@@ -1,29 +1,48 @@
1−{-# LANGUAGE BangPatterns #-}
1+{-# LANGUAGE BangPatterns, CPP, MagicHash, Rank2Types, UnboxedTuples #-}
2{-# OPTIONS_GHC -fno-warn-orphans #-}
2{-# OPTIONS_GHC -fno-warn-orphans #-}
3+#if __GLASGOW_HASKELL__ >= 702
4+{-# LANGUAGE Trustworthy #-}
6+#if __GLASGOW_HASKELL__ >= 708
7+{-# LANGUAGE TypeFamilies #-}
5-- Module : Data.Text
11-- Module : Data.Text
6−-- Copyright : (c) Tom Harper 2008-2009,
7−-- (c) Bryan O'Sullivan 2009,
12+-- Copyright : (c) 2009, 2010, 2011, 2012 Bryan O'Sullivan,
8−-- (c) Duncan Coutts 2009
13+-- (c) 2009 Duncan Coutts,
14+-- (c) 2008, 2009 Tom Harper
10-- License : BSD-style
16-- License : BSD-style
11−-- Maintainer : bos@serpentine.com, rtharper@aftereternity.co.uk,
17+-- Maintainer : bos@serpentine.com
13-- Stability : experimental
18-- Stability : experimental
14-- Portability : GHC
19-- Portability : GHC
16−-- A time and space-efficient implementation of Unicode text using
21+-- A time and space-efficient implementation of Unicode text.
17−-- packed Word16 arrays. Suitable for performance critical use, both
22+-- Suitable for performance critical use, both in terms of large data
18−-- in terms of large data quantities and high speed.
23+-- quantities and high speed.
25+-- /Note/: Read below the synopsis for important notes on the use of
20-- This module is intended to be imported @qualified@, to avoid name
28-- This module is intended to be imported @qualified@, to avoid name
21-- clashes with "Prelude" functions, e.g.
29-- clashes with "Prelude" functions, e.g.
23-- > import qualified Data.Text as T
31-- > import qualified Data.Text as T
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 >.
25module Data.Text
38module Data.Text
40+ -- * Strict vs lazy types
27 -- * Fusion
46 -- * Fusion
28 -- $fusion
47 -- $fusion
@@ -47,6 +66,7 @@ module Data.Text
51 -- * Transformations
71 -- * Transformations
@@ -61,6 +81,7 @@ module Data.Text
61 , toCaseFold
81 , toCaseFold
65 -- ** Justification
86 -- ** Justification
66 , justifyLeft
87 , justifyLeft
@@ -97,7 +118,6 @@ module Data.Text
98 -- ** Generation and unfolding
119 -- ** Generation and unfolding
99 , replicate
120 , replicate
101 , unfoldr
121 , unfoldr
102 , unfoldrN
122 , unfoldrN
@@ -105,8 +125,11 @@ module Data.Text
106 -- ** Breaking strings
126 -- ** Breaking strings
109 , takeWhile
131 , takeWhile
110 , dropWhile
133 , dropWhile
111 , dropWhileEnd
134 , dropWhileEnd
112 , dropAround
135 , dropAround
@@ -114,8 +137,10 @@ module Data.Text
114 , stripStart
137 , stripStart
115 , stripEnd
138 , stripEnd
116 , splitAt
139 , splitAt
117− , span
140+ , breakOn
120 , groupBy
145 , groupBy
@@ -123,10 +148,8 @@ module Data.Text
124 -- ** Breaking into many substrings
149 -- ** Breaking into many substrings
125 -- $split
150 -- $split
130 , chunksOf
153 , chunksOf
132 -- ** Breaking into lines and words
155 -- ** Breaking into lines and words
@@ -141,136 +164,309 @@ module Data.Text
141 , isSuffixOf
164 , isSuffixOf
142 , isInfixOf
165 , isInfixOf
144 -- * Searching
172 -- * Searching
147− , findAll
174+ , breakOnAll
149 , partition
176 , partition
151 -- , findSubstring
178 -- , findSubstring
153 -- * Indexing
180 -- * Indexing
154 -- $index
181 -- $index
156 , findIndex
183 , findIndex
162− -- * Zipping and unzipping
186+ -- * Zipping
164 , zipWith
188 , zipWith
166 -- -* Ordered text
190 -- -* Ordered text
167 -- , sort
191 -- , sort
193+ -- * Low level operations
170−import Prelude (Char, Bool(..), Functor(..), Int, Maybe(..), String,
198+import Prelude (Char, Bool(..), Int, Maybe(..), String,
171− Eq(..), Ord(..), (++),
199+ Eq(..), Ord(..), Ordering(..), (++),
172− Read(..), Show(..),
200+ Read(..),
173− (&&), (||), (+), (-), (.), ($),
201+ (&&), (||), (+), (-), (.), ($), ($!), (>>),
174− fromIntegral, div, not, return, otherwise)
202+ not, return, otherwise, quot)
203+#if defined(HAVE_DEEPSEQ)
204+import Control.DeepSeq (NFData(rnf))
175import Control.Exception (assert)
207import Control.Exception (assert)
176import Data.Char (isSpace)
209import 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)
177import Control.Monad.ST (ST)
213import Control.Monad.ST (ST)
178import qualified Data.Text.Array as A
214import qualified Data.Text.Array as A
179import qualified Data.List as L
215import qualified Data.List as L
216+import Data.Binary (Binary(get, put))
180import Data.Monoid (Monoid(..))
217import Data.Monoid (Monoid(..))
218+#if MIN_VERSION_base(4,9,0)
181−import Data.Word (Word16)
219+import Data.Semigroup (Semigroup(..))
182import Data.String (IsString(..))
221import Data.String (IsString(..))
184−import qualified Data.Text.Fusion as S
222+import qualified Data.Text.Internal.Fusion as S
185−import qualified Data.Text.Fusion.Common as S
223+import qualified Data.Text.Internal.Fusion.Common as S
224+import Data.Text.Encoding (decodeUtf8', encodeUtf8)
186−import Data.Text.Fusion (stream, reverseStream, unstream)
225+import Data.Text.Internal.Fusion (stream, reverseStream, unstream)
187−
226+import Data.Text.Internal.Private (span_)
188−import Data.Text.Internal (Text(..), empty, text, textP)
227+import Data.Text.Internal (Text(..), empty, firstf, mul, safe, text)
228+import Data.Text.Show (singleton, unpack, unpackCString#)
189import qualified Prelude as P
229import qualified Prelude as P
190−import Data.Text.Unsafe (iter, iter_, reverseIter, unsafeHead, unsafeTail)
230+import Data.Text.Unsafe (Iter(..), iter, iter_, lengthWord16, reverseIter,
231+ reverseIter_, unsafeHead, unsafeTail)
191−import Data.Text.UnsafeChar (unsafeChr)
232+import Data.Text.Internal.Unsafe.Char (unsafeChr)
192−import qualified Data.Text.Encoding.Utf16 as U16
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)
241+import GHC.Base (eqInt, neInt, gtInt, geInt, ltInt, leInt)
242+#if __GLASGOW_HASKELL__ >= 708
243+import qualified GHC.Exts as Exts
245+#if MIN_VERSION_base(4,7,0)
246+import Text.Printf (PrintfArg, formatArg, formatString)
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.
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.
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'
272+-- A 'Text' value is a sequence of Unicode scalar values, as defined
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
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
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\".
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 >.)
194-- $fusion
299-- $fusion
196-- Most of the functions in this module are subject to /fusion/,
301-- Most of the functions in this module are subject to /fusion/,
197-- meaning that a pipeline of such functions will usually allocate at
302-- meaning that a pipeline of such functions will usually allocate at
198-- most one 'Text' value.
303-- most one 'Text' value.
305+-- As an example, consider the following pipeline:
307+-- > import Data.Text as T
308+-- > import Data.Text.Encoding as E
309+-- > import Data.ByteString (ByteString)
311+-- > countChars :: ByteString -> Int
312+-- > countChars = T.length . T.toUpper . E.decodeUtf8
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'.
321+-- Functions that can be fused by the compiler are documented with the
322+-- phrase \"Subject to fusion\".
200instance Eq Text where
324instance Eq Text where
201− 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
202 {-# INLINE (==) #-}
328 {-# INLINE (==) #-}
204instance Ord Text where
330instance Ord Text where
205− compare t1 t2 = compare (stream t1) (stream t2)
331+ compare = compareText
206− {-# INLINE compare #-}
208−instance Show Text where
209− showsPrec p ps r = showsPrec p (unpack ps) r
211instance Read Text where
333instance Read Text where
212 readsPrec p str = [(pack x,y) | (x,y) <- readsPrec p str]
334 readsPrec p str = [(pack x,y) | (x,y) <- readsPrec p str]
336+#if MIN_VERSION_base(4,9,0)
337+-- Semigroup orphan instances for older GHCs are provided by
338+-- 'semigroups` package
340+instance Semigroup Text where
214instance Monoid Text where
344instance Monoid Text where
215 mempty = empty
345 mempty = empty
346+#if MIN_VERSION_base(4,9,0)
347+ mappend = (<>) -- future-proof definition
216 mappend = append
349 mappend = append
217 mconcat = concat
351 mconcat = concat
219instance IsString Text where
353instance IsString Text where
220 fromString = pack
354 fromString = pack
356+#if __GLASGOW_HASKELL__ >= 708
357+instance Exts.IsList Text where
358+ type Item Text = Char
363+#if defined(HAVE_DEEPSEQ)
364+instance NFData Text where rnf !_ = ()
367+instance Binary Text where
368+ put t = put (encodeUtf8 t)
371+ case decodeUtf8' bs of
372+ P.Left exn -> P.fail (P.show exn)
373+ P.Right a -> P.return a
375+-- | This instance preserves data abstraction at the cost of inefficiency.
376+-- We omit reflection services for the sake of data abstraction.
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
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? >
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, ... >
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
395+ _ -> P.error "gunfold"
396+ dataTypeOf _ = textDataType
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
405+packConstr = mkConstr textDataType "pack" [] Prefix
407+textDataType :: DataType
408+textDataType = mkDataType "Data.Text.Text" [packConstr]
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
417+ | i >= lenA || j >= lenB = compare lenA lenB
420+ | otherwise = go (i+di) (j+dj)
421+ where Iter a di = iter ta i
422+ Iter b dj = iter tb j
222-- -----------------------------------------------------------------------------
424-- -----------------------------------------------------------------------------
223-- * Conversion to/from 'Text'
425-- * Conversion to/from 'Text'
225−-- | /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.
226pack :: String -> Text
429pack :: String -> Text
227−pack = unstream . S.streamList
430+pack = unstream . S.map safe . S.streamList
228{-# INLINE [1] pack #-}
431{-# INLINE [1] pack #-}
230−-- | /O(n)/ Convert a Text into a String. Subject to fusion.
231−unpack :: Text -> String
232−unpack = S.unstreamList . stream
233−{-# INLINE [1] unpack #-}
235−-- | /O(1)/ Convert a character into a Text.
236−-- Subject to fusion.
237−singleton :: Char -> Text
238−singleton = unstream . S.singleton
239−{-# INLINE [1] singleton #-}
241-- -----------------------------------------------------------------------------
433-- -----------------------------------------------------------------------------
242-- * Basic functions
434-- * Basic functions
244-- | /O(n)/ Adds a character to the front of a 'Text'. This function
436-- | /O(n)/ Adds a character to the front of a 'Text'. This function
245-- is more costly than its 'List' counterpart because it requires
437-- is more costly than its 'List' counterpart because it requires
246−-- copying a new array. Subject to fusion.
438+-- copying a new array. Subject to fusion. Performs replacement on
439+-- invalid scalar values.
247cons :: Char -> Text -> Text
440cons :: Char -> Text -> Text
248−cons c t = unstream (S.cons c (stream t))
441+cons c t = unstream (S.cons (safe c) (stream t))
249{-# INLINE cons #-}
442{-# INLINE cons #-}
251-- | /O(n)/ Adds a character to the end of a 'Text'. This copies the
446-- | /O(n)/ Adds a character to the end of a 'Text'. This copies the
252-- entire array in the process, unless fused. Subject to fusion.
447-- entire array in the process, unless fused. Subject to fusion.
448+-- Performs replacement on invalid scalar values.
253snoc :: Text -> Char -> Text
449snoc :: Text -> Char -> Text
254−snoc t c = unstream (S.snoc (stream t) c)
450+snoc t c = unstream (S.snoc (stream t) (safe c))
255{-# INLINE snoc #-}
451{-# INLINE snoc #-}
257-- | /O(n)/ Appends one 'Text' to the other by copying both of them
453-- | /O(n)/ Appends one 'Text' to the other by copying both of them
258-- into a new 'Text'. Subject to fusion.
454-- into a new 'Text'. Subject to fusion.
259append :: Text -> Text -> Text
455append :: Text -> Text -> Text
260−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)
459+ | len > 0 = Text (A.run x) 0 len
460+ | otherwise = overflowError "append"
262 len = len1+len2
462 len = len1+len2
463+ x :: ST s (A.MArray s)
264− arr <- A.unsafeNew len :: ST s (A.MArray s Word16)
465+ arr <- A.new len
265− copy arr1 off1 (len1+off1) arr 0
466+ A.copyI arr 0 arr1 off1 len1
266− copy arr2 off2 (len2+off2) arr len1
467+ A.copyI arr len1 arr2 off2 len
267 return arr
468 return arr
269− copy arr i top arr' j
270− | i >= top = return ()
271− | otherwise = do A.unsafeWrite arr' j (arr `A.unsafeIndex` i)
272− copy arr (i+1) top arr' (j+1)
273−{-# INLINE append #-}
469+{-# NOINLINE append #-}
276"TEXT append -> fused" [~1] forall t1 t2.
472"TEXT append -> fused" [~1] forall t1 t2.
@@ -290,21 +486,14 @@ head t = S.head (stream t)
290uncons :: Text -> Maybe (Char, Text)
486uncons :: Text -> Maybe (Char, Text)
291uncons t@(Text arr off len)
487uncons t@(Text arr off len)
292 | len <= 0 = Nothing
488 | len <= 0 = Nothing
489+ | otherwise = Just $ let !(Iter c d) = iter t 0
293− | otherwise = Just (c, textP arr (off+d) (len-d))
490+ in (c, text arr (off+d) (len-d))
294− where (c,d) = iter t 0
295{-# INLINE [1] uncons #-}
491{-# INLINE [1] uncons #-}
297-- | Lifted from Control.Arrow and specialized.
493-- | Lifted from Control.Arrow and specialized.
298second :: (b -> c) -> (a,b) -> (a,c)
494second :: (b -> c) -> (a,b) -> (a,c)
299second f (a, b) = (a, f b)
495second f (a, b) = (a, f b)
302−"TEXT uncons -> fused" [~1] forall t.
303− uncons t = fmap (second unstream) (S.uncons (stream t))
304−"TEXT uncons -> unfused" [1] forall t.
305− fmap (second unstream) (S.uncons (stream t)) = uncons t
308-- | /O(1)/ Returns the last character of a 'Text', which must be
497-- | /O(1)/ Returns the last character of a 'Text', which must be
309-- non-empty. Subject to fusion.
498-- non-empty. Subject to fusion.
310last :: Text -> Char
499last :: Text -> Char
@@ -328,7 +517,7 @@ last (Text arr off len)
328tail :: Text -> Text
517tail :: Text -> Text
329tail t@(Text arr off len)
518tail t@(Text arr off len)
330 | len <= 0 = emptyError "tail"
519 | len <= 0 = emptyError "tail"
331− | otherwise = textP arr (off+d) (len-d)
520+ | otherwise = text arr (off+d) (len-d)
332 where d = iter_ t 0
521 where d = iter_ t 0
333{-# INLINE [1] tail #-}
522{-# INLINE [1] tail #-}
@@ -343,8 +532,8 @@ tail t@(Text arr off len)
343-- be non-empty. Subject to fusion.
532-- be non-empty. Subject to fusion.
344init :: Text -> Text
533init :: Text -> Text
345init (Text arr off len) | len <= 0 = emptyError "init"
534init (Text arr off len) | len <= 0 = emptyError "init"
346− | n >= 0xDC00 && n <= 0xDFFF = textP arr off (len-2)
535+ | n >= 0xDC00 && n <= 0xDFFF = text arr off (len-2)
347− | otherwise = textP arr off (len-1)
536+ | otherwise = text arr off (len-1)
349 n = A.unsafeIndex arr (off+len-1)
538 n = A.unsafeIndex arr (off+len-1)
350{-# INLINE [1] init #-}
539{-# INLINE [1] init #-}
@@ -359,7 +548,11 @@ init (Text arr off len) | len <= 0 = emptyError "init"
359-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to
548-- | /O(1)/ Tests whether a 'Text' is empty or not. Subject to
360-- fusion.
549-- fusion.
361null :: Text -> Bool
550null :: Text -> Bool
362−null (Text _arr _off len) = assert (len >= 0) $ len <= 0
551+null (Text _arr _off len) =
363{-# INLINE [1] null #-}
556{-# INLINE [1] null #-}
@@ -369,31 +562,86 @@ null (Text _arr _off len) = assert (len >= 0) $ len <= 0
369 S.null (stream t) = null t
562 S.null (stream t) = null t
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 #-}
372-- | /O(n)/ Returns the number of characters in a 'Text'.
571-- | /O(n)/ Returns the number of characters in a 'Text'.
373-- Subject to fusion.
572-- Subject to fusion.
374length :: Text -> Int
573length :: Text -> Int
375length t = S.length (stream t)
574length t = S.length (stream t)
376−{-# 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.
579+-- | /O(n)/ Compare the count of characters in a 'Text' to a number.
580+-- Subject to fusion.
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 #-}
590+"TEXT compareN/length -> compareLength" [~1] forall t n.
591+ compare (length t) n = compareLength t n
595+"TEXT ==N/length -> compareLength/==EQ" [~1] forall t n.
596+ eqInt (length t) n = compareLength t n == EQ
600+"TEXT /=N/length -> compareLength//=EQ" [~1] forall t n.
601+ neInt (length t) n = compareLength t n /= EQ
605+"TEXT <N/length -> compareLength/==LT" [~1] forall t n.
606+ ltInt (length t) n = compareLength t n == LT
610+"TEXT <=N/length -> compareLength//=GT" [~1] forall t n.
611+ leInt (length t) n = compareLength t n /= GT
615+"TEXT >N/length -> compareLength/==GT" [~1] forall t n.
616+ gtInt (length t) n = compareLength t n == GT
620+"TEXT >=N/length -> compareLength//=LT" [~1] forall t n.
621+ geInt (length t) n = compareLength t n /= LT
378-- -----------------------------------------------------------------------------
624-- -----------------------------------------------------------------------------
379-- * Transformations
625-- * Transformations
380-- | /O(n)/ 'map' @f@ @t@ is the 'Text' obtained by applying @f@ to
626-- | /O(n)/ 'map' @f@ @t@ is the 'Text' obtained by applying @f@ to
381−-- each element of @t@. Subject to fusion.
627+-- each element of @t@. Subject to fusion. Performs replacement on
628+-- invalid scalar values.
382map :: (Char -> Char) -> Text -> Text
629map :: (Char -> Char) -> Text -> Text
383−map f t = unstream (S.map f (stream t))
630+map f t = unstream (S.map (safe . f) (stream t))
384{-# INLINE [1] map #-}
631{-# INLINE [1] map #-}
386-- | /O(n)/ The 'intercalate' function takes a 'Text' and a list of
633-- | /O(n)/ The 'intercalate' function takes a 'Text' and a list of
387-- 'Text's and concatenates the list after interspersing the first
634-- 'Text's and concatenates the list after interspersing the first
388-- argument between each element of the list.
635-- argument between each element of the list.
389intercalate :: Text -> [Text] -> Text
636intercalate :: Text -> [Text] -> Text
390−intercalate t ts = unstream (S.intercalate (stream t) (L.map stream ts))
637+intercalate t = concat . (F.intersperse t)
391{-# INLINE intercalate #-}
638{-# INLINE intercalate #-}
393-- | /O(n)/ The 'intersperse' function takes a character and places it
640-- | /O(n)/ The 'intersperse' function takes a character and places it
394−-- 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.
395intersperse :: Char -> Text -> Text
643intersperse :: Char -> Text -> Text
396−intersperse c t = unstream (S.intersperse c (stream t))
644+intersperse c t = unstream (S.intersperse (safe c) (stream t))
397{-# INLINE intersperse #-}
645{-# INLINE intersperse #-}
399-- | /O(n)/ Reverse the characters of a string. Subject to fusion.
647-- | /O(n)/ Reverse the characters of a string. Subject to fusion.
@@ -401,13 +649,59 @@ reverse :: Text -> Text
401reverse t = S.reverse (stream t)
649reverse t = S.reverse (stream t)
402{-# INLINE reverse #-}
650{-# INLINE reverse #-}
404−-- | /O(m*n)/ Replace every occurrence of one substring with another.
652+-- | /O(m+n)/ Replace every non-overlapping occurrence of @needle@ in
653+-- @haystack@ with @replacement@.
655+-- This function behaves as though it was defined as follows:
658+-- replace needle replacement haystack =
659+-- 'intercalate' replacement ('splitOn' needle haystack)
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:
666+-- > replace "oo" "foo" "oo" == "foo"
668+-- In cases where several instances of @needle@ overlap, only the
669+-- first one will be replaced:
671+-- > replace "ofo" "bar" "ofofo" == "barfo"
673+-- In (unlikely) bad cases, this function's time complexity degrades
405−replace :: Text -- ^ Text to search for
675+replace :: Text
406− -> Text -- ^ Replacement text
676+ -- ^ @needle@ to search for. If this string is empty, an
677+ -- error will occur.
679+ -- ^ @replacement@ to replace @needle@ with.
407− -> Text -- ^ Input text
680+ -> Text
681+ -- ^ @haystack@ in which to search.
409−replace s d = intercalate d . split s
683+replace needle@(Text _ _ neeLen)
410−{-# INLINE replace #-}
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
691+ ixs = indices needle haystack
692+ len = hayLen - (neeLen - repLen) `mul` L.length ixs
693+ x :: ST s (A.MArray s)
696+ let loop (i:is) o d = do
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
412-- ----------------------------------------------------------------------------
706-- ----------------------------------------------------------------------------
413-- ** Case conversions (folds)
707-- ** Case conversions (folds)
@@ -422,10 +716,17 @@ replace s d = intercalate d . split s
422-- case conversion rules. As a result, these functions may map one
716-- case conversion rules. As a result, these functions may map one
423-- input character to two or three output characters. For examples,
717-- input character to two or three output characters. For examples,
424-- see the documentation of each function.
718-- see the documentation of each function.
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 >.
426−-- | /O(n)/ Convert a string to folded case. This function is mainly
726+-- | /O(n)/ Convert a string to folded case. Subject to fusion.
427−-- useful for performing caseless (also known as case insensitive)
728+-- This function is mainly useful for performing caseless (also known
428−-- string comparisons.
729+-- as case insensitive) string comparisons.
430-- A string @x@ is a caseless match for a string @y@ if and only if:
731-- A string @x@ is a caseless match for a string @y@ if and only if:
@@ -440,27 +741,56 @@ replace s d = intercalate d . split s
440-- instead of itself.
741-- instead of itself.
441toCaseFold :: Text -> Text
742toCaseFold :: Text -> Text
442toCaseFold t = unstream (S.toCaseFold (stream t))
743toCaseFold t = unstream (S.toCaseFold (stream t))
443−{-# INLINE [0] toCaseFold #-}
744+{-# INLINE toCaseFold #-}
445-- | /O(n)/ Convert a string to lower case, using simple case
746-- | /O(n)/ Convert a string to lower case, using simple case
747+-- conversion. Subject to fusion.
446−-- conversion. The result string may be longer than the input string.
749+-- The result string may be longer than the input string. For
447−-- For instance, \"İ\" (Latin capital letter I with dot above,
750+-- instance, \"İ\" (Latin capital letter I with dot above,
448−-- U+0130) maps to the sequence \"i\" (Latin small letter i, U+0069) followed
751+-- U+0130) maps to the sequence \"i\" (Latin small letter i, U+0069)
449−-- by \" ̇\" (combining dot above, U+0307).
752+-- followed by \" ̇\" (combining dot above, U+0307).
450toLower :: Text -> Text
753toLower :: Text -> Text
451toLower t = unstream (S.toLower (stream t))
754toLower t = unstream (S.toLower (stream t))
452{-# INLINE toLower #-}
755{-# INLINE toLower #-}
454-- | /O(n)/ Convert a string to upper case, using simple case
757-- | /O(n)/ Convert a string to upper case, using simple case
758+-- conversion. Subject to fusion.
455−-- conversion. The result string may be longer than the input string.
760+-- The result string may be longer than the input string. For
456−-- For instance, the German \"ß\" (eszett, U+00DF) maps to the
761+-- instance, the German \"ß\" (eszett, U+00DF) maps to the
457-- two-letter sequence \"SS\".
762-- two-letter sequence \"SS\".
458toUpper :: Text -> Text
763toUpper :: Text -> Text
459toUpper t = unstream (S.toUpper (stream t))
764toUpper t = unstream (S.toUpper (stream t))
460{-# INLINE toUpper #-}
765{-# INLINE toUpper #-}
767+-- | /O(n)/ Convert a string to title case, using simple case
768+-- conversion. Subject to fusion.
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
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).
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 #-}
462-- | /O(n)/ Left-justify a string to the given length, using the
789-- | /O(n)/ Left-justify a string to the given length, using the
463−-- 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.
465-- > justifyLeft 7 'x' "foo" == "fooxxxx"
795-- > justifyLeft 7 'x' "foo" == "fooxxxx"
466-- > justifyLeft 3 'x' "foobar" == "foobar"
796-- > justifyLeft 3 'x' "foobar" == "foobar"
@@ -479,7 +809,10 @@ justifyLeft k c t
481-- | /O(n)/ Right-justify a string to the given length, using the
811-- | /O(n)/ Right-justify a string to the given length, using the
482−-- specified fill character on the left. Examples:
812+-- specified fill character on the left. Performs replacement on
813+-- invalid scalar values.
484-- > justifyRight 7 'x' "bar" == "xxxxbar"
817-- > justifyRight 7 'x' "bar" == "xxxxbar"
485-- > justifyRight 3 'x' "foobar" == "foobar"
818-- > justifyRight 3 'x' "foobar" == "foobar"
@@ -490,8 +823,11 @@ justifyRight k c t
490 where len = length t
823 where len = length t
491{-# INLINE justifyRight #-}
824{-# INLINE justifyRight #-}
493−-- | /O(n)/ Center a string to the given length, using the
826+-- | /O(n)/ Center a string to the given length, using the specified
494−-- specified fill character on either side. Examples:
827+-- fill character on either side. Performs replacement on invalid
496-- > center 8 'x' "HS" = "xxxHSxxx"
832-- > center 8 'x' "HS" = "xxxHSxxx"
497center :: Int -> Char -> Text -> Text
833center :: Int -> Char -> Text -> Text
@@ -500,7 +836,7 @@ center k c t
500 | otherwise = replicateChar l c `append` t `append` replicateChar r c
836 | otherwise = replicateChar l c `append` t `append` replicateChar r c
501 where len = length t
837 where len = length t
502 d = k - len
838 d = k - len
503− r = d `div` 2
839+ r = d `quot` 2
504 l = d - r
840 l = d - r
505{-# INLINE center #-}
841{-# INLINE center #-}
@@ -518,12 +854,12 @@ transpose ts = P.map pack (L.transpose (P.map unpack ts))
518-- (typically the left-identity of the operator), and a 'Text',
854-- (typically the left-identity of the operator), and a 'Text',
519-- reduces the 'Text' using the binary operator, from left to right.
855-- reduces the 'Text' using the binary operator, from left to right.
520-- Subject to fusion.
856-- Subject to fusion.
521−foldl :: (b -> Char -> b) -> b -> Text -> b
857+foldl :: (a -> Char -> a) -> a -> Text -> a
522foldl f z t = S.foldl f z (stream t)
858foldl f z t = S.foldl f z (stream t)
523{-# INLINE foldl #-}
859{-# INLINE foldl #-}
525-- | /O(n)/ A strict version of 'foldl'. Subject to fusion.
861-- | /O(n)/ A strict version of 'foldl'. Subject to fusion.
526−foldl' :: (b -> Char -> b) -> b -> Text -> b
862+foldl' :: (a -> Char -> a) -> a -> Text -> a
527foldl' f z t = S.foldl' f z (stream t)
863foldl' f z t = S.foldl' f z (stream t)
528{-# INLINE foldl' #-}
864{-# INLINE foldl' #-}
@@ -542,12 +878,12 @@ foldl1' f t = S.foldl1' f (stream t)
542-- (typically the right-identity of the operator), and a 'Text',
878-- (typically the right-identity of the operator), and a 'Text',
543-- reduces the 'Text' using the binary operator, from right to left.
879-- reduces the 'Text' using the binary operator, from right to left.
544-- Subject to fusion.
880-- Subject to fusion.
545−foldr :: (Char -> b -> b) -> b -> Text -> b
881+foldr :: (Char -> a -> a) -> a -> Text -> a
546foldr f z t = S.foldr f z (stream t)
882foldr f z t = S.foldr f z (stream t)
547{-# INLINE foldr #-}
883{-# INLINE foldr #-}
549-- | /O(n)/ A variant of 'foldr' that has no starting value argument,
885-- | /O(n)/ A variant of 'foldr' that has no starting value argument,
550−-- 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
551-- fusion.
887-- fusion.
552foldr1 :: (Char -> Char -> Char) -> Text -> Char
888foldr1 :: (Char -> Char -> Char) -> Text -> Char
553foldr1 f t = S.foldr1 f (stream t)
889foldr1 f t = S.foldr1 f (stream t)
@@ -556,28 +892,36 @@ foldr1 f t = S.foldr1 f (stream t)
556-- -----------------------------------------------------------------------------
892-- -----------------------------------------------------------------------------
557-- ** Special folds
893-- ** Special folds
559−-- | /O(n)/ Concatenate a list of 'Text's. Subject to fusion.
895+-- | /O(n)/ Concatenate a list of 'Text's.
560concat :: [Text] -> Text
896concat :: [Text] -> Text
561−concat ts = unstream (S.concat (L.map stream ts))
897+concat ts = case ts' of
562−{-# INLINE concat #-}
898+ [] -> empty
900+ _ -> Text (A.run go) 0 len
902+ ts' = L.filter (not . null) ts
903+ len = sumP "concat" $ L.map lengthWord16 ts'
904+ go :: ST s (A.MArray s)
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
564-- | /O(n)/ Map a function over a 'Text' that results in a 'Text', and
911-- | /O(n)/ Map a function over a 'Text' that results in a 'Text', and
565−-- concatenate the results. Subject to fusion.
912+-- concatenate the results.
567−-- Note: if in 'concatMap' @f@ @t@, @f@ is defined in terms of fusible
568−-- functions, it will also be fusible.
569concatMap :: (Char -> Text) -> Text -> Text
913concatMap :: (Char -> Text) -> Text -> Text
570−concatMap f t = unstream (S.concatMap (stream . f) (stream t))
914+concatMap f = concat . foldr ((:) . f) []
571{-# INLINE concatMap #-}
915{-# INLINE concatMap #-}
573-- | /O(n)/ 'any' @p@ @t@ determines whether any character in the
917-- | /O(n)/ 'any' @p@ @t@ determines whether any character in the
574−-- 'Text' @t@ satisifes the predicate @p@. Subject to fusion.
918+-- 'Text' @t@ satisfies the predicate @p@. Subject to fusion.
575any :: (Char -> Bool) -> Text -> Bool
919any :: (Char -> Bool) -> Text -> Bool
576any p t = S.any p (stream t)
920any p t = S.any p (stream t)
577{-# INLINE any #-}
921{-# INLINE any #-}
579-- | /O(n)/ 'all' @p@ @t@ determines whether all characters in the
923-- | /O(n)/ 'all' @p@ @t@ determines whether all characters in the
580−-- 'Text' @t@ satisify the predicate @p@. Subject to fusion.
924+-- 'Text' @t@ satisfy the predicate @p@. Subject to fusion.
581all :: (Char -> Bool) -> Text -> Bool
925all :: (Char -> Bool) -> Text -> Bool
582all p t = S.all p (stream t)
926all p t = S.all p (stream t)
583{-# INLINE all #-}
927{-# INLINE all #-}
@@ -599,6 +943,7 @@ minimum t = S.minimum (stream t)
600-- | /O(n)/ 'scanl' is similar to 'foldl', but returns a list of
944-- | /O(n)/ 'scanl' is similar to 'foldl', but returns a list of
601-- successive reduced values from the left. Subject to fusion.
945-- successive reduced values from the left. Subject to fusion.
946+-- Performs replacement on invalid scalar values.
603-- > scanl f z [x1, x2, ...] == [z, z `f` x1, (z `f` x1) `f` x2, ...]
948-- > scanl f z [x1, x2, ...] == [z, z `f` x1, (z `f` x1) `f` x2, ...]
@@ -606,11 +951,13 @@ minimum t = S.minimum (stream t)
607-- > last (scanl f z xs) == foldl f z xs.
952-- > last (scanl f z xs) == foldl f z xs.
608scanl :: (Char -> Char -> Char) -> Char -> Text -> Text
953scanl :: (Char -> Char -> Char) -> Char -> Text -> Text
609−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)
610{-# INLINE scanl #-}
956{-# INLINE scanl #-}
612-- | /O(n)/ 'scanl1' is a variant of 'scanl' that has no starting
958-- | /O(n)/ 'scanl1' is a variant of 'scanl' that has no starting
613−-- value argument. Subject to fusion.
959+-- value argument. Subject to fusion. Performs replacement on
960+-- invalid scalar values.
615-- > scanl1 f [x1, x2, ...] == [x1, x1 `f` x2, ...]
962-- > scanl1 f [x1, x2, ...] == [x1, x1 `f` x2, ...]
616scanl1 :: (Char -> Char -> Char) -> Text -> Text
963scanl1 :: (Char -> Char -> Char) -> Text -> Text
@@ -618,48 +965,63 @@ scanl1 f t | null t = empty
618 | otherwise = scanl f (unsafeHead t) (unsafeTail t)
965 | otherwise = scanl f (unsafeHead t) (unsafeTail t)
619{-# INLINE scanl1 #-}
966{-# INLINE scanl1 #-}
621−-- | /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.
623-- > scanr f v == reverse . scanl (flip f) v . reverse
971-- > scanr f v == reverse . scanl (flip f) v . reverse
624scanr :: (Char -> Char -> Char) -> Char -> Text -> Text
972scanr :: (Char -> Char -> Char) -> Char -> Text -> Text
625−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)
626{-# INLINE scanr #-}
975{-# INLINE scanr #-}
628-- | /O(n)/ 'scanr1' is a variant of 'scanr' that has no starting
977-- | /O(n)/ 'scanr1' is a variant of 'scanr' that has no starting
629−-- value argument. Subject to fusion.
978+-- value argument. Subject to fusion. Performs replacement on
979+-- invalid scalar values.
630scanr1 :: (Char -> Char -> Char) -> Text -> Text
980scanr1 :: (Char -> Char -> Char) -> Text -> Text
631scanr1 f t | null t = empty
981scanr1 f t | null t = empty
632 | otherwise = scanr f (last t) (init t)
982 | otherwise = scanr f (last t) (init t)
633{-# INLINE scanr1 #-}
983{-# INLINE scanr1 #-}
635−-- | /O(n)/ Like a combination of 'map' and 'foldl'. Applies a
985+-- | /O(n)/ Like a combination of 'map' and 'foldl''. Applies a
636-- function to each element of a 'Text', passing an accumulating
986-- function to each element of a 'Text', passing an accumulating
637−-- 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.
638mapAccumL :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
989mapAccumL :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
639−mapAccumL f s t = case uncons t of
990+mapAccumL f z0 = S.mapAccumL g z0 . stream
640− Nothing -> (s, empty)
991+ where g a b = second safe (f a b)
641− Just (x, xs) -> (s'', cons y ys)
642− where (s', y ) = f s x
643− (s'',ys) = mapAccumL f s' xs
992+{-# INLINE mapAccumL #-}
645-- | The 'mapAccumR' function behaves like a combination of 'map' and
994-- | The 'mapAccumR' function behaves like a combination of 'map' and
646−-- 'foldr'; it applies a function to each element of a 'Text', passing
995+-- a strict 'foldr'; it applies a function to each element of a
647−-- an accumulating parameter from right to left, and returning a final
996+-- 'Text', passing an accumulating parameter from right to left, and
648−-- value of this accumulator together with the new 'Text'.
997+-- returning a final value of this accumulator together with the new
999+-- Performs replacement on invalid scalar values.
649mapAccumR :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
1000mapAccumR :: (a -> Char -> (a,Char)) -> a -> Text -> (a, Text)
650−mapAccumR f s t = case uncons t of
1001+mapAccumR f z0 = second reverse . S.mapAccumL g z0 . reverseStream
651− Nothing -> (s, empty)
1002+ where g a b = second safe (f a b)
652− Just (x, xs) -> (s'', cons y ys)
653− where (s'',y ) = f s' x
654− (s', ys) = mapAccumR f s xs
1003+{-# INLINE mapAccumR #-}
656-- -----------------------------------------------------------------------------
1005-- -----------------------------------------------------------------------------
657-- ** Generating and unfolding 'Text's
1006-- ** Generating and unfolding 'Text's
659-- | /O(n*m)/ 'replicate' @n@ @t@ is a 'Text' consisting of the input
1008-- | /O(n*m)/ 'replicate' @n@ @t@ is a 'Text' consisting of the input
660−-- @t@ repeated @n@ times. Subject to fusion.
1009+-- @t@ repeated @n@ times.
661replicate :: Int -> Text -> Text
1010replicate :: Int -> Text -> Text
662−replicate n t = unstream (S.replicateI (fromIntegral n) (S.stream t))
1011+replicate n t@(Text a o l)
1012+ | n <= 0 || l <= 0 = empty
1014+ | isSingleton t = replicateChar n (unsafeHead t)
1015+ | otherwise = Text (A.run x) 0 len
1018+ x :: ST s (A.MArray s)
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)
663{-# INLINE [1] replicate #-}
1025{-# INLINE [1] replicate #-}
665{-# RULES
1027{-# RULES
@@ -670,7 +1032,8 @@ replicate n t = unstream (S.replicateI (fromIntegral n) (S.stream t))
670-- | /O(n)/ 'replicateChar' @n@ @c@ is a 'Text' of length @n@ with @c@ the
1032-- | /O(n)/ 'replicateChar' @n@ @c@ is a 'Text' of length @n@ with @c@ the
671-- value of every element. Subject to fusion.
1033-- value of every element. Subject to fusion.
672replicateChar :: Int -> Char -> Text
1034replicateChar :: Int -> Char -> Text
673−replicateChar n c = unstream (S.replicateCharI n c)
1035+replicateChar n c = unstream (S.replicateCharI n (safe c))
1036+{-# INLINE replicateChar #-}
675-- | /O(n)/, where @n@ is the length of the result. The 'unfoldr'
1038-- | /O(n)/, where @n@ is the length of the result. The 'unfoldr'
676-- function is analogous to the List 'L.unfoldr'. 'unfoldr' builds a
1039-- function is analogous to the List 'L.unfoldr'. 'unfoldr' builds a
@@ -678,9 +1041,9 @@ replicateChar n c = unstream (S.replicateCharI n c)
678-- returns 'Nothing' if it is done producing the 'Text', otherwise
1041-- returns 'Nothing' if it is done producing the 'Text', otherwise
679-- 'Just' @(a,b)@. In this case, @a@ is the next 'Char' in the
1042-- 'Just' @(a,b)@. In this case, @a@ is the next 'Char' in the
680-- string, and @b@ is the seed value for further production. Subject
1043-- string, and @b@ is the seed value for further production. Subject
681−-- to fusion.
1044+-- to fusion. Performs replacement on invalid scalar values.
682unfoldr :: (a -> Maybe (Char,a)) -> a -> Text
1045unfoldr :: (a -> Maybe (Char,a)) -> a -> Text
683−unfoldr f s = unstream (S.unfoldr f s)
1046+unfoldr f s = unstream (S.unfoldr (firstf safe . f) s)
684{-# INLINE unfoldr #-}
1047{-# INLINE unfoldr #-}
686-- | /O(n)/ Like 'unfoldr', 'unfoldrN' builds a 'Text' from a seed
1049-- | /O(n)/ Like 'unfoldr', 'unfoldrN' builds a 'Text' from a seed
@@ -688,9 +1051,9 @@ unfoldr f s = unstream (S.unfoldr f s)
688-- first argument to 'unfoldrN'. This function is more efficient than
1051-- first argument to 'unfoldrN'. This function is more efficient than
689-- 'unfoldr' when the maximum length of the result is known and
1052-- 'unfoldr' when the maximum length of the result is known and
690-- correct, otherwise its performance is similar to 'unfoldr'. Subject
1053-- correct, otherwise its performance is similar to 'unfoldr'. Subject
691−-- to fusion.
1054+-- to fusion. Performs replacement on invalid scalar values.
692unfoldrN :: Int -> (a -> Maybe (Char,a)) -> a -> Text
1055unfoldrN :: Int -> (a -> Maybe (Char,a)) -> a -> Text
693−unfoldrN n f s = unstream (S.unfoldrN n f s)
1056+unfoldrN n f s = unstream (S.unfoldrN n (firstf safe . f) s)
694{-# INLINE unfoldrN #-}
1057{-# INLINE unfoldrN #-}
696-- -----------------------------------------------------------------------------
1059-- -----------------------------------------------------------------------------
@@ -703,14 +1066,16 @@ take :: Int -> Text -> Text
703take n t@(Text arr off len)
1066take n t@(Text arr off len)
704 | n <= 0 = empty
1067 | n <= 0 = empty
705 | n >= len = t
1068 | n >= len = t
706− | otherwise = Text arr off (loop 0 0)
1069+ | otherwise = text arr off (iterN n t)
709− | i >= len || cnt >= n = i
710− | otherwise = loop (i+d) (cnt+1)
712{-# INLINE [1] take #-}
1070{-# INLINE [1] take #-}
1072+iterN :: Int -> Text -> Int
1073+iterN n t@(Text _arr _off len) = loop 0 0
1075+ | i >= len || cnt >= n = i
1076+ | otherwise = loop (i+d) (cnt+1)
1077+ where d = iter_ t i
714{-# RULES
1079{-# RULES
715"TEXT take -> fused" [~1] forall n t.
1080"TEXT take -> fused" [~1] forall n t.
716 take n t = unstream (S.take n (stream t))
1081 take n t = unstream (S.take n (stream t))
@@ -718,18 +1083,36 @@ take n t@(Text arr off len)
718 unstream (S.take n (stream t)) = take n t
1083 unstream (S.take n (stream t)) = take n t
1086+-- | /O(n)/ 'takeEnd' @n@ @t@ returns the suffix remaining after
1087+-- taking @n@ characters from the end of @t@.
1091+-- > takeEnd 3 "foobar" == "bar"
1092+takeEnd :: Int -> Text -> Text
1093+takeEnd n t@(Text arr off len)
1096+ | otherwise = text arr (off+i) (len-i)
1097+ where i = iterNEnd n t
1099+iterNEnd :: Int -> Text -> Int
1100+iterNEnd n t@(Text _arr _off len) = loop (len-1) n
1104+ | otherwise = loop (i+d) (m-1)
1105+ where d = reverseIter_ t i
721-- | /O(n)/ 'drop' @n@, applied to a 'Text', returns the suffix of the
1107-- | /O(n)/ 'drop' @n@, applied to a 'Text', returns the suffix of the
722−-- 'Text' of length @n@, or the empty 'Text' if @n@ is greater than the
1108+-- 'Text' after the first @n@ characters, or the empty 'Text' if @n@
723−-- length of the 'Text'. Subject to fusion.
1109+-- is greater than the length of the 'Text'. Subject to fusion.
724drop :: Int -> Text -> Text
1110drop :: Int -> Text -> Text
725drop n t@(Text arr off len)
1111drop n t@(Text arr off len)
726 | n <= 0 = t
1112 | n <= 0 = t
727 | n >= len = empty
1113 | n >= len = empty
728− | otherwise = loop 0 0
1114+ | otherwise = text arr (off+i) (len-i)
730− | i >= len || cnt >= n = Text arr (off+i) (len-i)
731− | otherwise = loop (i+d) (cnt+1)
732− where d = iter_ t i
1115+ where i = iterN n t
733{-# INLINE [1] drop #-}
1116{-# INLINE [1] drop #-}
735{-# RULES
1118{-# RULES
@@ -739,6 +1122,18 @@ drop n t@(Text arr off len)
739 unstream (S.drop n (stream t)) = drop n t
1122 unstream (S.drop n (stream t)) = drop n t
1125+-- | /O(n)/ 'dropEnd' @n@ @t@ returns the prefix remaining after
1126+-- dropping @n@ characters from the end of @t@.
1130+-- > dropEnd 3 "foobar" == "foo"
1131+dropEnd :: Int -> Text -> Text
1132+dropEnd n t@(Text arr off len)
1135+ | otherwise = text arr off (iterNEnd n t)
742-- | /O(n)/ 'takeWhile', applied to a predicate @p@ and a 'Text',
1137-- | /O(n)/ 'takeWhile', applied to a predicate @p@ and a 'Text',
743-- returns the longest prefix (possibly empty) of elements that
1138-- returns the longest prefix (possibly empty) of elements that
744-- satisfy @p@. Subject to fusion.
1139-- satisfy @p@. Subject to fusion.
@@ -746,8 +1141,8 @@ takeWhile :: (Char -> Bool) -> Text -> Text
746takeWhile p t@(Text arr off len) = loop 0
1141takeWhile p t@(Text arr off len) = loop 0
747 where loop !i | i >= len = t
1142 where loop !i | i >= len = t
748 | p c = loop (i+d)
1143 | p c = loop (i+d)
749− | otherwise = textP arr off i
1144+ | otherwise = text arr off i
750− where (c,d) = iter t i
1145+ where Iter c d = iter t i
751{-# INLINE [1] takeWhile #-}
1146{-# INLINE [1] takeWhile #-}
753{-# RULES
1148{-# RULES
@@ -757,6 +1152,27 @@ takeWhile p t@(Text arr off len) = loop 0
757 unstream (S.takeWhile p (stream t)) = takeWhile p t
1152 unstream (S.takeWhile p (stream t)) = takeWhile p t
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.
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 #-}
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
760-- | /O(n)/ 'dropWhile' @p@ @t@ returns the suffix remaining after
1176-- | /O(n)/ 'dropWhile' @p@ @t@ returns the suffix remaining after
761-- 'takeWhile' @p@ @t@. Subject to fusion.
1177-- 'takeWhile' @p@ @t@. Subject to fusion.
762dropWhile :: (Char -> Bool) -> Text -> Text
1178dropWhile :: (Char -> Bool) -> Text -> Text
@@ -764,7 +1180,7 @@ dropWhile p t@(Text arr off len) = loop 0 0
764 where loop !i !l | l >= len = empty
1180 where loop !i !l | l >= len = empty
765 | p c = loop (i+d) (l+d)
1181 | p c = loop (i+d) (l+d)
766 | otherwise = Text arr (off+i) (len-l)
1182 | otherwise = Text arr (off+i) (len-l)
767− where (c,d) = iter t i
1183+ where Iter c d = iter t i
768{-# INLINE [1] dropWhile #-}
1184{-# INLINE [1] dropWhile #-}
770{-# RULES
1186{-# RULES
@@ -775,8 +1191,9 @@ dropWhile p t@(Text arr off len) = loop 0 0
777-- | /O(n)/ 'dropWhileEnd' @p@ @t@ returns the prefix remaining after
1193-- | /O(n)/ 'dropWhileEnd' @p@ @t@ returns the prefix remaining after
778−-- dropping characters that fail the predicate @p@ from the end of
1194+-- dropping characters that satisfy the predicate @p@ from the end of
779-- @t@. Subject to fusion.
1195-- @t@. Subject to fusion.
780-- Examples:
1197-- Examples:
782-- > dropWhileEnd (=='.') "foo..." == "foo"
1199-- > dropWhileEnd (=='.') "foo..." == "foo"
@@ -796,7 +1213,7 @@ dropWhileEnd p t@(Text arr off len) = loop (len-1) len
798-- | /O(n)/ 'dropAround' @p@ @t@ returns the substring remaining after
1215-- | /O(n)/ 'dropAround' @p@ @t@ returns the substring remaining after
799−-- dropping characters that fail the predicate @p@ from both the
1216+-- dropping characters that satisfy the predicate @p@ from both the
800-- beginning and end of @t@. Subject to fusion.
1217-- beginning and end of @t@. Subject to fusion.
801dropAround :: (Char -> Bool) -> Text -> Text
1218dropAround :: (Char -> Bool) -> Text -> Text
802dropAround p = dropWhile p . dropWhileEnd p
1219dropAround p = dropWhile p . dropWhileEnd p
@@ -831,28 +1248,20 @@ splitAt :: Int -> Text -> (Text, Text)
831splitAt n t@(Text arr off len)
1248splitAt n t@(Text arr off len)
832 | n <= 0 = (empty, t)
1249 | n <= 0 = (empty, t)
833 | n >= len = (t, empty)
1250 | n >= len = (t, empty)
834− | otherwise = (Text arr off k, Text arr (off+k) (len-k))
1251+ | otherwise = let k = iterN n t
835− where k = loop 0 0
1252+ in (text arr off k, text arr (off+k) (len-k))
837− | i >= len || cnt >= n = i
838− | otherwise = loop (i+d) (cnt+1)
840−{-# INLINE splitAt #-}
842−-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns a
1254+-- | /O(n)/ 'span', applied to a predicate @p@ and text @t@, returns
843−-- pair whose first element is the longest prefix (possibly empty) of
1255+-- a pair whose first element is the longest prefix (possibly empty)
844−-- @t@ of elements that satisfy @p@, and whose second is the remainder
1256+-- of @t@ of elements that satisfy @p@, and whose second is the
845−-- of the list.
1257+-- remainder of the list.
846span :: (Char -> Bool) -> Text -> (Text, Text)
1258span :: (Char -> Bool) -> Text -> (Text, Text)
847−span p t@(Text arr off len) = (textP arr off k, textP arr (off+k) (len-k))
1259+span p t = case span_ p t of
848− where k = loop 0
1260+ (# hd,tl #) -> (hd,tl)
849− loop !i | i >= len || not (p c) = i
850− | otherwise = loop (i+d)
851− where (c,d) = iter t i
852{-# INLINE span #-}
1261{-# INLINE span #-}
854−-- | /O(n)/ 'break' is like 'span', but the prefix returned is over
1263+-- | /O(n)/ 'break' is like 'span', but the prefix returned is
855−-- elements that fail the predicate @p@.
1264+-- over elements that fail the predicate @p@.
856break :: (Char -> Bool) -> Text -> (Text, Text)
1265break :: (Char -> Bool) -> Text -> (Text, Text)
857break p = span (not . p)
1266break p = span (not . p)
858{-# INLINE break #-}
1267{-# INLINE break #-}
@@ -864,7 +1273,7 @@ groupBy p = loop
864 loop t@(Text arr off len)
1273 loop t@(Text arr off len)
865 | null t = []
1274 | null t = []
866 | otherwise = text arr off n : loop (text arr (off+n) (len-n))
1275 | otherwise = text arr off n : loop (text arr (off+n) (len-n))
867− where (c,d) = iter t 0
1276+ where Iter c d = iter t 0
868 n = d + findAIndexOrEnd (not . p c) (Text arr (off+d) (len-d))
1277 n = d + findAIndexOrEnd (not . p c) (Text arr (off+d) (len-d))
870-- | Returns the /array/ index (in units of 'Word16') at which a
1279-- | Returns the /array/ index (in units of 'Word16') at which a
@@ -874,8 +1283,8 @@ findAIndexOrEnd :: (Char -> Bool) -> Text -> Int
874findAIndexOrEnd q t@(Text _arr _off len) = go 0
1283findAIndexOrEnd q t@(Text _arr _off len) = go 0
875 where go !i | i >= len || q c = i
1284 where go !i | i >= len || q c = i
876 | otherwise = go (i+d)
1285 | otherwise = go (i+d)
877− where (c,d) = iter t i
1286+ where Iter c d = iter t i
879-- | /O(n)/ Group characters in a string by equality.
1288-- | /O(n)/ Group characters in a string by equality.
880group :: Text -> [Text]
1289group :: Text -> [Text]
881group = groupBy (==)
1290group = groupBy (==)
@@ -899,107 +1308,59 @@ tails t | null t = [empty]
899-- copies to create substrings; they just construct new 'Text's that
1308-- copies to create substrings; they just construct new 'Text's that
900-- are slices of the original.
1309-- are slices of the original.
902−-- | /O(m*n)/ Break a 'Text' into pieces separated by the first
1311+-- | /O(m+n)/ Break a 'Text' into pieces separated by the first 'Text'
903−-- 'Text' argument, consuming the delimiter. An empty delimiter is
1312+-- argument (which cannot be empty), consuming the delimiter. An empty
904−-- invalid, and will cause an error to be raised.
1313+-- delimiter is invalid, and will cause an error to be raised.
906-- Examples:
1315-- Examples:
908−-- > split "\r\n" "a\r\nb\r\nd\r\ne" == ["a","b","d","e"]
1317+-- > splitOn "\r\n" "a\r\nb\r\nd\r\ne" == ["a","b","d","e"]
909−-- > split "aaa" "aaaXaaaXaaaXaaa" == ["","X","X","X",""]
1318+-- > splitOn "aaa" "aaaXaaaXaaaXaaa" == ["","X","X","X",""]
910−-- > split "x" "x" == ["",""]
1319+-- > splitOn "x" "x" == ["",""]
914−-- > intercalate s . split s == id
1323+-- > intercalate s . splitOn s == id
915−-- > split (singleton c) == splitWith (==c)
1324+-- > splitOn (singleton c) == split (==c)
916−split :: Text -- ^ Text to split on
1325+--
1326+-- (Note: the string @s@ to split on above cannot be empty.)
1328+-- In (unlikely) bad cases, this function's time complexity degrades
1329+-- towards /O(n*m)/.
1331+ -- ^ String to split on. If this string is empty, an error
917− -> Text -- ^ Input text
1334+ -- ^ Input text.
918− -> [Text]
1335+ -> [Text]
919−split pat src0
1336+splitOn pat@(Text _ _ l) src@(Text arr off len)
920− | null pat = emptyError "split"
1337+ | l <= 0 = emptyError "splitOn"
921− | l == 1 = splitWith (== (unsafeHead pat)) src0
1338+ | isSingleton pat = split (== unsafeHead pat) src
922− | otherwise = go src0
1339+ | otherwise = go 0 (indices pat src)
924− l = length pat
1341+ go !s (x:xs) = text arr (s+off) (x-s) : go (x+l) xs
925− go src = search 0 src
1342+ go s _ = [text arr (s+off) (len-s)]
928− | null s = [src] -- not found
929− | pat `isPrefixOf` s = take n src : go (drop l s)
930− | otherwise = search (n+1) (unsafeTail s)
931−{-# INLINE [1] split #-}
1343+{-# INLINE [1] splitOn #-}
933{-# RULES
1345{-# RULES
934−"TEXT split/singleton -> splitWith/==" [~1] forall c t.
1346+"TEXT splitOn/singleton -> split/==" [~1] forall c t.
935− split (singleton c) t = splitWith (==c) t
1347+ splitOn (singleton c) t = split (==c) t
938−-- | /O(m*n)/ Break a 'Text' into pieces at most @k@ times,
939−-- treating the first 'Text' argument as the delimiter to break on,
940−-- and consuming the delimiter. The last element of the list contains
941−-- the remaining text after the number of times to split has been
942−-- reached. A value of zero or less for @k@ causes no splitting to
943−-- occur. An empty delimiter is invalid, and will cause an error to be
948−-- > splitTimes 0 "//" "a//b//c" == ["a//b//c"]
949−-- > splitTimes 2 ":" "a:b:c:d:e" == ["a","b","c:d:e"]
950−-- > splitTimes 100 "???" "a????b" == ["a","?b"]
954−-- > intercalate s . splitTimes k s == id
955−splitTimes :: Int -- ^ Maximum number of times to split
956− -> Text -- ^ Text to split on
957− -> Text -- ^ Input text
959−splitTimes k pat src0
961− | null pat = emptyError "splitTimes"
962− | otherwise = go k src0
965− go !i src = search 0 src
968− | i == 0 || null s = [src] -- not found or limit reached
969− | pat `isPrefixOf` s = take n src : go (i-1) (drop l s)
970− | otherwise = search (n+1) (unsafeTail s)
971−{-# INLINE splitTimes #-}
973−-- | /O(m*n)/ Break a 'Text' into pieces at most @k@ times, like
974−-- 'splitTimes', but start from the end of the input and work towards
979−-- > splitTimes 2 "::" "a::b::c::d::e" == ["a","b","c::d::e"]
980−-- > splitTimesEnd 2 "::" "a::b::c::d::e" == ["a::b::c","d","e"]
981−splitTimesEnd :: Int -- ^ Maximum number of times to split
982− -> Text -- ^ Text to split on
983− -> Text -- ^ Input text
985−splitTimesEnd k pat src =
986− L.reverse . L.map reverse $ splitTimes k (reverse pat) (reverse src)
987−{-# INLINE splitTimesEnd #-}
989-- | /O(n)/ Splits a 'Text' into components delimited by separators,
1350-- | /O(n)/ Splits a 'Text' into components delimited by separators,
990-- where the predicate returns True for a separator element. The
1351-- where the predicate returns True for a separator element. The
991-- resulting components do not contain the separators. Two adjacent
1352-- resulting components do not contain the separators. Two adjacent
992-- separators result in an empty component in the output. eg.
1353-- separators result in an empty component in the output. eg.
994−-- > splitWith (=='a') "aabbaca" == ["","","bb","c",""]
1355+-- > split (=='a') "aabbaca" == ["","","bb","c",""]
995−-- > splitWith (=='a') "" == [""]
1356+-- > split (=='a') "" == [""]
996−splitWith :: (Char -> Bool) -> Text -> [Text]
1357+split :: (Char -> Bool) -> Text -> [Text]
997−splitWith _ t@(Text _off _arr 0) = [t]
1358+split _ t@(Text _off _arr 0) = [t]
998−splitWith p t = loop t
1359+split p t = loop t
999 where loop s | null s' = [l]
1360 where loop s | null s' = [l]
1000 | otherwise = l : loop (unsafeTail s')
1361 | otherwise = l : loop (unsafeTail s')
1001− where (l, s') = break p s
1362+ where (# l, s' #) = span_ (not . p) s
1002−{-# INLINE splitWith #-}
1363+{-# INLINE split #-}
1004-- | /O(n)/ Splits a 'Text' into components of length @k@. The last
1365-- | /O(n)/ Splits a 'Text' into components of length @k@. The last
1005-- element may be shorter than the other chunks, depending on the
1366-- element may be shorter than the other chunks, depending on the
@@ -1018,22 +1379,14 @@ chunksOf k = go
1018-- ----------------------------------------------------------------------------
1379-- ----------------------------------------------------------------------------
1019-- * Searching
1380-- * Searching
1021−-------------------------------------------------------------------------------
1022−-- ** Searching by equality
1024−-- | /O(n)/ 'elem' is the 'Text' membership predicate.
1025−elem :: Char -> Text -> Bool
1026−elem c t = S.elem c (stream t)
1029-------------------------------------------------------------------------------
1382-------------------------------------------------------------------------------
1030-- ** Searching with a predicate
1383-- ** Searching with a predicate
1032−-- | /O(n)/ The 'find' function takes a predicate and a 'Text',
1385+-- | /O(n)/ The 'find' function takes a predicate and a 'Text', and
1033−-- and returns the first element in matching the predicate, or 'Nothing'
1386+-- returns the first element matching the predicate, or 'Nothing' if
1034−-- if there is no such element.
1387+-- there is no such element.
1035find :: (Char -> Bool) -> Text -> Maybe Char
1388find :: (Char -> Bool) -> Text -> Maybe Char
1036−find p t = S.find p (stream t)
1389+find p t = S.findBy p (stream t)
1037{-# INLINE find #-}
1390{-# INLINE find #-}
1039-- | /O(n)/ The 'partition' function takes a predicate and a 'Text',
1392-- | /O(n)/ The 'partition' function takes a predicate and a 'Text',
@@ -1052,38 +1405,76 @@ filter :: (Char -> Bool) -> Text -> Text
1052filter p t = unstream (S.filter p (stream t))
1405filter p t = unstream (S.filter p (stream t))
1053{-# INLINE filter #-}
1406{-# INLINE filter #-}
1055−-- /O(n*m)/ Find all non-overlapping instances of @needle@ in
1408+-- | /O(n+m)/ Find the first instance of @needle@ (which must be
1056−-- @haystack@. The first element of the returned tuple is the prefix
1409+-- non-'null') in @haystack@. The first element of the returned tuple
1057−-- of @haystack@ before any matches of @needle@. The second element
1410+-- is the prefix of @haystack@ before @needle@ is matched. The second
1058−-- of the tuple is a list of non-overlapping substrings from
1411+-- is the remainder of @haystack@, starting with the match.
1059−-- @haystack@. Each begins with @needle@ and ends just before the
1060−-- following match. If @needle@ is not present in @haystack@, the
1061−-- list will be empty and the prefix will contain @haystack@.
1063-- Examples:
1413-- Examples:
1065−-- > findAll "::" "a::b::c" ==> ("a", ["::b", "::c"])
1415+-- > breakOn "::" "a::b::c" ==> ("a", "::b::c")
1066−-- > findAll "/" "foobar" ==> ("foobar", [])
1416+-- > breakOn "/" "foobar" ==> ("foobar", "")
1070−-- > concat (prefix : matches) == haystack
1420+-- > append prefix match == haystack
1071−-- > where (prefix, matches) = findAll needle haystack
1421+-- > where (prefix, match) = breakOn needle haystack
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.
1427+-- In (unlikely) bad cases, this function's time complexity degrades
1428+-- towards /O(n*m)/.
1072−findAll :: Text -> Text -> (Text, [Text])
1429+breakOn :: Text -> Text -> (Text, Text)
1073−findAll pat@(Text _ _ plen) src@(Text sarr soff slen)
1430+breakOn pat src@(Text arr off len)
1431+ | null pat = emptyError "breakOn"
1432+ | otherwise = case indices pat src of
1434+ (x:_) -> (text arr off x, text arr (off+x) (len-x))
1435+{-# INLINE breakOn #-}
1437+-- | /O(n+m)/ Similar to 'breakOn', but searches from the end of the
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.
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 #-}
1450+-- | /O(n+m)/ Find all non-overlapping instances of @needle@ in
1451+-- @haystack@. Each element of the returned list consists of a pair:
1453+-- * The entire string prior to the /k/th match (i.e. the prefix)
1455+-- * The /k/th match, followed by the remainder of the string
1459+-- > breakOnAll "::" ""
1461+-- > breakOnAll "/" "a/b/c/"
1462+-- > ==> [("a", "/b/c/"), ("a/b", "/c/"), ("a/b/c", "/")]
1464+-- In (unlikely) bad cases, this function's time complexity degrades
1465+-- towards /O(n*m)/.
1467+-- The @needle@ parameter may not be empty.
1468+breakOnAll :: Text -- ^ @needle@ to search for
1469+ -> Text -- ^ @haystack@ in which to search
1471+breakOnAll pat src@(Text arr off slen)
1074− | plen <= 0 = emptyError "findAll"
1472+ | null pat = emptyError "breakOnAll"
1075− | otherwise = (h,t)
1473+ | otherwise = L.map step (indices pat src)
1077− (h:t) = go 0 (search 0)
1475+ step x = (chunk 0 x, chunk x (slen-x))
1078− go k (x:xs) = Text sarr k (x-k) : go x xs
1476+ chunk !n !l = text arr (n+off) l
1081− | i >= slen = [slen] -- not found
1082− | pat `isPrefixOf` s = i : search (i+plen)
1083− | otherwise = search (i+d)
1084− where d = iter_ src i
1085− s = Text sarr (soff+i) (slen-i)
1086−{-# INLINE findAll #-}
1477+{-# INLINE breakOnAll #-}
1088-------------------------------------------------------------------------------
1479-------------------------------------------------------------------------------
1089-- ** Indexing 'Text's
1480-- ** Indexing 'Text's
@@ -1102,8 +1493,7 @@ findAll pat@(Text _ _ plen) src@(Text sarr soff slen)
1102-- For example, suppose you have a string that you want to split on
1493-- For example, suppose you have a string that you want to split on
1103-- the substring @\"::\"@, such as @\"foo::bar::quux\"@. Instead of
1494-- the substring @\"::\"@, such as @\"foo::bar::quux\"@. Instead of
1104-- searching for the index of @\"::\"@ and taking the substrings
1495-- searching for the index of @\"::\"@ and taking the substrings
1105−-- before and after that index, you would instead use @splitTimes 1
1496+-- before and after that index, you would instead use @breakOnAll \"::\"@.
1108-- | /O(n)/ 'Text' index (subscript) operator, starting from 0.
1498-- | /O(n)/ 'Text' index (subscript) operator, starting from 0.
1109index :: Text -> Int -> Char
1499index :: Text -> Int -> Char
@@ -1117,42 +1507,17 @@ findIndex :: (Char -> Bool) -> Text -> Maybe Int
1117findIndex p t = S.findIndex p (stream t)
1507findIndex p t = S.findIndex p (stream t)
1118{-# INLINE findIndex #-}
1508{-# INLINE findIndex #-}
1120−-- | The 'findIndices' function extends 'findIndex', by returning the
1121−-- indices of all elements satisfying the predicate, in ascending
1122−-- order. Subject to fusion.
1123−findIndices :: (Char -> Bool) -> Text -> [Int]
1124−findIndices p t = S.findIndices p (stream t)
1125−{-# INLINE findIndices #-}
1127−-- | /O(n)/ The 'elemIndex' function returns the index of the first
1128−-- element in the given 'Text' which is equal to the query element, or
1129−-- 'Nothing' if there is no such element. Subject to fusion.
1130−elemIndex :: Char -> Text -> Maybe Int
1131−elemIndex c t = S.elemIndex c (stream t)
1132−{-# INLINE elemIndex #-}
1134−-- | /O(n)/ The 'elemIndices' function returns the index of every
1135−-- element in the given 'Text' which is equal to the query
1136−-- element. Subject to fusion.
1137−elemIndices :: Char -> Text -> [Int]
1138−elemIndices c t = S.elemIndices c (stream t)
1139−{-# INLINE elemIndices #-}
1141−-- | /O(n*m)/ The 'count' function returns the number of times the
1510+-- | /O(n+m)/ The 'count' function returns the number of times the
1142-- query string appears in the given 'Text'. An empty query string is
1511-- query string appears in the given 'Text'. An empty query string is
1143-- invalid, and will cause an error to be raised.
1512-- invalid, and will cause an error to be raised.
1514+-- In (unlikely) bad cases, this function's time complexity degrades
1515+-- towards /O(n*m)/.
1144count :: Text -> Text -> Int
1516count :: Text -> Text -> Int
1145−count pat src0
1517+count pat src
1146− | null pat = emptyError "count"
1518+ | null pat = emptyError "count"
1147− | l == 1 = countChar (unsafeHead pat) src0
1519+ | isSingleton pat = countChar (unsafeHead pat) src
1148− | otherwise = go 0 src0
1520+ | otherwise = L.length (indices pat src)
1151− go !n src = search src
1153− search s | null s = n
1154− | pat `isPrefixOf` s = go (n+1) (drop l s)
1155− | otherwise = search (unsafeTail s)
1156{-# INLINE [1] count #-}
1521{-# INLINE [1] count #-}
1158{-# RULES
1523{-# RULES
@@ -1175,13 +1540,15 @@ countChar c t = S.countChar c (stream t)
1175-- equivalent to a pair of 'unpack' operations.
1540-- equivalent to a pair of 'unpack' operations.
1176zip :: Text -> Text -> [(Char,Char)]
1541zip :: Text -> Text -> [(Char,Char)]
1177zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
1542zip a b = S.unstreamList $ S.zipWith (,) (stream a) (stream b)
1178−{-# INLINE [0] zip #-}
1543+{-# INLINE zip #-}
1180-- | /O(n)/ 'zipWith' generalises 'zip' by zipping with the function
1545-- | /O(n)/ 'zipWith' generalises 'zip' by zipping with the function
1181-- given as the first argument, instead of a tupling function.
1546-- given as the first argument, instead of a tupling function.
1547+-- Performs replacement on invalid scalar values.
1182zipWith :: (Char -> Char -> Char) -> Text -> Text -> Text
1548zipWith :: (Char -> Char -> Char) -> Text -> Text -> Text
1183−zipWith f t1 t2 = unstream (S.zipWith f (stream t1) (stream t2))
1549+zipWith f t1 t2 = unstream (S.zipWith g (stream t1) (stream t2))
1550+ where g a b = safe (f a b)
1184−{-# INLINE [0] zipWith #-}
1551+{-# INLINE zipWith #-}
1186-- | /O(n)/ Breaks a 'Text' up into a list of words, delimited by 'Char's
1553-- | /O(n)/ Breaks a 'Text' up into a list of words, delimited by 'Char's
1187-- representing white space.
1554-- representing white space.
@@ -1197,7 +1564,7 @@ words t@(Text arr off len) = loop 0 0
1197 then loop (start+1) (start+1)
1564 then loop (start+1) (start+1)
1198 else Text arr (start+off) (n-start) : loop (n+d) (n+d)
1565 else Text arr (start+off) (n-start) : loop (n+d) (n+d)
1199 | otherwise = loop start (n+d)
1566 | otherwise = loop start (n+d)
1200− where (c,d) = iter t n
1567+ where Iter c d = iter t n
1201{-# INLINE words #-}
1568{-# INLINE words #-}
1203-- | /O(n)/ Breaks a 'Text' up into a list of 'Text's at
1570-- | /O(n)/ Breaks a 'Text' up into a list of 'Text's at
@@ -1207,9 +1574,10 @@ lines ps | null ps = []
1207 | otherwise = h : if null t
1574 | otherwise = h : if null t
1209 else lines (unsafeTail t)
1576 else lines (unsafeTail t)
1210− where (h,t) = span (/= '\n') ps
1577+ where (# h,t #) = span_ (/= '\n') ps
1211{-# INLINE lines #-}
1578{-# INLINE lines #-}
1213-- | /O(n)/ Portably breaks a 'Text' up into a list of 'Text's at line
1581-- | /O(n)/ Portably breaks a 'Text' up into a list of 'Text's at line
1214-- boundaries.
1582-- boundaries.
@@ -1217,7 +1585,6 @@ lines ps | null ps = []
1217-- return immediately followed by a line feed, or a carriage return.
1585-- return immediately followed by a line feed, or a carriage return.
1218-- This accounts for both Unix and Windows line ending conventions,
1586-- This accounts for both Unix and Windows line ending conventions,
1219-- and for the old convention used on Mac OS 9 and earlier.
1587-- and for the old convention used on Mac OS 9 and earlier.
1221lines' :: Text -> [Text]
1588lines' :: Text -> [Text]
1222lines' ps | null ps = []
1589lines' ps | null ps = []
1223 | otherwise = h : case uncons t of
1590 | otherwise = h : case uncons t of
@@ -1244,8 +1611,7 @@ unwords = intercalate (singleton ' ')
1244{-# INLINE unwords #-}
1611{-# INLINE unwords #-}
1246-- | /O(n)/ The 'isPrefixOf' function takes two 'Text's and returns
1613-- | /O(n)/ The 'isPrefixOf' function takes two 'Text's and returns
1247−-- 'True' iff the first is a prefix of the second. This function is
1614+-- 'True' iff the first is a prefix of the second. Subject to fusion.
1248−-- subject to fusion.
1249isPrefixOf :: Text -> Text -> Bool
1615isPrefixOf :: Text -> Text -> Bool
1250isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
1616isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
1251 alen <= blen && S.isPrefixOf (stream a) (stream b)
1617 alen <= blen && S.isPrefixOf (stream a) (stream b)
@@ -1254,8 +1620,6 @@ isPrefixOf a@(Text _ _ alen) b@(Text _ _ blen) =
1254{-# RULES
1620{-# RULES
1255"TEXT isPrefixOf -> fused" [~1] forall s t.
1621"TEXT isPrefixOf -> fused" [~1] forall s t.
1256 isPrefixOf s t = S.isPrefixOf (stream s) (stream t)
1622 isPrefixOf s t = S.isPrefixOf (stream s) (stream t)
1257−"TEXT isPrefixOf -> unfused" [1] forall s t.
1258− S.isPrefixOf (stream s) (stream t) = isPrefixOf s t
1261-- | /O(n)/ The 'isSuffixOf' function takes two 'Text's and returns
1625-- | /O(n)/ The 'isSuffixOf' function takes two 'Text's and returns
@@ -1268,13 +1632,124 @@ isSuffixOf a@(Text _aarr _aoff alen) b@(Text barr boff blen) =
1268 | otherwise = Text barr (boff+d) alen
1632 | otherwise = Text barr (boff+d) alen
1269{-# INLINE isSuffixOf #-}
1633{-# INLINE isSuffixOf #-}
1271−-- | /O(n)/ The 'isInfixOf' function takes two 'Text's and returns
1635+-- | /O(n+m)/ The 'isInfixOf' function takes two 'Text's and returns
1272-- 'True' iff the first is contained, wholly and intact, anywhere
1636-- 'True' iff the first is contained, wholly and intact, anywhere
1273-- within the second.
1637-- within the second.
1639+-- In (unlikely) bad cases, this function's time complexity degrades
1640+-- towards /O(n*m)/.
1274isInfixOf :: Text -> Text -> Bool
1641isInfixOf :: Text -> Text -> Bool
1275−isInfixOf needle haystack = L.any (isPrefixOf needle) (tails haystack)
1642+isInfixOf needle haystack
1276−{-# INLINE isInfixOf #-}
1643+ | null needle = True
1277−-- TODO: a better implementation
1644+ | isSingleton needle = S.elem (unsafeHead needle) . S.stream $ haystack
1645+ | otherwise = not . L.null . indices needle $ haystack
1646+{-# INLINE [1] isInfixOf #-}
1649+"TEXT isInfixOf/singleton -> S.elem/S.stream" [~1] forall n h.
1650+ isInfixOf (singleton n) h = S.elem n (S.stream h)
1653+-------------------------------------------------------------------------------
1656+-- | /O(n)/ Return the suffix of the second string if its prefix
1657+-- matches the entire first string.
1661+-- > stripPrefix "foo" "foobar" == Just "bar"
1662+-- > stripPrefix "" "baz" == Just "baz"
1663+-- > stripPrefix "foo" "quux" == Nothing
1665+-- This is particularly useful with the @ViewPatterns@ extension to
1668+-- > {-# LANGUAGE ViewPatterns #-}
1669+-- > import Data.Text as T
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
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
1683+-- If the strings do not have a common prefix or either one is empty,
1684+-- this function returns 'Nothing'.
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
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
1702+-- | /O(n)/ Return the prefix of the second string if its suffix
1703+-- matches the entire first string.
1707+-- > stripSuffix "bar" "foobar" == Just "foo"
1708+-- > stripSuffix "" "baz" == Just "baz"
1709+-- > stripSuffix "foo" "quux" == Nothing
1711+-- This is particularly useful with the @ViewPatterns@ extension to
1714+-- > {-# LANGUAGE ViewPatterns #-}
1715+-- > import Data.Text as T
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
1725+-- | Add a list of non-negative numbers. Errors out on overflow.
1726+sumP :: String -> [Int] -> Int
1729+ | ax >= 0 = go ax xs
1730+ | otherwise = overflowError fun
1279emptyError :: String -> a
1734emptyError :: String -> a
1280−emptyError fun = P.error ("Data.Text." ++ fun ++ ": empty input")
1735+emptyError fun = P.error $ "Data.Text." ++ fun ++ ": empty input"
1737+overflowError :: String -> a
1738+overflowError fun = P.error $ "Data.Text." ++ fun ++ ": size overflow"
1740+-- | /O(n)/ Make a distinct copy of the given string, sharing no
1741+-- storage with the original string.
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
1751+ go :: ST s (A.MArray s)
1754+ A.copyI marr 0 arr off len