@@ -1,10 +1,3 @@
1−{-# OPTIONS_GHC -Wno-deprecations #-}
2−
3−#undef FUSION_CHECK
4−#ifdef FUSION_CHECK
5−{-# OPTIONS_GHC -ddump-simpl -ddump-to-file -dsuppress-all #-}
6−#endif
7−
81-- |
92-- Module : Streamly.Benchmark.Data.Parser
103-- Copyright : (c) 2020 Composewell Technologies
@@ -14,42 +7,43 @@
147
158{-# LANGUAGE CPP #-}
169{-# LANGUAGE FlexibleContexts #-}
17−{-# LANGUAGE ScopedTypeVariables #-}
18−{-# OPTIONS_GHC -Wno-orphans #-}
1910
2011module Main
2112 (
2213 main
2314 ) where
2415
25−import Control.Applicative ((<|>))
2616import Control.DeepSeq (NFData(..))
27−import Data.Function ((&))
17+import Control.Monad.Catch (MonadCatch)
18+import Data.Foldable (asum)
2819import Data.Functor (($>))
20+import Data.Maybe (fromMaybe)
2921import Data.Monoid (Sum(..))
3022import GHC.Magic (inline)
23+#if __GLASGOW_HASKELL__ >= 802
3124import GHC.Magic (noinline)
25+#else
26+#define noinline
27+#endif
3228import System.IO (Handle)
3329import System.Random (randomRIO)
34−import Streamly.Internal.Data.Fold (Fold(..))
35−import Streamly.Internal.Data.Parser
36− (ParseError(..), Parser(..), Initial(..), Step(..))
37−import Streamly.Internal.Data.Stream (Stream)
38−import Prelude hiding
39− (any, all, take, sequence, sequence_, sequenceA, takeWhile, dropWhile, span)
30+import Prelude
31+ hiding (any, all, take, sequence, sequence_, sequenceA, takeWhile)
4032
41−import qualified Control.Applicative as AP
42−import qualified Data.Foldable as F
4333import qualified Data.Traversable as TR
34+import qualified Data.Foldable as F
35+import qualified Control.Applicative as AP
4436import qualified Streamly.FileSystem.Handle as Handle
45−import qualified Streamly.Internal.Data.Array as Array
46−import qualified Streamly.Internal.Data.Fold as Fold
37+import qualified Streamly.Prelude as S
38+import qualified Streamly.Internal.Data.Array.Foreign as Array
39+import qualified Streamly.Internal.Data.Fold as FL
4740import qualified Streamly.Internal.Data.Parser as PR
48−import qualified Streamly.Data.Stream as Stream
41+import qualified Streamly.Internal.Data.Stream.IsStream as IP
4942import qualified Streamly.Internal.Data.Producer as Producer
50−import qualified Streamly.Internal.Data.Stream as Stream
43+import qualified Streamly.Internal.Data.Producer.Source as Source
5144
52−import Test.Tasty.Bench hiding (env)
45+import Gauge hiding (env)
46+import Streamly.Prelude (SerialT)
5347import Streamly.Benchmark.Common
5448import Streamly.Benchmark.Common.Handle
5549
@@ -60,59 +54,19 @@ import Streamly.Benchmark.Common.Handle
6054-- XXX these can be moved to the common module
6155
6256{-# INLINE sourceUnfoldrM #-}
63−sourceUnfoldrM :: Monad m => Int -> Int -> Stream m Int
64−sourceUnfoldrM value n = Stream.unfoldrM step n
57+sourceUnfoldrM :: (S.IsStream t, S.MonadAsync m) => Int -> Int -> t m Int
58+sourceUnfoldrM value n = S.unfoldrM step n
6559 where
6660 step cnt =
6761 if cnt > n + value
6862 then return Nothing
6963 else return (Just (cnt, cnt + 1))
7064
71−-- | Generates something like this: { { \{ \{ } }. The stream consists of
72−-- three parts, the first part is contains a sequence of `{`. The second part
73−-- contains a sequence pf escaped values `\{`. The third part contains a
74−-- sequence of `}`.
75−{-# INLINE sourceEscapedFrames #-}
76−sourceEscapedFrames ::
77− Monad m
78− => Int
79− -> Int
80− -> Stream m Char
81−sourceEscapedFrames value n = Stream.unfoldrM step n
82− where
83−
84− bs = '\\'
85− cbOpen = '{'
86− cbClose = '}'
87− value1 = value `div` 4
88−
89− step cnt
90− | cnt > 4 * value1 = return Nothing
91− | cnt <= value1 = return $ Just (cbOpen, cnt + 1)
92− | cnt > 3 * value1 = return $ Just (cbClose, cnt + 1)
93− | otherwise =
94− return
95− $ Just
96− $ if (cnt - value1) `mod` 2 == 1
97− then (bs, cnt + 1)
98− else (cbOpen, cnt + 1)
99−
100−{-# INLINE benchIOSrc #-}
101−benchIOSrc
102− :: NFData b
103− => (Int -> Int -> Stream IO a)
104− -> Int
105− -> String
106− -> (Stream IO a -> IO b)
107− -> Benchmark
108−benchIOSrc src value name f =
109− bench name $ nfIO $ randomRIO (1,1) >>= f . src value
110−
11165-- | Takes a fold method, and uses it with a default source.
11266{-# INLINE benchIOSink #-}
11367benchIOSink
114− :: NFData b
115− => Int -> String -> (Stream IO Int -> IO b) -> Benchmark
68+ :: (S.IsStream t, NFData b)
69+ => Int -> String -> (t IO Int -> IO b) -> Benchmark
11670benchIOSink value name f =
11771 bench name $ nfIO $ randomRIO (1,1) >>= f . sourceUnfoldrM value
11872
@@ -121,451 +75,188 @@ benchIOSink value name f =
12175-------------------------------------------------------------------------------
12276
12377{-# INLINE takeBetween #-}
124−takeBetween :: Monad m => Int -> Stream m a -> m (Either ParseError ())
125−takeBetween value = Stream.parse (PR.takeBetween 0 value Fold.drain)
78+takeBetween :: MonadCatch m => Int -> SerialT m a -> m ()
79+takeBetween value = IP.parse (PR.takeBetween 0 value FL.drain)
12680
12781{-# INLINE takeEQ #-}
128−takeEQ :: Monad m => Int -> Stream m a -> m (Either ParseError ())
129−takeEQ value = Stream.parse (PR.takeEQ value Fold.drain)
130−
131−{-# INLINE takeGE #-}
132−takeGE :: Monad m => Int -> Stream m a -> m (Either ParseError ())
133−takeGE value = Stream.parse (PR.takeGE value Fold.drain)
134−
135−{-# INLINE dropWhile #-}
136−dropWhile :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
137−dropWhile value = Stream.parse (PR.dropWhile (<= value))
82+takeEQ :: MonadCatch m => Int -> SerialT m a -> m ()
83+takeEQ value = IP.parse (PR.takeEQ value FL.drain)
13884
139−{-# INLINE takeStartBy #-}
140−takeStartBy :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
141−takeStartBy value stream = do
142− let stream2 = value `Stream.cons` stream
143− Stream.parse (PR.takeStartBy (== value) Fold.drain) stream2
85+{-# INLINE drainWhile #-}
86+drainWhile :: MonadCatch m => Int -> SerialT m Int -> m ()
87+drainWhile value = IP.parse (PR.drainWhile (<= value))
14488
145−takeFramedByEsc_ :: Monad m => Int -> Stream m Char -> m (Either ParseError ())
146−takeFramedByEsc_ _ = Stream.parse parser
147−
148− where
149−
150− isEsc = (== '\\')
151− isBegin = (== '{')
152− isEnd = (== '}')
153−
154− parser = PR.takeFramedByEsc_ isEsc isBegin isEnd Fold.drain
155−
156−{-# INLINE listEqBy #-}
157−listEqBy :: Int -> Stream IO Int -> IO (Either ParseError [Int])
158−listEqBy len = Stream.parse (PR.listEqBy (==) [1 .. len])
159−
160−{-# INLINE streamEqBy #-}
161−streamEqBy :: Int -> Stream IO Int -> IO (Either ParseError ())
162−streamEqBy len = Stream.parse (PR.streamEqBy (==) (Stream.enumerateFromTo 1 len))
89+{-# INLINE sliceBeginWith #-}
90+sliceBeginWith :: MonadCatch m => Int -> SerialT m Int -> m ()
91+sliceBeginWith value stream = do
92+ stream1 <- return . fromMaybe (S.fromPure (value + 1)) =<< S.tail stream
93+ let stream2 = value `S.cons` stream1
94+ IP.parse (PR.sliceBeginWith (== value) FL.drain) stream2
16395
16496{-# INLINE takeWhile #-}
165−takeWhile :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
166−takeWhile value = Stream.parse (PR.takeWhile (<= value) Fold.drain)
167−
168−takeWhileP :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
169−takeWhileP value =
170− Stream.parse (PR.takeWhileP (<= value) (PR.takeWhile (<= value - 1) Fold.drain))
97+takeWhile :: MonadCatch m => Int -> SerialT m Int -> m ()
98+takeWhile value = IP.parse (PR.takeWhile (<= value) FL.drain)
17199
172100{-# INLINE takeP #-}
173−takeP :: Monad m => Int -> Stream m a -> m (Either ParseError ())
174−takeP value = Stream.parse (PR.takeP value (PR.fromFold Fold.drain))
101+takeP :: MonadCatch m => Int -> SerialT m a -> m ()
102+takeP value = IP.parse (PR.takeP value (PR.fromFold FL.drain))
175103
176104{-# INLINE groupBy #-}
177−groupBy :: Monad m => Stream m Int -> m (Either ParseError ())
178−groupBy = Stream.parse (PR.groupBy (<=) Fold.drain)
105+groupBy :: MonadCatch m => SerialT m Int -> m ()
106+groupBy = IP.parse (PR.groupBy (<=) FL.drain)
179107
180108{-# INLINE groupByRolling #-}
181−groupByRolling :: Monad m => Stream m Int -> m (Either ParseError ())
182−groupByRolling = Stream.parse (PR.groupByRolling (<=) Fold.drain)
109+groupByRolling :: MonadCatch m => SerialT m Int -> m ()
110+groupByRolling = IP.parse (PR.groupByRolling (<=) FL.drain)
183111
184112{-# INLINE wordBy #-}
185−wordBy :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
186−wordBy value = Stream.parse (PR.wordBy (>= value) Fold.drain)
187−
188−{-# INLINE sepByWords #-}
189−sepByWords :: Monad m => Stream m Int -> m (Either ParseError ())
190−sepByWords = Stream.parse (wrds even Fold.drain)
191− where
192− wrds p = PR.sepBy (PR.takeWhile (not . p) Fold.drain) (PR.dropWhile p)
193−
194−{-# INLINE sepByAllWords #-}
195−sepByAllWords :: Monad m => Stream m Int -> m (Either ParseError ())
196−sepByAllWords = Stream.parse (wrds even Fold.drain)
197− where
198− wrds p = PR.sepByAll (PR.takeWhile (not . p) Fold.drain) (PR.dropWhile p)
199−
200−-- Returning a list to compare with the sepBy1 in ParserK
201−{-# INLINE sepBy1 #-}
202−sepBy1 :: Monad m => Stream m Int -> m (Either ParseError [Int])
203−sepBy1 xs = do
204− Stream.parse (PR.sepBy1 (PR.satisfy odd) (PR.satisfy even) Fold.toList) xs
205−
206−{-# INLINE sepByWords1 #-}
207−sepByWords1 :: Monad m => Stream m Int -> m (Either ParseError ())
208−sepByWords1 = Stream.parse (wrds even Fold.drain)
209− where
210− wrds p = PR.sepBy1 (PR.takeWhile (not . p) Fold.drain) (PR.dropWhile p)
211−
212−{-# INLINE deintercalate #-}
213−deintercalate :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
214−deintercalate _ = Stream.parse (partition even)
215−
216− where
217−
218− partition p =
219− PR.deintercalate
220− (PR.takeWhile (not . p) Fold.sum) (PR.takeWhile p Fold.sum) Fold.drain
221−
222−{-# INLINE deintercalate1 #-}
223−deintercalate1 :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
224−deintercalate1 _ = Stream.parse (partition even)
225−
226− where
227−
228− partition p =
229− PR.deintercalate1
230− (PR.takeWhile (not . p) Fold.sum) (PR.takeWhile p Fold.sum) Fold.drain
231−
232−{-# INLINE deintercalateAll #-}
233−deintercalateAll :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
234−deintercalateAll _ = Stream.parse (partition even)
235−
236− where
237−
238− partition p =
239− PR.deintercalateAll
240− (PR.takeWhile (not . p) Fold.sum) (PR.takeWhile p Fold.sum) Fold.drain
113+wordBy :: MonadCatch m => Int -> SerialT m Int -> m ()
114+wordBy value = IP.parse (PR.wordBy (>= value) FL.drain)
241115
242116{-# INLINE manyWordByEven #-}
243−manyWordByEven :: Monad m => Stream m Int -> m (Either ParseError ())
244−manyWordByEven = Stream.parse (PR.many (PR.wordBy even Fold.drain) Fold.drain)
117+manyWordByEven :: MonadCatch m => SerialT m Int -> m ()
118+manyWordByEven = IP.parse (PR.many (PR.wordBy (Prelude.even) FL.drain) FL.drain)
245119
246120{-# INLINE many #-}
247−many :: Monad m => Stream m Int -> m (Either ParseError Int)
248−many = Stream.parse (PR.many (PR.satisfy (> 0)) Fold.length)
121+many :: MonadCatch m => SerialT m Int -> m Int
122+many = IP.parse (PR.many (PR.satisfy (> 0)) FL.length)
249123
250124{-# INLINE manyAlt #-}
251−manyAlt :: Monad m => Stream m Int -> m Int
125+manyAlt :: MonadCatch m => SerialT m Int -> m Int
252126manyAlt xs = do
253− x <- Stream.parse (AP.many (PR.satisfy (> 0))) xs
127+ x <- IP.parse (AP.many (PR.satisfy (> 0))) xs
254128 return $ Prelude.length x
255129
256130{-# INLINE some #-}
257−some :: Monad m => Stream m Int -> m (Either ParseError Int)
258−some = Stream.parse (PR.some (PR.satisfy (> 0)) Fold.length)
131+some :: MonadCatch m => SerialT m Int -> m Int
132+some = IP.parse (PR.some (PR.satisfy (> 0)) FL.length)
259133
260134{-# INLINE someAlt #-}
261−someAlt :: Monad m => Stream m Int -> m Int
135+someAlt :: MonadCatch m => SerialT m Int -> m Int
262136someAlt xs = do
263− x <- Stream.parse (AP.some (PR.satisfy (> 0))) xs
137+ x <- IP.parse (AP.some (PR.satisfy (> 0))) xs
264138 return $ Prelude.length x
265139
266140{-# INLINE manyTill #-}
267−manyTill :: Monad m => Int -> Stream m Int -> m (Either ParseError Int)
141+manyTill :: MonadCatch m => Int -> SerialT m Int -> m Int
268142manyTill value =
269− Stream.parse (PR.manyTill (PR.satisfy (> 0)) (PR.satisfy (== value)) Fold.length)
143+ IP.parse (PR.manyTill (PR.satisfy (> 0)) (PR.satisfy (== value)) FL.length)
270144
271−{-# INLINE splitAp2 #-}
272−splitAp2 :: Monad m
273− => Int -> Stream m Int -> m (Either ParseError ((), ()))
274−splitAp2 value =
275− Stream.parse
145+{-# INLINE splitAp #-}
146+splitAp :: MonadCatch m
147+ => Int -> SerialT m Int -> m ((), ())
148+splitAp value =
149+ IP.parse
276150 ((,)
277− <$> PR.dropWhile (<= (value `div` 2))
278− <*> PR.dropWhile (<= value)
279− )
280−
281−{-# INLINE splitAp4 #-}
282−splitAp4 :: Monad m
283− => Int -> Stream m Int -> m (Either ParseError ())
284−splitAp4 value =
285− Stream.parse
286− ( (\() () () () -> ())
287− <$> PR.dropWhile (<= (value * 1 `div` 4))
288− <*> PR.dropWhile (<= (value * 2 `div` 4))
289− <*> PR.dropWhile (<= (value * 3 `div` 4))
290− <*> PR.dropWhile (<= value)
291− )
292−
293−{-# INLINE splitAp8 #-}
294−splitAp8 :: Monad m
295− => Int -> Stream m Int -> m (Either ParseError ())
296−splitAp8 value =
297− Stream.parse
298− ( (\() () () () () () () () -> ())
299− <$> PR.dropWhile (<= (value * 1 `div` 8))
300− <*> PR.dropWhile (<= (value * 2 `div` 8))
301− <*> PR.dropWhile (<= (value * 3 `div` 8))
302− <*> PR.dropWhile (<= (value * 4 `div` 8))
303− <*> PR.dropWhile (<= (value * 5 `div` 8))
304− <*> PR.dropWhile (<= (value * 6 `div` 8))
305− <*> PR.dropWhile (<= (value * 7 `div` 8))
306− <*> PR.dropWhile (<= value)
151+ <$> PR.drainWhile (<= (value `div` 2))
152+ <*> PR.drainWhile (<= value)
307153 )
308154
309155{-# INLINE splitApBefore #-}
310−splitApBefore :: Monad m
311− => Int -> Stream m Int -> m (Either ParseError ())
156+splitApBefore :: MonadCatch m
157+ => Int -> SerialT m Int -> m ()
312158splitApBefore value =
313− Stream.parse
314− ( PR.dropWhile (<= (value `div` 2))
315− *> PR.dropWhile (<= value)
159+ IP.parse
160+ ( PR.drainWhile (<= (value `div` 2))
161+ *> PR.drainWhile (<= value)
316162 )
317163
318164{-# INLINE splitApAfter #-}
319−splitApAfter :: Monad m
320− => Int -> Stream m Int -> m (Either ParseError ())
165+splitApAfter :: MonadCatch m
166+ => Int -> SerialT m Int -> m ()
321167splitApAfter value =
322− Stream.parse
323− ( PR.dropWhile (<= (value `div` 2))
324− <* PR.dropWhile (<= value)
168+ IP.parse
169+ ( PR.drainWhile (<= (value `div` 2))
170+ <* PR.drainWhile (<= value)
325171 )
326172
327−{-# INLINE splitWith2 #-}
328−splitWith2 :: Monad m
329− => Int -> Stream m Int -> m (Either ParseError ((), ()))
330−splitWith2 value =
331− Stream.parse
332− (PR.splitWith (,)
333− (PR.dropWhile (<= (value `div` 2)))
334− (PR.dropWhile (<= value))
173+{-# INLINE serialWith #-}
174+serialWith :: MonadCatch m
175+ => Int -> SerialT m Int -> m ((), ())
176+serialWith value =
177+ IP.parse
178+ (PR.serialWith (,)
179+ (PR.drainWhile (<= (value `div` 2)))
180+ (PR.drainWhile (<= value))
335181 )
336182
337183{-# INLINE split_ #-}
338−split_ :: Monad m
339− => Int -> Stream m Int -> m (Either ParseError ())
184+split_ :: MonadCatch m
185+ => Int -> SerialT m Int -> m ()
340186split_ value =
341− Stream.parse
187+ IP.parse
342188 (PR.split_
343− (PR.dropWhile (<= (value `div` 2)))
344− (PR.dropWhile (<= value))
189+ (PR.drainWhile (<= (value `div` 2)))
190+ (PR.drainWhile (<= value))
345191 )
346192
347−-- XXX dropWhile with applicative does not fuse
348−-- PR.dropWhile (<= (value * 1 `div` 4)) *> PR.die "alt"
349−{-# INLINE takeWhileFail #-}
350−takeWhileFail :: Monad m => (a -> Bool) -> Fold m a b -> Parser a m b
351−takeWhileFail predicate (Fold fstep finitial _ ffinal) =
352− Parser step initial extract
353−
354− where
193+{-# INLINE sliceSepByP #-}
194+sliceSepByP :: MonadCatch m
195+ => Int -> SerialT m Int -> m()
196+sliceSepByP value = IP.parse (PR.sliceSepByP (>= value) (PR.fromFold FL.drain))
355197
356− initial = do
357− res <- finitial
358− return $ case res of
359− Fold.Partial s -> IPartial s
360− Fold.Done b -> IDone b
361−
362− step s a =
363− if predicate a
364− then do
365− fres <- fstep s a
366− return
367− $ case fres of
368− Fold.Partial s1 -> Partial 0 s1
369− Fold.Done b -> Done 0 b
370− else return $ Error "fail"
371−
372− extract s = fmap (Done 0) (ffinal s)
373−
374−{-# INLINE alt2 #-}
375−alt2 :: Monad m
376− => Int -> Stream m Int -> m (Either ParseError ())
377−alt2 value =
378− Stream.parse
379− (PR.alt
380− (takeWhileFail (<= (value `div` 2)) Fold.drain)
381− (PR.dropWhile (<= value))
382− )
383−
384−{-# INLINE alt4 #-}
385−alt4 :: Monad m
386− => Int -> Stream m Int -> m (Either ParseError ())
387−alt4 value =
388− Stream.parse
389− ( takeWhileFail (<= (value * 1 `div` 4)) Fold.drain
390− <|> takeWhileFail (<= (value * 2 `div` 4)) Fold.drain
391− <|> takeWhileFail (<= (value * 3 `div` 4)) Fold.drain
392− <|> PR.dropWhile (<= value)
393− )
394−
395−{-# INLINE alt8 #-}
396−alt8 :: Monad m
397− => Int -> Stream m Int -> m (Either ParseError ())
398−alt8 value =
399− Stream.parse
400− ( takeWhileFail (<= (value * 1 `div` 8)) Fold.drain
401− <|> takeWhileFail (<= (value * 2 `div` 8)) Fold.drain
402− <|> takeWhileFail (<= (value * 3 `div` 8)) Fold.drain
403− <|> takeWhileFail (<= (value * 4 `div` 8)) Fold.drain
404− <|> takeWhileFail (<= (value * 5 `div` 8)) Fold.drain
405− <|> takeWhileFail (<= (value * 6 `div` 8)) Fold.drain
406− <|> takeWhileFail (<= (value * 7 `div` 8)) Fold.drain
407− <|> PR.dropWhile (<= value)
408− )
409−
410−{-# INLINE alt16 #-}
411−alt16 :: Monad m
412− => Int -> Stream m Int -> m (Either ParseError ())
413−alt16 value =
414− Stream.parse
415− ( takeWhileFail (<= (value * 1 `div` 16)) Fold.drain
416− <|> takeWhileFail (<= (value * 2 `div` 16)) Fold.drain
417− <|> takeWhileFail (<= (value * 3 `div` 16)) Fold.drain
418− <|> takeWhileFail (<= (value * 4 `div` 16)) Fold.drain
419− <|> takeWhileFail (<= (value * 5 `div` 16)) Fold.drain
420− <|> takeWhileFail (<= (value * 6 `div` 16)) Fold.drain
421− <|> takeWhileFail (<= (value * 8 `div` 16)) Fold.drain
422− <|> takeWhileFail (<= (value * 9 `div` 16)) Fold.drain
423− <|> takeWhileFail (<= (value * 10 `div` 16)) Fold.drain
424− <|> takeWhileFail (<= (value * 11 `div` 16)) Fold.drain
425− <|> takeWhileFail (<= (value * 12 `div` 16)) Fold.drain
426− <|> takeWhileFail (<= (value * 13 `div` 16)) Fold.drain
427− <|> takeWhileFail (<= (value * 14 `div` 16)) Fold.drain
428− <|> takeWhileFail (<= (value * 15 `div` 16)) Fold.drain
429− <|> PR.dropWhile (<= value)
430− )
431−
432−{-# INLINE altSmall #-}
433−altSmall :: Monad m
434− => Int -> Stream m Int -> m ()
435−altSmall value =
436− Stream.fold Fold.drain .
437− Stream.parseMany
438− (PR.alt
439− (PR.satisfy (>= value) *> PR.die "alt")
440− (PR.satisfy (<= value))
441− )
442−
443−{-# INLINE monad #-}
444−monad :: Monad m
445− => Int -> Stream m Int -> m (Either ParseError ())
446−monad value =
447− Stream.parse
448− $ do
449− PR.dropWhile (<= (value `div` 2))
450− PR.dropWhile (<= value)
451−
452−{-# INLINE monad4 #-}
453−monad4 :: Monad m
454− => Int -> Stream m Int -> m (Either ParseError ())
455−monad4 value =
456− Stream.parse $ do
457− PR.dropWhile (<= (value `div` 4))
458− PR.dropWhile (<= (value `div` 2))
459− PR.dropWhile (<= (value * 3 `div` 4))
460− PR.dropWhile (<= value)
461−
462−{-# INLINE monad8 #-}
463−monad8 :: Monad m
464− => Int -> Stream m Int -> m (Either ParseError ())
465−monad8 value =
466− Stream.parse $ do
467− PR.dropWhile (<= (value * 1 `div` 8))
468− PR.dropWhile (<= (value * 2 `div` 8))
469− PR.dropWhile (<= (value * 3 `div` 8))
470− PR.dropWhile (<= (value * 4 `div` 8))
471− PR.dropWhile (<= (value * 5 `div` 8))
472− PR.dropWhile (<= (value * 6 `div` 8))
473− PR.dropWhile (<= (value * 7 `div` 8))
474− PR.dropWhile (<= value)
475−
476−{-# INLINE monad16 #-}
477−monad16 :: Monad m
478− => Int -> Stream m Int -> m (Either ParseError ())
479−monad16 value =
480− Stream.parse $ do
481− PR.dropWhile (<= (value * 1 `div` 16))
482− PR.dropWhile (<= (value * 2 `div` 16))
483− PR.dropWhile (<= (value * 3 `div` 16))
484− PR.dropWhile (<= (value * 4 `div` 16))
485− PR.dropWhile (<= (value * 5 `div` 16))
486− PR.dropWhile (<= (value * 6 `div` 16))
487− PR.dropWhile (<= (value * 7 `div` 16))
488− PR.dropWhile (<= (value * 8 `div` 16))
489− PR.dropWhile (<= (value * 9 `div` 16))
490− PR.dropWhile (<= (value * 10 `div` 16))
491− PR.dropWhile (<= (value * 11 `div` 16))
492− PR.dropWhile (<= (value * 12 `div` 16))
493− PR.dropWhile (<= (value * 13 `div` 16))
494− PR.dropWhile (<= (value * 14 `div` 16))
495− PR.dropWhile (<= (value * 15 `div` 16))
496− PR.dropWhile (<= value)
497−
498−{-# INLINE takeEndBy_ #-}
499−takeEndBy_ :: Monad m
500− => Int -> Stream m Int -> m (Either ParseError ())
501−takeEndBy_ value = Stream.parse (PR.takeEndBy_ (>= value) (PR.fromFold Fold.drain))
502−
503−{-
504198{-# INLINE teeAllAny #-}
505−teeAllAny :: Monad m
506− => Int -> Stream m Int -> m ((), ())
199+teeAllAny :: MonadCatch m
200+ => Int -> SerialT m Int -> m ((), ())
507201teeAllAny value =
508− Stream.parse
202+ IP.parse
509203 (PR.teeWith (,)
510− (PR.dropWhile (<= value))
511− (PR.dropWhile (<= value))
204+ (PR.drainWhile (<= value))
205+ (PR.drainWhile (<= value))
512206 )
513207
514208{-# INLINE teeFstAllAny #-}
515−teeFstAllAny :: Monad m
516− => Int -> Stream m Int -> m ((), ())
209+teeFstAllAny :: MonadCatch m
210+ => Int -> SerialT m Int -> m ((), ())
517211teeFstAllAny value =
518− Stream.parse
212+ IP.parse
519213 (PR.teeWithFst (,)
520− (PR.dropWhile (<= value))
521− (PR.dropWhile (<= value))
214+ (PR.drainWhile (<= value))
215+ (PR.drainWhile (<= value))
522216 )
523217
524218{-# INLINE shortestAllAny #-}
525−shortestAllAny :: Monad m
526− => Int -> Stream m Int -> m (Either ParseError ())
219+shortestAllAny :: MonadCatch m
220+ => Int -> SerialT m Int -> m ()
527221shortestAllAny value =
528− Stream.parse
222+ IP.parse
529223 (PR.shortest
530− (PR.dropWhile (<= value))
531− (PR.dropWhile (<= value))
224+ (PR.drainWhile (<= value))
225+ (PR.drainWhile (<= value))
532226 )
533227
534228{-# INLINE longestAllAny #-}
535−longestAllAny :: Monad m
536− => Int -> Stream m Int -> m (Either ParseError ())
229+longestAllAny :: MonadCatch m
230+ => Int -> SerialT m Int -> m ()
537231longestAllAny value =
538− Stream.parse
232+ IP.parse
539233 (PR.longest
540− (PR.dropWhile (<= value))
541− (PR.dropWhile (<= value))
234+ (PR.drainWhile (<= value))
235+ (PR.drainWhile (<= value))
542236 )
543−-}
237+
238+parseManyChunksOfSum :: Int -> Handle -> IO Int
239+parseManyChunksOfSum n inh =
240+ S.length
241+ $ IP.parseMany
242+ (PR.fromFold $ FL.take n FL.sum)
243+ (S.unfold Handle.read inh)
544244
545245-------------------------------------------------------------------------------
546−-- Spanning
246+-- Parsing with unfolds
547247-------------------------------------------------------------------------------
548248
549−{-# INLINE span #-}
550−span :: Monad m => Int -> Stream m Int -> m (Either ParseError ((), ()))
551−span value = Stream.parse (PR.span (<= (value `div` 2)) Fold.drain Fold.drain)
552−
553−{-# INLINE spanBy #-}
554−spanBy :: Monad m => Int -> Stream m Int -> m (Either ParseError ((), ()))
555−spanBy value =
556− Stream.parse (PR.spanBy (\_ i -> i <= (value `div` 2)) Fold.drain Fold.drain)
557−
558−{-# INLINE spanByRolling #-}
559−spanByRolling :: Monad m => Int -> Stream m Int -> m (Either ParseError ((), ()))
560−spanByRolling value =
561− Stream.parse (PR.spanByRolling (\_ i -> i <= value `div` 2) Fold.drain Fold.drain)
562−
563−parseManyChunksOfSum :: Int -> Handle -> IO Int
564−parseManyChunksOfSum n inh =
565− Stream.fold Fold.length
566− $ Stream.parseMany
567− (PR.fromFold $ Fold.take n Fold.sum)
568− (Stream.unfold Handle.reader inh)
249+{-# INLINE parseManyUnfoldArrays #-}
250+parseManyUnfoldArrays :: Int -> [Array.Array Int] -> IO ()
251+parseManyUnfoldArrays count arrays = do
252+ let src = Source.source (Just (Producer.OuterLoop arrays))
253+ let parser = PR.fromFold (FL.take count FL.drain)
254+ let readSrc =
255+ Source.producer
256+ $ Producer.concat Producer.fromList Array.producer
257+ let streamParser =
258+ Producer.simplify (Source.parseMany parser readSrc)
259+ S.drain $ S.unfold streamParser src
569260
570261-------------------------------------------------------------------------------
571262-- Parsers in which -fspec-constr-recursive=16 is problematic
@@ -576,129 +267,63 @@ parseManyChunksOfSum n inh =
576267-- not have to rely on it.
577268--
578269{-# INLINE lookAhead #-}
579−lookAhead :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
270+lookAhead :: MonadCatch m => Int -> SerialT m Int -> m ()
580271lookAhead value =
581− Stream.parse (PR.lookAhead (PR.takeWhile (<= value) Fold.drain) $> ())
272+ IP.parse (PR.lookAhead (PR.takeWhile (<= value) FL.drain) $> ())
582273
583−-- XXX The timing of this increased 3x after the stepify extract changes.
584−{-# INLINE sequenceA_ #-}
585−sequenceA_ :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
586−sequenceA_ value =
587− Stream.parse (F.sequenceA_ $ replicate value (PR.satisfy (> 0)))
588−
589−-- quadratic complexity
590274{-# INLINE sequenceA #-}
591−sequenceA :: Monad m => Int -> Stream m Int -> m Int
275+sequenceA :: MonadCatch m => Int -> SerialT m Int -> m Int
592276sequenceA value xs = do
593− x <- Stream.parse (TR.sequenceA (replicate value (PR.satisfy (> 0)))) xs
277+ x <- IP.parse (TR.sequenceA (replicate value (PR.satisfy (> 0)))) xs
594278 return $ length x
595279
596−-- quadratic complexity
280+{-# INLINE sequenceA_ #-}
281+sequenceA_ :: MonadCatch m => Int -> SerialT m Int -> m ()
282+sequenceA_ value =
283+ IP.parse (F.sequenceA_ $ replicate value (PR.satisfy (> 0)))
284+
597285{-# INLINE sequence #-}
598−sequence :: Monad m => Int -> Stream m Int -> m Int
286+sequence :: MonadCatch m => Int -> SerialT m Int -> m Int
599287sequence value xs = do
600− x <- Stream.parse (TR.sequence (replicate value (PR.satisfy (> 0)))) xs
288+ x <- IP.parse (TR.sequence (replicate value (PR.satisfy (> 0)))) xs
601289 return $ length x
602290
603291{-# INLINE sequence_ #-}
604−sequence_ :: Monad m => Int -> Stream m Int -> m (Either ParseError ())
605−sequence_ value xs =
606− Stream.parse (foldr f (return ()) (replicate value (PR.takeBetween 0 1 Fold.drain))) xs
292+sequence_ :: MonadCatch m => Int -> SerialT m Int -> m ()
293+sequence_ value =
294+ IP.parse (F.sequence_ $ replicate value (PR.satisfy (> 0)))
607295
608− where
609−
610− {-# INLINE f #-}
611− f m k = m >>= (\_ -> k)
612−
613−-- choice using the "Alternative" instance with direct style parser type has
614−-- quadratic performance complexity.
615−--
616296{-# INLINE choiceAsum #-}
617−choiceAsum :: Monad m => Int -> Stream m Int -> m (Either ParseError Int)
297+choiceAsum :: MonadCatch m => Int -> SerialT m Int -> m Int
618298choiceAsum value =
619− Stream.parse (F.asum (replicate value (PR.satisfy (< 0)))
299+ IP.parse (asum (replicate value (PR.satisfy (< 0)))
620300 AP.<|> PR.satisfy (> 0))
621301
622−{-
623302{-# INLINE choice #-}
624−choice :: Monad m => Int -> Stream m Int -> m (Either ParseError Int)
303+choice :: MonadCatch m => Int -> SerialT m Int -> m Int
625304choice value =
626− Stream.parse
305+ IP.parse
627306 (PR.choice (replicate value (PR.satisfy (< 0))) AP.<|> PR.satisfy (> 0))
628−-}
629−
630−-------------------------------------------------------------------------------
631−-- Parsing with unfolds
632−-------------------------------------------------------------------------------
633−
634−{-# INLINE parseManyUnfoldArrays #-}
635−parseManyUnfoldArrays :: Int -> [Array.Array Int] -> IO ()
636−parseManyUnfoldArrays count arrays = do
637− let src = Producer.source (Just (Producer.OuterLoop arrays))
638− let parser = PR.fromFold (Fold.take count Fold.drain)
639− let readSrc =
640− Producer.producer
641− $ Producer.concat Producer.fromList Array.producer
642− let streamParser =
643− Producer.simplify (Producer.parseMany parser readSrc)
644− Stream.fold Fold.drain $ Stream.unfold streamParser src
645307
646308-------------------------------------------------------------------------------
647309-- Stream transformation
648310-------------------------------------------------------------------------------
649311
650312{-# INLINE parseMany #-}
651−parseMany :: Monad m => Int -> Stream m Int -> m ()
313+parseMany :: MonadCatch m => Int -> SerialT m Int -> m ()
652314parseMany n =
653− Stream.fold Fold.drain
654− . fmap getSum
655− . Stream.catRights . Stream.parseMany (PR.fromFold $ Fold.take n Fold.mconcat)
656− . fmap Sum
657−
658−{-# INLINE parseManyGroupBy #-}
659−parseManyGroupBy :: Monad m => (Int -> Int -> Bool) -> Stream m Int -> m ()
660−parseManyGroupBy cmp =
661− Stream.fold Fold.drain . Stream.parseMany (PR.groupBy cmp Fold.drain)
662−
663−{-# INLINE parseManyGroupsRolling #-}
664−parseManyGroupsRolling :: Monad m => Bool -> Stream m Int -> m ()
665−parseManyGroupsRolling b =
666− Stream.fold Fold.drain
667− . Stream.parseMany (PR.groupByRolling (\_ _ -> b) Fold.drain)
668−
669−{-# INLINE parseManyGroupsRollingEither #-}
670−parseManyGroupsRollingEither :: Monad m =>
671− (Int -> Int -> Bool) -> Int -> m ()
672−parseManyGroupsRollingEither cmp value = do
673− sourceUnfoldrM value 1
674− & Stream.parseMany (PR.groupByRollingEither cmp Fold.drain Fold.drain)
675− & Stream.fold Fold.drain
676−
677−{-# INLINE parseManyGroupsRollingEitherAlt #-}
678−parseManyGroupsRollingEitherAlt :: Monad m =>
679− (Int -> Int -> Bool) -> Int -> m ()
680−parseManyGroupsRollingEitherAlt cmp value = do
681− sourceUnfoldrM value 1
682− -- Make the input unsorted.
683− & fmap (\x -> if even x then x + 2 else x)
684− & Stream.parseMany (PR.groupByRollingEither cmp Fold.drain Fold.drain)
685− & Stream.fold Fold.drain
315+ S.drain
316+ . S.map getSum
317+ . IP.parseMany (PR.fromFold $ FL.take n FL.mconcat)
318+ . S.map Sum
686319
687320{-# INLINE parseIterate #-}
688−parseIterate :: Monad m => Int -> Stream m Int -> m ()
321+parseIterate :: MonadCatch m => Int -> SerialT m Int -> m ()
689322parseIterate n =
690− Stream.fold Fold.drain
691− . fmap getSum
692− . Stream.catRights
693− . Stream.parseIterate
694− (PR.fromFold . Fold.take n . Fold.sconcat)
695− (Sum 0)
696− . fmap Sum
697−
698−{-# INLINE concatSequence #-}
699−concatSequence :: Monad m => Stream m Int -> m (Either ParseError ())
700−concatSequence =
701− Stream.parse $ PR.sequence (Stream.repeat PR.one) Fold.drain
323+ S.drain
324+ . S.map getSum
325+ . IP.parseIterate (PR.fromFold . FL.take n . FL.sconcat) (Sum 0)
326+ . S.map Sum
702327
703328-------------------------------------------------------------------------------
704329-- Benchmarks
@@ -707,95 +332,41 @@ concatSequence =
707332moduleName :: String
708333moduleName = "Data.Parser"
709334
710−instance NFData ParseError where
711− {-# INLINE rnf #-}
712− rnf (ParseError x) = rnf x
713−
714335o_1_space_serial :: Int -> [Benchmark]
715336o_1_space_serial value =
716337 [ benchIOSink value "takeBetween" $ takeBetween value
338+ , benchIOSink value "takeEQ" $ takeEQ value
717339 , benchIOSink value "takeWhile" $ takeWhile value
718− , benchIOSink value "takeWhileP" $ takeWhileP value
719340 , benchIOSink value "takeP" $ takeP value
720− , benchIOSink value "dropWhile" $ dropWhile value
721− , benchIOSink value "takeStartBy" $ takeStartBy value
722− , benchIOSink value "takeEndBy_" $ takeEndBy_ value
341+ , benchIOSink value "drainWhile" $ drainWhile value
342+ , benchIOSink value "sliceBeginWith" $ sliceBeginWith value
723343 , benchIOSink value "groupBy" $ groupBy
724344 , benchIOSink value "groupByRolling" $ groupByRolling
725345 , benchIOSink value "wordBy" $ wordBy value
726− , benchIOSink value "sepBy (words)" sepByWords
727− , benchIOSink value "sepByAll (words)" sepByAllWords
728− , benchIOSink value "sepBy1 (words)" sepByWords1
729− , benchIOSink value "deintercalate" $ deintercalate value
730− , benchIOSink value "deintercalate1" $ deintercalate1 value
731− , benchIOSink value "deintercalateAll" $ deintercalateAll value
732− -- Applicative and Monad
733− , benchIOSink value "splitAp2" $ splitAp2 value
734− , benchIOSink value "splitAp4" $ splitAp4 value
735− , benchIOSink value "splitAp8" $ splitAp8 value
346+ , benchIOSink value "splitAp" $ splitAp value
736347 , benchIOSink value "splitApBefore" $ splitApBefore value
737348 , benchIOSink value "splitApAfter" $ splitApAfter value
738− , benchIOSink value "splitWith2" $ splitWith2 value
739− , benchIOSink value "span" $ span value
740− , benchIOSink value "spanBy" $ spanBy value
741− , benchIOSink value "spanByRolling" $ spanByRolling value
742− , benchIOSink value "monad2" $ monad value
743− , benchIOSink value "monad4" $ monad4 value
744− , benchIOSink value "monad8" $ monad8 value
745− -- Alternative
746− , benchIOSink value "alt2parseMany" $ altSmall value
747− , benchIOSink value "alt2" $ alt2 value
748− , benchIOSink value "alt4" $ alt4 value
749− , benchIOSink value "alt8" $ alt8 value
750− , benchIOSink value "alt16" $ alt16 value
349+ , benchIOSink value "serialWith" $ serialWith value
350+ , benchIOSink value "sliceSepByP" $ sliceSepByP value
751351 , benchIOSink value "many" many
752352 , benchIOSink value "many (wordBy even)" $ manyWordByEven
753353 , benchIOSink value "some" some
754354 , benchIOSink value "manyTill" $ manyTill value
755− , benchIOSink value "parseMany" $ parseMany value
756− , benchIOSink value "parseMany (take 1)" (parseMany 1)
757− , benchIOSink value "parseMany (take all)" (parseMany value)
758− , benchIOSink value "parseMany (groupBy (<))" (parseManyGroupBy (<))
759− , benchIOSink value "parseMany (groupBy (==))" (parseManyGroupBy (==))
760− , benchIOSink value "parseMany groupRollingBy (bound groups)"
761− $ parseManyGroupsRolling False
762− , benchIOSink value "parseMany groupRollingBy (1 group)"
763− $ parseManyGroupsRolling True
764− , bench "parseMany groupRollingByEither (Left)"
765− $ nfIO $ parseManyGroupsRollingEitherLeft
766− , bench "parseMany groupRollingByEither (Right)"
767− $ nfIO $ parseManyGroupsRollingEitherRight
768− , bench "parseMany groupRollingByEither (Alternating)"
769− $ nfIO $ parseManyGroupsRollingEitherAlt1
770− , benchIOSink value "parseIterate (take 1)" (parseIterate 1)
771− , benchIOSink value "parseIterate (take all)" (parseIterate value)
772− , benchIOSink value "concatSequence" concatSequence
773− {-
774355 , benchIOSink value "tee" $ teeAllAny value
775356 , benchIOSink value "teeFst" $ teeFstAllAny value
776357 , benchIOSink value "shortest" $ shortestAllAny value
777358 , benchIOSink value "longest" $ longestAllAny value
778− -}
779− , benchIOSink value "streamEqBy" (streamEqBy value)
359+ , benchIOSink value "parseMany (take 1)" (parseMany 1)
360+ , benchIOSink value "parseMany (take all)" (parseMany value)
361+ , benchIOSink value "parseIterate (take 1)" (parseIterate 1)
362+ , benchIOSink value "parseIterate (take all)" (parseIterate value)
780363 ]
781364
782− where
783−
784− {-# NOINLINE parseManyGroupsRollingEitherLeft #-}
785− parseManyGroupsRollingEitherLeft = parseManyGroupsRollingEither (<) value
786−
787− {-# NOINLINE parseManyGroupsRollingEitherRight #-}
788− parseManyGroupsRollingEitherRight = parseManyGroupsRollingEither (>) value
789−
790− {-# NOINLINE parseManyGroupsRollingEitherAlt1 #-}
791− parseManyGroupsRollingEitherAlt1 =
792− parseManyGroupsRollingEitherAlt (>) value
793−
794365o_1_space_filesystem :: BenchEnv -> [Benchmark]
795366o_1_space_filesystem env =
796− [ mkBench ("S.parseMany (Fold.take " ++ show (bigSize env) ++ " Fold.sum)") env
367+ [ mkBench ("S.parseMany (FL.take " ++ show (bigSize env) ++ " FL.sum)") env
797368 $ \inh _ -> noinline parseManyChunksOfSum (bigSize env) inh
798− , mkBench "S.parseMany (Fold.take 1 Fold.sum)" env
369+ , mkBench "S.parseMany (FL.take 1 FL.sum)" env
799370 $ \inh _ -> inline parseManyChunksOfSum 1 inh
800371 ]
801372
@@ -810,38 +381,26 @@ o_1_space_serial_unfold bound arrays =
810381o_n_heap_serial :: Int -> [Benchmark]
811382o_n_heap_serial value =
812383 [
813− benchIOSink value "takeEQ" $ takeEQ value
814− , benchIOSink value "takeGE" $ takeGE value
815−
816384 -- lookahead benchmark holds the entire input till end
817− , benchIOSink value "lookAhead" $ lookAhead value
385+ benchIOSink value "lookAhead" $ lookAhead value
818386
819− -- o-n-heap because of backtracking
820− , benchIOSrc sourceEscapedFrames value "takeFramedByEsc_"
821− $ takeFramedByEsc_ value
387+ -- accumulates the results in a list
388+ , benchIOSink value "sequence" $ sequence value
389+ , benchIOSink value "sequenceA" $ sequenceA value
822390
391+ -- XXX why should this take O(n) heap, it discards the results?
392+ , benchIOSink value "sequence_" $ sequence_ value
393+ , benchIOSink value "sequenceA_" $ sequenceA_ value
823394 -- non-linear time complexity (parserD)
824395 , benchIOSink value "split_" $ split_ value
825− -- XXX Takes lot of space when run on a long stream, why?
826− , benchIOSink value "monad16" $ monad16 value
396+ -- XXX why O(n) heap?
397+ , benchIOSink value "choice (asum)" $ choiceAsum value
398+ , benchIOSink value "choice" $ choice value
827399
828400 -- These show non-linear time complexity.
829401 -- They accumulate the results in a list.
830− , benchIOSink value "sepBy1" sepBy1
831402 , benchIOSink value "manyAlt" manyAlt
832403 , benchIOSink value "someAlt" someAlt
833− , benchIOSink value "listEqBy" (listEqBy value)
834− ]
835−
836−-- accumulate results in a list in IO
837−o_n_space_serial :: Int -> [Benchmark]
838−o_n_space_serial value =
839− [ benchIOSink value "sequenceA/100" $ sequenceA (value `div` 100)
840− , benchIOSink value "sequenceA_/100" $ sequenceA_ (value `div` 100)
841− , benchIOSink value "sequence/100" $ sequence (value `div` 100)
842− , benchIOSink value "sequence_/100" $ sequence_ (value `div` 100)
843− , benchIOSink value "choice (asum)/100" $ choiceAsum (value `div` 100)
844− -- , benchIOSink value "choice/100" $ choice (value `div` 100)
845404 ]
846405
847406-------------------------------------------------------------------------------
@@ -850,13 +409,12 @@ o_n_space_serial value =
850409
851410main :: IO ()
852411main = do
853−#ifndef FUSION_CHECK
854412 env <- mkHandleBenchEnv
855413 runWithCLIOptsEnv defaultStreamSize alloc (allBenchmarks env)
856414
857415 where
858416
859− alloc value = Stream.fold Fold.toList $ Stream.chunksOf 100 $ sourceUnfoldrM value 0
417+ alloc value = IP.toList $ IP.arraysOf 100 $ sourceUnfoldrM value 0
860418
861419 allBenchmarks env arrays value =
862420 [ bgroup (o_1_space_prefix moduleName) (o_1_space_serial value)
@@ -866,22 +424,4 @@ main = do
866424 , bgroup (o_1_space_prefix moduleName)
867425 (o_1_space_serial_unfold value arrays)
868426 , bgroup (o_n_heap_prefix moduleName) (o_n_heap_serial value)
869− , bgroup (o_n_space_prefix moduleName) (o_n_space_serial value)
870427 ]
871−#else
872− -- Enable FUSION_CHECK macro at the beginning of the file
873− -- Enable one benchmark below, and run the benchmark
874− -- Check the .dump-simpl output
875− let value = 100000
876− -- let input = sourceUnfoldrM value 1
877− -- manyTill value input
878− -- deintercalate value input
879− -- deintercalate1 value input
880− -- deintercalateAll value input
881− -- sepByWords input
882− -- sepByAllWords input
883− -- sepBy1 input
884− -- sepByWords1 input
885− takeFramedByEsc_ value (sourceEscapedFrames value 1)
886− return ()
887−#endif