words
@@ -140,162 +164,309 @@ module Data.Text
140164 , isSuffixOf
141165 , isInfixOf
142166
167+ -- ** View patterns
168+ , stripPrefix
169+ , stripSuffix
170+ , commonPrefixes
171+
143172 -- * Searching
144173 , filter
174+ , breakOnAll
145175 , find
146− , findBy
147− , partitionBy
176+ , partition
148177
149178 -- , findSubstring
150−
179+
151180 -- * Indexing
152181 -- $index
153182 , index
154183 , findIndex
155184 , count
156185
157− -- * Zipping and unzipping
186+ -- * Zipping
158187 , zip
159188 , zipWith
160189
161190 -- -* Ordered text
162191 -- , sort
192+
193+ -- * Low level operations
194+ , copy
195+ , unpackCString#
163196 ) where
164197
165−import Prelude (Char, Bool(..), Functor(..), Int, Maybe(..), String,
166− Eq(..), Ord(..), (++),
167− Read(..), Show(..),
168− (&&), (||), (+), (-), (.), ($), (>>), (*),
169− div, error, not, return, otherwise)
198+import Prelude (Char, Bool(..), Int, Maybe(..), String,
199+ Eq(..), Ord(..), Ordering(..), (++),
200+ Read(..),
201+ (&&), (||), (+), (-), (.), ($), ($!), (>>),
202+ not, return, otherwise, quot)
170203#if defined(HAVE_DEEPSEQ)
171−import Control.DeepSeq (NFData)
204+import Control.DeepSeq (NFData(rnf))
172205#endif
206+#if defined(ASSERTS)
173207import Control.Exception (assert)
174−import Data.Char (isSpace)
175−import Data.Data (Data(gfoldl, toConstr, gunfold, dataTypeOf))
176−#if __GLASGOW_HASKELL__ >= 612
177−import Data.Data (mkNoRepType)
178−#else
179−import Data.Data (mkNorepType)
180208#endif
209+import Data.Char (isSpace)
210+import Data.Data (Data(gfoldl, toConstr, gunfold, dataTypeOf), constrIndex,
211+ Constr, mkConstr, DataType, mkDataType, Fixity(Prefix))
181212import Control.Monad (foldM)
182213import Control.Monad.ST (ST)
183214import qualified Data.Text.Array as A
184215import qualified Data.List as L
216+import Data.Binary (Binary(get, put))
185217import Data.Monoid (Monoid(..))
186−import Data.Word (Word16)
218+#if MIN_VERSION_base(4,9,0)
219+import Data.Semigroup (Semigroup(..))
220+#endif
187221import Data.String (IsString(..))
188−import qualified Data.Text.Fusion as S
189−import qualified Data.Text.Fusion.Common as S
190−import Data.Text.Fusion (stream, reverseStream, unstream)
191−import Data.Text.Internal (Text(..), empty, text, textP)
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#)
192229import qualified Prelude as P
193−import Data.Text.Unsafe (iter, iter_, reverseIter, unsafeHead, unsafeTail)
194−import Data.Text.UnsafeChar (unsafeChr)
195−import qualified Data.Text.Encoding.Utf16 as U16
196−import Data.Text.Search (indices)
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.
269+
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 >.)
197298
198299-- $fusion
199300--
200301-- Most of the functions in this module are subject to /fusion/,
201302-- meaning that a pipeline of such functions will usually allocate at
202303-- 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\".
203323
204324instance Eq Text where
205− t1 == t2 = stream t1 == stream t2
325+ Text arrA offA lenA == Text arrB offB lenB
326+ | lenA == lenB = A.equal arrA offA arrB offB lenA
327+ | otherwise = False
206328 {-# INLINE (==) #-}
207329
208330instance Ord Text where
209− compare t1 t2 = compare (stream t1) (stream t2)
210− {-# INLINE compare #-}
211−
212−instance Show Text where
213− showsPrec p ps r = showsPrec p (unpack ps) r
331+ compare = compareText
214332
215333instance Read Text where
216334 readsPrec p str = [(pack x,y) | (x,y) <- readsPrec p str]
217335
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+
218344instance Monoid Text where
219345 mempty = empty
346+#if MIN_VERSION_base(4,9,0)
347+ mappend = (<>) -- future-proof definition
348+#else
220349 mappend = append
350+#endif
221351 mconcat = concat
222352
223353instance IsString Text where
224354 fromString = pack
225355
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+
226363#if defined(HAVE_DEEPSEQ)
227−instance NFData Text
364+instance NFData Text where rnf !_ = ()
228365#endif
229366
230−-- This instance preserves data abstraction at the cost of inefficiency.
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.
231376-- We omit reflection services for the sake of data abstraction.
232−--
233−-- This instance was created by copying the behavior of Data.Set and
234−-- Data.Map. If you feel a mistake has been made, please feel free to
235−-- submit improvements.
236377--
237−-- Original discussion is archived here:
238−
239−-- "could we get a Data instance for Data.Text.Text?"
240−-- http://groups.google.com/group/haskell-cafe/browse_thread/thread/b5bbb1b28a7e525d/0639d46852575b93
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, ... >
241389
242390instance Data Text where
243391 gfoldl f z txt = z pack `f` (unpack txt)
244− toConstr _ = error "Data.Text.Text.toConstr"
245− gunfold _ _ = error "Data.Text.Text.gunfold"
246−#if __GLASGOW_HASKELL__ >= 612
247− dataTypeOf _ = mkNoRepType "Data.Text.Text"
248−#else
249− dataTypeOf _ = mkNorepType "Data.Text.Text"
392+ toConstr _ = packConstr
393+ gunfold k z c = case constrIndex c of
394+ 1 -> k (z pack)
395+ _ -> P.error "gunfold"
396+ dataTypeOf _ = textDataType
397+
398+#if MIN_VERSION_base(4,7,0)
399+-- | Only defined for @base-4.7.0.0@ and later
400+instance PrintfArg Text where
401+ formatArg txt = formatString $ unpack txt
250402#endif
251403
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+
252424-- -----------------------------------------------------------------------------
253425-- * Conversion to/from 'Text'
254426
255−-- | /O(n)/ Convert a 'String' into a 'Text'. Subject to fusion.
427+-- | /O(n)/ Convert a 'String' into a 'Text'. Subject to
428+-- fusion. Performs replacement on invalid scalar values.
256429pack :: String -> Text
257−pack = unstream . S.streamList
430+pack = unstream . S.map safe . S.streamList
258431{-# INLINE [1] pack #-}
259432
260−-- | /O(n)/ Convert a Text into a String. Subject to fusion.
261−unpack :: Text -> String
262−unpack = S.unstreamList . stream
263−{-# INLINE [1] unpack #-}
264−
265−-- | /O(1)/ Convert a character into a Text.
266−-- Subject to fusion.
267−singleton :: Char -> Text
268−singleton = unstream . S.singleton
269−{-# INLINE [1] singleton #-}
270−
271433-- -----------------------------------------------------------------------------
272434-- * Basic functions
273435
274436-- | /O(n)/ Adds a character to the front of a 'Text'. This function
275437-- is more costly than its 'List' counterpart because it requires
276−-- copying a new array. Subject to fusion.
438+-- copying a new array. Subject to fusion. Performs replacement on
439+-- invalid scalar values.
277440cons :: Char -> Text -> Text
278−cons c t = unstream (S.cons c (stream t))
441+cons c t = unstream (S.cons (safe c) (stream t))
279442{-# INLINE cons #-}
280443
444+infixr 5 `cons`
445+
281446-- | /O(n)/ Adds a character to the end of a 'Text'. This copies the
282447-- entire array in the process, unless fused. Subject to fusion.
448+-- Performs replacement on invalid scalar values.
283449snoc :: Text -> Char -> Text
284−snoc t c = unstream (S.snoc (stream t) c)
450+snoc t c = unstream (S.snoc (stream t) (safe c))
285451{-# INLINE snoc #-}
286452
287453-- | /O(n)/ Appends one 'Text' to the other by copying both of them
288454-- into a new 'Text'. Subject to fusion.
289455append :: Text -> Text -> Text
290−append (Text arr1 off1 len1) (Text arr2 off2 len2) = Text (A.run x) 0 len
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"
291461 where
292462 len = len1+len2
463+ x :: ST s (A.MArray s)
293464 x = do
294− arr <- A.unsafeNew len
295− copy arr 0 arr1 off1 len1
296− copy arr len1 arr2 off2 (len1+len2)
465+ arr <- A.new len
466+ A.copyI arr 0 arr1 off1 len1
467+ A.copyI arr len1 arr2 off2 len
297468 return arr
298−{-# INLINE append #-}
469+{-# NOINLINE append #-}
299470
300471{-# RULES
301472"TEXT append -> fused" [~1] forall t1 t2.
@@ -304,14 +475,6 @@ append (Text arr1 off1 len1) (Text arr2 off2 len2) = Text (A.run x) 0 len
304475 unstream (S.append (stream t1) (stream t2)) = append t1 t2
305476 #-}
306477
307−copy :: forall s. A.MArray s Word16 -> Int -> A.Array Word16 -> Int -> Int
308− -> ST s ()
309−copy dest i0 src j0 top = go i0 j0
310− where
311− go i j | i >= top = return ()
312− | otherwise = do A.unsafeWrite dest i (src `A.unsafeIndex` j)
313− go (i+1) (j+1)
314−
315478-- | /O(1)/ Returns the first character of a 'Text', which must be
316479-- non-empty. Subject to fusion.
317480head :: Text -> Char
@@ -323,21 +486,14 @@ head t = S.head (stream t)
323486uncons :: Text -> Maybe (Char, Text)
324487uncons t@(Text arr off len)
325488 | len <= 0 = Nothing
326− | otherwise = Just (c, textP arr (off+d) (len-d))
327− where (c,d) = iter t 0
489+ | otherwise = Just $ let !(Iter c d) = iter t 0
490+ in (c, text arr (off+d) (len-d))
328491{-# INLINE [1] uncons #-}
329492
330493-- | Lifted from Control.Arrow and specialized.
331494second :: (b -> c) -> (a,b) -> (a,c)
332495second f (a, b) = (a, f b)
333496
334−{-# RULES
335−"TEXT uncons -> fused" [~1] forall t.
336− uncons t = fmap (second unstream) (S.uncons (stream t))
337−"TEXT uncons -> unfused" [1] forall t.
338− fmap (second unstream) (S.uncons (stream t)) = uncons t
339− #-}
340−
341497-- | /O(1)/ Returns the last character of a 'Text', which must be
342498-- non-empty. Subject to fusion.
343499last :: Text -> Char
@@ -361,7 +517,7 @@ last (Text arr off len)
361517tail :: Text -> Text
362518tail t@(Text arr off len)
363519 | len <= 0 = emptyError "tail"
364− | otherwise = textP arr (off+d) (len-d)
520+ | otherwise = text arr (off+d) (len-d)
365521 where d = iter_ t 0
366522{-# INLINE [1] tail #-}
367523
@@ -376,8 +532,8 @@ tail t@(Text arr off len)
376532-- be non-empty. Subject to fusion.
377533init :: Text -> Text
378534init (Text arr off len) | len <= 0 = emptyError "init"
379− | n >= 0xDC00 && n <= 0xDFFF = textP arr off (len-2)
380− | otherwise = textP arr off (len-1)
535+ | n >= 0xDC00 && n <= 0xDFFF = text arr off (len-2)
536+ | otherwise = text arr off (len-1)
381537 where
382538 n = A.unsafeIndex arr (off+len-1)
383539{-# INLINE [1] init #-}
@@ -392,7 +548,11 @@ init (Text arr off len) | len <= 0 = emptyError "init"
392548-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to
393549-- fusion.
394550null :: Text -> Bool
395−null (Text _arr _off len) = assert (len >= 0) $ len <= 0
551+null (Text _arr _off len) =
552+#if defined(ASSERTS)
553+ assert (len >= 0) $
554+#endif
555+ len <= 0
396556{-# INLINE [1] null #-}
397557
398558{-# RULES
@@ -412,27 +572,76 @@ isSingleton = S.isSingleton . stream
412572-- Subject to fusion.
413573length :: Text -> Int
414574length t = S.length (stream t)
415−{-# INLINE length #-}
575+{-# INLINE [0] length #-}
576+-- length needs to be phased after the compareN/length rules otherwise
577+-- it may inline before the rules have an opportunity to fire.
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+ #-}
416623
417624-- -----------------------------------------------------------------------------
418625-- * Transformations
419626-- | /O(n)/ 'map' @f@ @t@ is the 'Text' obtained by applying @f@ to
420−-- each element of @t@. Subject to fusion.
627+-- each element of @t@. Subject to fusion. Performs replacement on
628+-- invalid scalar values.
421629map :: (Char -> Char) -> Text -> Text
422−map f t = unstream (S.map f (stream t))
630+map f t = unstream (S.map (safe . f) (stream t))
423631{-# INLINE [1] map #-}
424632
425633-- | /O(n)/ The 'intercalate' function takes a 'Text' and a list of
426634-- 'Text's and concatenates the list after interspersing the first
427635-- argument between each element of the list.
428636intercalate :: Text -> [Text] -> Text
429−intercalate t = concat . (L.intersperse t)
637+intercalate t = concat . (F.intersperse t)
430638{-# INLINE intercalate #-}
431639
432640-- | /O(n)/ The 'intersperse' function takes a character and places it
433−-- between the characters of a 'Text'. Subject to fusion.
641+-- between the characters of a 'Text'. Subject to fusion. Performs
642+-- replacement on invalid scalar values.
434643intersperse :: Char -> Text -> Text
435−intersperse c t = unstream (S.intersperse c (stream t))
644+intersperse c t = unstream (S.intersperse (safe c) (stream t))
436645{-# INLINE intersperse #-}
437646
438647-- | /O(n)/ Reverse the characters of a string. Subject to fusion.
@@ -440,13 +649,59 @@ reverse :: Text -> Text
440649reverse t = S.reverse (stream t)
441650{-# INLINE reverse #-}
442651
443−-- | /O(m+n)/ Replace every occurrence of one substring with another.
444−replace :: Text -- ^ Text to search for
445− -> Text -- ^ Replacement text
446− -> Text -- ^ Input text
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.
447678 -> Text
448−replace s d = intercalate d . split s
449−{-# INLINE replace #-}
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
450705
451706-- ----------------------------------------------------------------------------
452707-- ** Case conversions (folds)
@@ -461,10 +716,17 @@ replace s d = intercalate d . split s
461716-- case conversion rules. As a result, these functions may map one
462717-- input character to two or three output characters. For examples,
463718-- 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 >.
464725
465−-- | /O(n)/ Convert a string to folded case. This function is mainly
466−-- useful for performing caseless (also known as case insensitive)
467−-- string comparisons.
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.
468730--
469731-- A string @x@ is a caseless match for a string @y@ if and only if:
470732--
@@ -479,27 +741,56 @@ replace s d = intercalate d . split s
479741-- instead of itself.
480742toCaseFold :: Text -> Text
481743toCaseFold t = unstream (S.toCaseFold (stream t))
482−{-# INLINE [0] toCaseFold #-}
744+{-# INLINE toCaseFold #-}
483745
484746-- | /O(n)/ Convert a string to lower case, using simple case
485−-- conversion. The result string may be longer than the input string.
486−-- For instance, \"İ\" (Latin capital letter I with dot above,
487−-- U+0130) maps to the sequence \"i\" (Latin small letter i, U+0069) followed
488−-- by \" ̇\" (combining dot above, U+0307).
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).
489753toLower :: Text -> Text
490754toLower t = unstream (S.toLower (stream t))
491755{-# INLINE toLower #-}
492756
493757-- | /O(n)/ Convert a string to upper case, using simple case
494−-- conversion. The result string may be longer than the input string.
495−-- For instance, the German \"ß\" (eszett, U+00DF) maps to the
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
496762-- two-letter sequence \"SS\".
497763toUpper :: Text -> Text
498764toUpper t = unstream (S.toUpper (stream t))
499765{-# INLINE toUpper #-}
500766
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+
501789-- | /O(n)/ Left-justify a string to the given length, using the
502−-- specified fill character on the right. Subject to fusion. Examples:
790+-- specified fill character on the right. Subject to fusion.
791+-- Performs replacement on invalid scalar values.
792+--
793+-- Examples:
503794--
504795-- > justifyLeft 7 'x' "foo" == "fooxxxx"
505796-- > justifyLeft 3 'x' "foobar" == "foobar"
@@ -518,7 +809,10 @@ justifyLeft k c t
518809 #-}
519810
520811-- | /O(n)/ Right-justify a string to the given length, using the
521−-- specified fill character on the left. Examples:
812+-- specified fill character on the left. Performs replacement on
813+-- invalid scalar values.
814+--
815+-- Examples:
522816--
523817-- > justifyRight 7 'x' "bar" == "xxxxbar"
524818-- > justifyRight 3 'x' "foobar" == "foobar"
@@ -529,8 +823,11 @@ justifyRight k c t
529823 where len = length t
530824{-# INLINE justifyRight #-}
531825
532−-- | /O(n)/ Center a string to the given length, using the
533−-- specified fill character on either side. Examples:
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:
534831--
535832-- > center 8 'x' "HS" = "xxxHSxxx"
536833center :: Int -> Char -> Text -> Text
@@ -539,7 +836,7 @@ center k c t
539836 | otherwise = replicateChar l c `append` t `append` replicateChar r c
540837 where len = length t
541838 d = k - len
542− r = d `div` 2
839+ r = d `quot` 2
543840 l = d - r
544841{-# INLINE center #-}
545842
@@ -557,12 +854,12 @@ transpose ts = P.map pack (L.transpose (P.map unpack ts))
557854-- (typically the left-identity of the operator), and a 'Text',
558855-- reduces the 'Text' using the binary operator, from left to right.
559856-- Subject to fusion.
560−foldl :: (b -> Char -> b) -> b -> Text -> b
857+foldl :: (a -> Char -> a) -> a -> Text -> a
561858foldl f z t = S.foldl f z (stream t)
562859{-# INLINE foldl #-}
563860
564861-- | /O(n)/ A strict version of 'foldl'. Subject to fusion.
565−foldl' :: (b -> Char -> b) -> b -> Text -> b
862+foldl' :: (a -> Char -> a) -> a -> Text -> a
566863foldl' f z t = S.foldl' f z (stream t)
567864{-# INLINE foldl' #-}
568865
@@ -581,12 +878,12 @@ foldl1' f t = S.foldl1' f (stream t)
581878-- (typically the right-identity of the operator), and a 'Text',
582879-- reduces the 'Text' using the binary operator, from right to left.
583880-- Subject to fusion.
584−foldr :: (Char -> b -> b) -> b -> Text -> b
881+foldr :: (Char -> a -> a) -> a -> Text -> a
585882foldr f z t = S.foldr f z (stream t)
586883{-# INLINE foldr #-}
587884
588885-- | /O(n)/ A variant of 'foldr' that has no starting value argument,
589−-- and thust must be applied to a non-empty 'Text'. Subject to
886+-- and thus must be applied to a non-empty 'Text'. Subject to
590887-- fusion.
591888foldr1 :: (Char -> Char -> Char) -> Text -> Char
592889foldr1 f t = S.foldr1 f (stream t)
@@ -597,14 +894,19 @@ foldr1 f t = S.foldr1 f (stream t)
597894
598895-- | /O(n)/ Concatenate a list of 'Text's.
599896concat :: [Text] -> Text
600−concat ts = Text (A.run go) 0 len
897+concat ts = case ts' of
898+ [] -> empty
899+ [t] -> t
900+ _ -> Text (A.run go) 0 len
601901 where
602− len = L.sum (L.map (\(Text _ _ l) -> l) ts)
902+ ts' = L.filter (not . null) ts
903+ len = sumP "concat" $ L.map lengthWord16 ts'
904+ go :: ST s (A.MArray s)
603905 go = do
604− arr <- A.unsafeNew len
605− let step i (Text a o l) = let j = i + l in copy arr i a o j >> return j
606− foldM step 0 ts >> return arr
607−{-# INLINE concat #-}
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
608910
609911-- | /O(n)/ Map a function over a 'Text' that results in a 'Text', and
610912-- concatenate the results.
@@ -613,13 +915,13 @@ concatMap f = concat . foldr ((:) . f) []
613915{-# INLINE concatMap #-}
614916
615917-- | /O(n)/ 'any' @p@ @t@ determines whether any character in the
616−-- 'Text' @t@ satisifes the predicate @p@. Subject to fusion.
918+-- 'Text' @t@ satisfies the predicate @p@. Subject to fusion.
617919any :: (Char -> Bool) -> Text -> Bool
618920any p t = S.any p (stream t)
619921{-# INLINE any #-}
620922
621923-- | /O(n)/ 'all' @p@ @t@ determines whether all characters in the
622−-- 'Text' @t@ satisify the predicate @p@. Subject to fusion.
924+-- 'Text' @t@ satisfy the predicate @p@. Subject to fusion.
623925all :: (Char -> Bool) -> Text -> Bool
624926all p t = S.all p (stream t)
625927{-# INLINE all #-}
@@ -641,6 +943,7 @@ minimum t = S.minimum (stream t)
641943
642944-- | /O(n)/ 'scanl' is similar to 'foldl', but returns a list of
643945-- successive reduced values from the left. Subject to fusion.
946+-- Performs replacement on invalid scalar values.
644947--
645948-- > scanl f z [x1, x2, ...] == [z, z `f` x1, (z `f` x1) `f` x2, ...]
646949--
@@ -648,11 +951,13 @@ minimum t = S.minimum (stream t)
648951--
649952-- > last (scanl f z xs) == foldl f z xs.
650953scanl :: (Char -> Char -> Char) -> Char -> Text -> Text
651−scanl f z t = unstream (S.scanl f z (stream t))
954+scanl f z t = unstream (S.scanl g z (stream t))
955+ where g a b = safe (f a b)
652956{-# INLINE scanl #-}
653957
654958-- | /O(n)/ 'scanl1' is a variant of 'scanl' that has no starting
655−-- value argument. Subject to fusion.
959+-- value argument. Subject to fusion. Performs replacement on
960+-- invalid scalar values.
656961--
657962-- > scanl1 f [x1, x2, ...] == [x1, x1 `f` x2, ...]
658963scanl1 :: (Char -> Char -> Char) -> Text -> Text
@@ -660,40 +965,42 @@ scanl1 f t | null t = empty
660965 | otherwise = scanl f (unsafeHead t) (unsafeTail t)
661966{-# INLINE scanl1 #-}
662967
663−-- | /O(n)/ 'scanr' is the right-to-left dual of 'scanl'.
968+-- | /O(n)/ 'scanr' is the right-to-left dual of 'scanl'. Performs
969+-- replacement on invalid scalar values.
664970--
665971-- > scanr f v == reverse . scanl (flip f) v . reverse
666972scanr :: (Char -> Char -> Char) -> Char -> Text -> Text
667−scanr f z = S.reverse . S.reverseScanr f z . reverseStream
973+scanr f z = S.reverse . S.reverseScanr g z . reverseStream
974+ where g a b = safe (f a b)
668975{-# INLINE scanr #-}
669976
670977-- | /O(n)/ 'scanr1' is a variant of 'scanr' that has no starting
671−-- value argument. Subject to fusion.
978+-- value argument. Subject to fusion. Performs replacement on
979+-- invalid scalar values.
672980scanr1 :: (Char -> Char -> Char) -> Text -> Text
673981scanr1 f t | null t = empty
674982 | otherwise = scanr f (last t) (init t)
675983{-# INLINE scanr1 #-}
676984
677−-- | /O(n)/ Like a combination of 'map' and 'foldl'. Applies a
985+-- | /O(n)/ Like a combination of 'map' and 'foldl''. Applies a
678986-- function to each element of a 'Text', passing an accumulating
679−-- parameter from left to right, and returns a final 'Text'.
987+-- parameter from left to right, and returns a final 'Text'. Performs
988+-- replacement on invalid scalar values.
680989mapAccumL :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
681−mapAccumL f s t = case uncons t of
682− Nothing -> (s, empty)
683− Just (x, xs) -> (s'', cons y ys)
684− where (s', y ) = f s x
685− (s'',ys) = mapAccumL f s' xs
990+mapAccumL f z0 = S.mapAccumL g z0 . stream
991+ where g a b = second safe (f a b)
992+{-# INLINE mapAccumL #-}
686993
687994-- | The 'mapAccumR' function behaves like a combination of 'map' and
688−-- 'foldr'; it applies a function to each element of a 'Text', passing
689−-- an accumulating parameter from right to left, and returning a final
690−-- value of this accumulator together with the new 'Text'.
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.
6911000mapAccumR :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
692−mapAccumR f s t = case uncons t of
693− Nothing -> (s, empty)
694− Just (x, xs) -> (s'', cons y ys)
695− where (s'',y ) = f s' x
696− (s', ys) = mapAccumR f s xs
1001+mapAccumR f z0 = second reverse . S.mapAccumL g z0 . reverseStream
1002+ where g a b = second safe (f a b)
1003+{-# INLINE mapAccumR #-}
6971004
6981005-- -----------------------------------------------------------------------------
6991006-- ** Generating and unfolding 'Text's
@@ -702,17 +1009,18 @@ mapAccumR f s t = case uncons t of
7021009-- @t@ repeated @n@ times.
7031010replicate :: Int -> Text -> Text
7041011replicate n t@(Text a o l)
705− | n <= 0 || l <= 0 = empty
706− | n == 1 = t
707− | isSingleton t = replicateChar n (unsafeHead t)
708− | otherwise = Text (A.run x) 0 len
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
7091016 where
710− len = l * n
1017+ len = l `mul` n
1018+ x :: ST s (A.MArray s)
7111019 x = do
712− arr <- A.unsafeNew len
1020+ arr <- A.new len
7131021 let loop !d !i | i >= n = return arr
7141022 | otherwise = let m = d + l
715− in copy arr d a o m >> loop m (i+1)
1023+ in A.copyI arr d a o m >> loop m (i+1)
7161024 loop 0 0
7171025{-# INLINE [1] replicate #-}
7181026
@@ -724,7 +1032,7 @@ replicate n t@(Text a o l)
7241032-- | /O(n)/ 'replicateChar' @n@ @c@ is a 'Text' of length @n@ with @c@ the
7251033-- value of every element. Subject to fusion.
7261034replicateChar :: Int -> Char -> Text
727−replicateChar n c = unstream (S.replicateCharI n c)
1035+replicateChar n c = unstream (S.replicateCharI n (safe c))
7281036{-# INLINE replicateChar #-}
7291037
7301038-- | /O(n)/, where @n@ is the length of the result. The 'unfoldr'
@@ -733,9 +1041,9 @@ replicateChar n c = unstream (S.replicateCharI n c)
7331041-- returns 'Nothing' if it is done producing the 'Text', otherwise
7341042-- 'Just' @(a,b)@. In this case, @a@ is the next 'Char' in the
7351043-- string, and @b@ is the seed value for further production. Subject
736−-- to fusion.
1044+-- to fusion. Performs replacement on invalid scalar values.
7371045unfoldr :: (a -> Maybe (Char,a)) -> a -> Text
738−unfoldr f s = unstream (S.unfoldr f s)
1046+unfoldr f s = unstream (S.unfoldr (firstf safe . f) s)
7391047{-# INLINE unfoldr #-}
7401048
7411049-- | /O(n)/ Like 'unfoldr', 'unfoldrN' builds a 'Text' from a seed
@@ -743,9 +1051,9 @@ unfoldr f s = unstream (S.unfoldr f s)
7431051-- first argument to 'unfoldrN'. This function is more efficient than
7441052-- 'unfoldr' when the maximum length of the result is known and
7451053-- correct, otherwise its performance is similar to 'unfoldr'. Subject
746−-- to fusion.
1054+-- to fusion. Performs replacement on invalid scalar values.
7471055unfoldrN :: Int -> (a -> Maybe (Char,a)) -> a -> Text
748−unfoldrN n f s = unstream (S.unfoldrN n f s)
1056+unfoldrN n f s = unstream (S.unfoldrN n (firstf safe . f) s)
7491057{-# INLINE unfoldrN #-}
7501058
7511059-- -----------------------------------------------------------------------------
@@ -758,14 +1066,16 @@ take :: Int -> Text -> Text
7581066take n t@(Text arr off len)
7591067 | n <= 0 = empty
7601068 | n >= len = t
761− | otherwise = Text arr off (loop 0 0)
762− where
763− loop !i !cnt
764− | i >= len || cnt >= n = i
765− | otherwise = loop (i+d) (cnt+1)
766− where d = iter_ t i
1069+ | otherwise = text arr off (iterN n t)
7671070{-# INLINE [1] take #-}
7681071
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+
7691079{-# RULES
7701080"TEXT take -> fused" [~1] forall n t.
7711081 take n t = unstream (S.take n (stream t))
@@ -773,18 +1083,36 @@ take n t@(Text arr off len)
7731083 unstream (S.take n (stream t)) = take n t
7741084 #-}
7751085
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+
7761107-- | /O(n)/ 'drop' @n@, applied to a 'Text', returns the suffix of the
777−-- 'Text' of length @n@, or the empty 'Text' if @n@ is greater than the
778−-- length of the 'Text'. Subject to fusion.
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.
7791110drop :: Int -> Text -> Text
7801111drop n t@(Text arr off len)
7811112 | n <= 0 = t
7821113 | n >= len = empty
783− | otherwise = loop 0 0
784− where loop !i !cnt
785− | i >= len || cnt >= n = Text arr (off+i) (len-i)
786− | otherwise = loop (i+d) (cnt+1)
787− where d = iter_ t i
1114+ | otherwise = text arr (off+i) (len-i)
1115+ where i = iterN n t
7881116{-# INLINE [1] drop #-}
7891117
7901118{-# RULES
@@ -794,6 +1122,18 @@ drop n t@(Text arr off len)
7941122 unstream (S.drop n (stream t)) = drop n t
7951123 #-}
7961124
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+
7971137-- | /O(n)/ 'takeWhile', applied to a predicate @p@ and a 'Text',
7981138-- returns the longest prefix (possibly empty) of elements that
7991139-- satisfy @p@. Subject to fusion.
@@ -801,8 +1141,8 @@ takeWhile :: (Char -> Bool) -> Text -> Text
8011141takeWhile p t@(Text arr off len) = loop 0
8021142 where loop !i | i >= len = t
8031143 | p c = loop (i+d)
804− | otherwise = textP arr off i
805− where (c,d) = iter t i
1144+ | otherwise = text arr off i
1145+ where Iter c d = iter t i
8061146{-# INLINE [1] takeWhile #-}
8071147
8081148{-# RULES
@@ -812,6 +1152,27 @@ takeWhile p t@(Text arr off len) = loop 0
8121152 unstream (S.takeWhile p (stream t)) = takeWhile p t
8131153 #-}
8141154
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+
8151176-- | /O(n)/ 'dropWhile' @p@ @t@ returns the suffix remaining after
8161177-- 'takeWhile' @p@ @t@. Subject to fusion.
8171178dropWhile :: (Char -> Bool) -> Text -> Text
@@ -819,7 +1180,7 @@ dropWhile p t@(Text arr off len) = loop 0 0
8191180 where loop !i !l | l >= len = empty
8201181 | p c = loop (i+d) (l+d)
8211182 | otherwise = Text arr (off+i) (len-l)
822− where (c,d) = iter t i
1183+ where Iter c d = iter t i
8231184{-# INLINE [1] dropWhile #-}
8241185
8251186{-# RULES
@@ -830,8 +1191,9 @@ dropWhile p t@(Text arr off len) = loop 0 0
8301191 #-}
8311192
8321193-- | /O(n)/ 'dropWhileEnd' @p@ @t@ returns the prefix remaining after
833−-- dropping characters that fail the predicate @p@ from the end of
1194+-- dropping characters that satisfy the predicate @p@ from the end of
8341195-- @t@. Subject to fusion.
1196+--
8351197-- Examples:
8361198--
8371199-- > dropWhileEnd (=='.') "foo..." == "foo"
@@ -851,7 +1213,7 @@ dropWhileEnd p t@(Text arr off len) = loop (len-1) len
8511213 #-}
8521214
8531215-- | /O(n)/ 'dropAround' @p@ @t@ returns the substring remaining after
854−-- dropping characters that fail the predicate @p@ from both the
1216+-- dropping characters that satisfy the predicate @p@ from both the
8551217-- beginning and end of @t@. Subject to fusion.
8561218dropAround :: (Char -> Bool) -> Text -> Text
8571219dropAround p = dropWhile p . dropWhileEnd p
@@ -886,31 +1248,23 @@ splitAt :: Int -> Text -> (Text, Text)
8861248splitAt n t@(Text arr off len)
8871249 | n <= 0 = (empty, t)
8881250 | n >= len = (t, empty)
889− | otherwise = (Text arr off k, Text arr (off+k) (len-k))
890− where k = loop 0 0
891− loop !i !cnt
892− | i >= len || cnt >= n = i
893− | otherwise = loop (i+d) (cnt+1)
894− where d = iter_ t i
895−{-# INLINE splitAt #-}
1251+ | otherwise = let k = iterN n t
1252+ in (text arr off k, text arr (off+k) (len-k))
8961253
897−-- | /O(n)/ 'spanBy', applied to a predicate @p@ and text @t@, returns
1254+-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns
8981255-- a pair whose first element is the longest prefix (possibly empty)
8991256-- of @t@ of elements that satisfy @p@, and whose second is the
9001257-- remainder of the list.
901−spanBy :: (Char -> Bool) -> Text -> (Text, Text)
902−spanBy p t@(Text arr off len) = (textP arr off k, textP arr (off+k) (len-k))
903− where k = loop 0
904− loop !i | i >= len || not (p c) = i
905− | otherwise = loop (i+d)
906− where (c,d) = iter t i
907−{-# INLINE spanBy #-}
908−
909−-- | /O(n)/ 'breakBy' is like 'spanBy', but the prefix returned is
1258+span :: (Char -> Bool) -> Text -> (Text, Text)
1259+span p t = case span_ p t of
1260+ (# hd,tl #) -> (hd,tl)
1261+{-# INLINE span #-}
1262+
1263+-- | /O(n)/ 'break' is like 'span', but the prefix returned is
9101264-- over elements that fail the predicate @p@.
911−breakBy :: (Char -> Bool) -> Text -> (Text, Text)
912−breakBy p = spanBy (not . p)
913−{-# INLINE breakBy #-}
1265+break :: (Char -> Bool) -> Text -> (Text, Text)
1266+break p = span (not . p)
1267+{-# INLINE break #-}
9141268
9151269-- | /O(n)/ Group characters in a string according to a predicate.
9161270groupBy :: (Char -> Char -> Bool) -> Text -> [Text]
@@ -919,7 +1273,7 @@ groupBy p = loop
9191273 loop t@(Text arr off len)
9201274 | null t = []
9211275 | otherwise = text arr off n : loop (text arr (off+n) (len-n))
922− where (c,d) = iter t 0
1276+ where Iter c d = iter t 0
9231277 n = d + findAIndexOrEnd (not . p c) (Text arr (off+d) (len-d))
9241278
9251279-- | Returns the /array/ index (in units of 'Word16') at which a
@@ -929,8 +1283,8 @@ findAIndexOrEnd :: (Char -> Bool) -> Text -> Int
9291283findAIndexOrEnd q t@(Text _arr _off len) = go 0
9301284 where go !i | i >= len || q c = i
9311285 | otherwise = go (i+d)
932− where (c,d) = iter t i
933−
1286+ where Iter c d = iter t i
1287+
9341288-- | /O(n)/ Group characters in a string by equality.
9351289group :: Text -> [Text]
9361290group = groupBy (==)
@@ -954,36 +1308,43 @@ tails t | null t = [empty]
9541308-- copies to create substrings; they just construct new 'Text's that
9551309-- are slices of the original.
9561310
957−-- | /O(m+n)/ Break a 'Text' into pieces separated by the first
958−-- 'Text' argument, consuming the delimiter. An empty delimiter is
959−-- invalid, and will cause an error to be raised.
1311+-- | /O(m+n)/ Break a 'Text' into pieces separated by the first 'Text'
1312+-- argument (which cannot be empty), consuming the delimiter. An empty
1313+-- delimiter is invalid, and will cause an error to be raised.
9601314--
9611315-- Examples:
9621316--
963−-- > split "\r\n" "a\r\nb\r\nd\r\ne" == ["a","b","d","e"]
964−-- > split "aaa" "aaaXaaaXaaaXaaa" == ["","X","X","X",""]
965−-- > split "x" "x" == ["",""]
966−--
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" == ["",""]
1320+--
9671321-- and
9681322--
969−-- > intercalate s . split s == id
970−-- > split (singleton c) == splitBy (==c)
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.)
9711327--
9721328-- In (unlikely) bad cases, this function's time complexity degrades
9731329-- towards /O(n*m)/.
974−split :: Text -> Text -> [Text]
975−split pat@(Text _ _ l) src@(Text arr off len)
976− | l <= 0 = emptyError "split"
977− | isSingleton pat = splitBy (== unsafeHead pat) src
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
9781339 | otherwise = go 0 (indices pat src)
9791340 where
980− go !s (x:xs) = textP arr (s+off) (x-s) : go (x+l) xs
981− go s _ = [textP arr (s+off) (len-s)]
982−{-# INLINE [1] split #-}
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 #-}
9831344
9841345{-# RULES
985−"TEXT split/singleton -> splitBy/==" [~1] forall c t.
986− split (singleton c) t = splitBy (==c) t
1346+"TEXT splitOn/singleton -> split/==" [~1] forall c t.
1347+ splitOn (singleton c) t = split (==c) t
9871348 #-}
9881349
9891350-- | /O(n)/ Splits a 'Text' into components delimited by separators,
@@ -991,15 +1352,15 @@ split pat@(Text _ _ l) src@(Text arr off len)
9911352-- resulting components do not contain the separators. Two adjacent
9921353-- separators result in an empty component in the output. eg.
9931354--
994−-- > splitBy (=='a') "aabbaca" == ["","","bb","c",""]
995−-- > splitBy (=='a') "" == [""]
996−splitBy :: (Char -> Bool) -> Text -> [Text]
997−splitBy _ t@(Text _off _arr 0) = [t]
998−splitBy p t = loop t
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
9991360 where loop s | null s' = [l]
10001361 | otherwise = l : loop (unsafeTail s')
1001− where (l, s') = breakBy p s
1002−{-# INLINE splitBy #-}
1362+ where (# l, s' #) = span_ (not . p) s
1363+{-# INLINE split #-}
10031364
10041365-- | /O(n)/ Splits a 'Text' into components of length @k@. The last
10051366-- element may be shorter than the other chunks, depending on the
@@ -1021,21 +1382,21 @@ chunksOf k = go
10211382-------------------------------------------------------------------------------
10221383-- ** Searching with a predicate
10231384
1024−-- | /O(n)/ The 'findBy' function takes a predicate and a 'Text', and
1025−-- returns the first element in matching the predicate, or 'Nothing'
1026−-- if there is no such element.
1027−findBy :: (Char -> Bool) -> Text -> Maybe Char
1028−findBy p t = S.findBy p (stream t)
1029−{-# INLINE findBy #-}
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.
1388+find :: (Char -> Bool) -> Text -> Maybe Char
1389+find p t = S.findBy p (stream t)
1390+{-# INLINE find #-}
10301391
1031−-- | /O(n)/ The 'partitionBy' function takes a predicate and a 'Text',
1392+-- | /O(n)/ The 'partition' function takes a predicate and a 'Text',
10321393-- and returns the pair of 'Text's with elements which do and do not
10331394-- satisfy the predicate, respectively; i.e.
10341395--
1035−-- > partitionBy p t == (filter p t, filter (not . p) t)
1036−partitionBy :: (Char -> Bool) -> Text -> (Text, Text)
1037−partitionBy p t = (filter p t, filter (not . p) t)
1038−{-# INLINE partitionBy #-}
1396+-- > partition p t == (filter p t, filter (not . p) t)
1397+partition :: (Char -> Bool) -> Text -> (Text, Text)
1398+partition p t = (filter p t, filter (not . p) t)
1399+{-# INLINE partition #-}
10391400
10401401-- | /O(n)/ 'filter', applied to a predicate and a 'Text',
10411402-- returns a 'Text' containing those characters that satisfy the
@@ -1051,39 +1412,40 @@ filter p t = unstream (S.filter p (stream t))
10511412--
10521413-- Examples:
10531414--
1054−-- > break "::" "a::b::c" ==> ("a", "::b::c")
1055−-- > break "/" "foobar" ==> ("foobar", "")
1415+-- > breakOn "::" "a::b::c" ==> ("a", "::b::c")
1416+-- > breakOn "/" "foobar" ==> ("foobar", "")
10561417--
10571418-- Laws:
10581419--
10591420-- > append prefix match == haystack
1060−-- > where (prefix, match) = break needle haystack
1421+-- > where (prefix, match) = breakOn needle haystack
10611422--
10621423-- If you need to break a string by a substring repeatedly (e.g. you
1063−-- want to break on every instance of a substring), use 'find'
1424+-- want to break on every instance of a substring), use 'breakOnAll'
10641425-- instead, as it has lower startup overhead.
10651426--
10661427-- In (unlikely) bad cases, this function's time complexity degrades
10671428-- towards /O(n*m)/.
1068−break :: Text -> Text -> (Text, Text)
1069−break pat src@(Text arr off len)
1070− | null pat = emptyError "break"
1429+breakOn :: Text -> Text -> (Text, Text)
1430+breakOn pat src@(Text arr off len)
1431+ | null pat = emptyError "breakOn"
10711432 | otherwise = case indices pat src of
10721433 [] -> (src, empty)
1073− (x:_) -> (textP arr off x, textP arr (off+x) (len-x))
1074−{-# INLINE break #-}
1434+ (x:_) -> (text arr off x, text arr (off+x) (len-x))
1435+{-# INLINE breakOn #-}
10751436
1076−-- | /O(n+m)/ Similar to 'break', but searches from the end of the string.
1437+-- | /O(n+m)/ Similar to 'breakOn', but searches from the end of the
1438+-- string.
10771439--
10781440-- The first element of the returned tuple is the prefix of @haystack@
10791441-- up to and including the last match of @needle@. The second is the
10801442-- remainder of @haystack@, following the match.
10811443--
1082−-- > breakEnd "::" "a::b::c" ==> ("a::b::", "c")
1083−breakEnd :: Text -> Text -> (Text, Text)
1084−breakEnd pat src = let (a,b) = break (reverse pat) (reverse src)
1085− in (reverse b, reverse a)
1086−{-# INLINE breakEnd #-}
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 #-}
10871449
10881450-- | /O(n+m)/ Find all non-overlapping instances of @needle@ in
10891451-- @haystack@. Each element of the returned list consists of a pair:
@@ -1094,25 +1456,25 @@ breakEnd pat src = let (a,b) = break (reverse pat) (reverse src)
10941456--
10951457-- Examples:
10961458--
1097−-- > find "::" ""
1459+-- > breakOnAll "::" ""
10981460-- > ==> []
1099−-- > find "/" "a/b/c/"
1461+-- > breakOnAll "/" "a/b/c/"
11001462-- > ==> [("a", "/b/c/"), ("a/b", "/c/"), ("a/b/c", "/")]
11011463--
11021464-- In (unlikely) bad cases, this function's time complexity degrades
11031465-- towards /O(n*m)/.
11041466--
11051467-- The @needle@ parameter may not be empty.
1106−find :: Text -- ^ @needle@ to search for
1107− -> Text -- ^ @haystack@ in which to search
1108− -> [(Text, Text)]
1109−find pat@(Text _ _ plen) src@(Text arr off slen)
1110− | null pat = emptyError "find"
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"
11111473 | otherwise = L.map step (indices pat src)
11121474 where
11131475 step x = (chunk 0 x, chunk x (slen-x))
1114− chunk !n !l = textP arr (n+off) l
1115−{-# INLINE find #-}
1476+ chunk !n !l = text arr (n+off) l
1477+{-# INLINE breakOnAll #-}
11161478
11171479-------------------------------------------------------------------------------
11181480-- ** Indexing 'Text's
@@ -1131,7 +1493,7 @@ find pat@(Text _ _ plen) src@(Text arr off slen)
11311493-- For example, suppose you have a string that you want to split on
11321494-- the substring @\"::\"@, such as @\"foo::bar::quux\"@. Instead of
11331495-- searching for the index of @\"::\"@ and taking the substrings
1134−-- before and after that index, you would instead use @find \"::\"@.
1496+-- before and after that index, you would instead use @breakOnAll \"::\"@.
11351497
11361498-- | /O(n)/ 'Text' index (subscript) operator, starting from 0.
11371499index :: Text -> Int -> Char
@@ -1178,13 +1540,15 @@ countChar c t = S.countChar c (stream t)
11781540-- equivalent to a pair of 'unpack' operations.
11791541zip :: Text -> Text -> [(Char,Char)]
11801542zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
1181−{-# INLINE [0] zip #-}
1543+{-# INLINE zip #-}
11821544
11831545-- | /O(n)/ 'zipWith' generalises 'zip' by zipping with the function
11841546-- given as the first argument, instead of a tupling function.
1547+-- Performs replacement on invalid scalar values.
11851548zipWith :: (Char -> Char -> Char) -> Text -> Text -> Text
1186−zipWith f t1 t2 = unstream (S.zipWith f (stream t1) (stream t2))
1187−{-# INLINE [0] zipWith #-}
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 #-}
11881552
11891553-- | /O(n)/ Breaks a 'Text' up into a list of words, delimited by 'Char's
11901554-- representing white space.
@@ -1200,7 +1564,7 @@ words t@(Text arr off len) = loop 0 0
12001564 then loop (start+1) (start+1)
12011565 else Text arr (start+off) (n-start) : loop (n+d) (n+d)
12021566 | otherwise = loop start (n+d)
1203− where (c,d) = iter t n
1567+ where Iter c d = iter t n
12041568{-# INLINE words #-}
12051569
12061570-- | /O(n)/ Breaks a 'Text' up into a list of 'Text's at
@@ -1210,7 +1574,7 @@ lines ps | null ps = []
12101574 | otherwise = h : if null t
12111575 then []
12121576 else lines (unsafeTail t)
1213− where (h,t) = spanBy (/= '\n') ps
1577+ where (# h,t #) = span_ (/= '\n') ps
12141578{-# INLINE lines #-}
12151579
12161580{-
@@ -1247,8 +1611,7 @@ unwords = intercalate (singleton ' ')
12471611{-# INLINE unwords #-}
12481612
12491613-- | /O(n)/ The 'isPrefixOf' function takes two 'Text's and returns
1250−-- 'True' iff the first is a prefix of the second. This function is
1251−-- subject to fusion.
1614+-- 'True' iff the first is a prefix of the second. Subject to fusion.
12521615isPrefixOf :: Text -> Text -> Bool
12531616isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
12541617 alen <= blen && S.isPrefixOf (stream a) (stream b)
@@ -1257,8 +1620,6 @@ isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
12571620{-# RULES
12581621"TEXT isPrefixOf -> fused" [~1] forall s t.
12591622 isPrefixOf s t = S.isPrefixOf (stream s) (stream t)
1260−"TEXT isPrefixOf -> unfused" [1] forall s t.
1261− S.isPrefixOf (stream s) (stream t) = isPrefixOf s t
12621623 #-}
12631624
12641625-- | /O(n)/ The 'isSuffixOf' function takes two 'Text's and returns
@@ -1289,5 +1650,106 @@ isInfixOf needle haystack
12891650 isInfixOf (singleton n) h = S.elem n (S.stream h)
12901651 #-}
12911652
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+
12921734emptyError :: String -> a
1293−emptyError fun = P.error ("Data.Text." ++ fun ++ ": empty input")
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