on 98259cc help sign in
5 files changed+62−305
since splitsnapshots
unifiedsplit
Data/Syntax.hs+12−104 Data/Syntax/Char.hs−8 Data/Syntax/Combinator.hs+11−149 Data/Syntax/Indent.hs+19−22 syntax.cabal+20−22
Data/Syntax.hs+12−104
@@ -10,28 +10,22 @@ License : MIT
1010Maintainer : Paweł Nowak <pawel834@gmail.com>
1111Stability : experimental
1212
13−Reversible parsing and pretty-printing.
13+Abstract syntax descriptions based on semi-isomorphisms.
1414-}
1515module Data.Syntax (
1616 -- * Syntax.
1717 Syntax(..),
18− Isolable(..),
1918 -- * Common isomorphisms.
2019 packed
2120 ) where
2221
23−import Prelude hiding (take, takeWhile, id, (.))
22+import Prelude hiding (take, takeWhile)
2423
25−import Control.Category
26−import Control.Category.Reader
27−import Control.Category.Structures
28−import Control.Lens.Iso
29−import Control.Lens.SemiIso
30−import Control.SIArrow
31−import Data.MonoTraversable
32−import Data.Sequences hiding (take, takeWhile, replicate)
33−import qualified Data.Vector as V
34−import qualified Data.Vector.Unboxed as VU
24+import Control.Lens.Iso
25+import Control.Lens.SemiIso
26+import Control.SIArrow
27+import Data.MonoTraversable
28+import Data.Sequences hiding (take, takeWhile)
3529
3630-- | An isomorphism between a sequence and a list of its elements.
3731packed :: IsSequence seq => Iso' seq [Element seq]
@@ -39,11 +33,11 @@ packed = iso otoList fromList
3933
4034-- | An abstract syntax description based on semi-isomorphisms.
4135--
42−-- This class can be implemented by both parsers and printers.
43−--
36+-- This class can be implemented by both parsers and printers (and maybe more?).
37+--
4438-- The usual use is to write a polymorphic syntax description and instantiate it
45−-- both as a parser and a printer. Examples are available in 'syntax-example' and
46−-- 'syntax-example-json' packages.
39+-- both as a parser and a printer. An example is available in the 'syntax-example'
40+-- package.
4741--
4842-- Methods of this class try to mimic "Data.Attoparsec.Text" interface.
4943class ( SIArrow syn
@@ -76,7 +70,7 @@ class ( SIArrow syn
7670 satisfyWith ai p = bifiltered p . ai /$/ anyChar
7771
7872 -- | A specific string.
79− string :: Seq syn -> syn () ()
73+ string :: (Seq syn) -> syn () ()
8074 string s = rev (exact s) /$/ take (olength s)
8175
8276 -- | A string of length @n@.
@@ -99,90 +93,4 @@ class ( SIArrow syn
9993 takeTill1 :: (Element (Seq syn) -> Bool) -> syn () (Seq syn)
10094 takeTill1 p = takeWhile1 (not . p)
10195
102− -- | Constant size vector. The default implementation uses lists, but
103− -- "syntax-attoparsec" and "syntax-printer" override it with an efficient
104− -- implementation that works on a vector directly.
105− --
106− -- @vecN n e@ describes a size @n@ vector with elements @e@.
107− --
108− -- Also see 'Data.Syntax.Combinator.vec'.
109− vecN :: Int -> syn () a -> syn () (V.Vector a)
110− vecN n e = packed /$/ sireplicate n e
111−
112− -- | Constant size vector with index-aware element. The default implementation
113− -- uses lists, but "syntax-attoparsec" and "syntax-printer" override it with an
114− -- efficient implementation that works on a vector directly.
115− --
116− -- @ivecN n e@ describes a size @n@ vector with elements @e@. Each element
117− -- gets its index and should output a value and the index unchanged.
118− --
119− -- Also see 'Data.Syntax.Combinator.ivec'.
120− ivecN :: Int -> syn Int (Int, a) -> syn () (V.Vector a)
121− ivecN n e = (packed /$/)
122− $ sisequence
123− $ map (\(i, e') -> constant i ^>> e'
124− >>> first (sipure (constant i))
125− >># unit . swapped)
126− $ zip [0 .. n-1] (replicate n e)
127−
128− -- | Constant size unboxed vector. The default implementation uses lists, but
129− -- "syntax-attoparsec" and "syntax-printer" override it with an efficient
130− -- implementation that works on a vector directly.
131− --
132− -- @vecN n e@ describes a size @n@ vector with elements @e@.
133− --
134− -- Also see 'Data.Syntax.Combinator.vec'.
135− uvecN :: VU.Unbox a => Int -> syn () a -> syn () (VU.Vector a)
136− uvecN n e = packed /$/ sireplicate n e
137−
138− -- | Constant size unboxed vector with index-aware element. The default implementation
139− -- uses lists, but "syntax-attoparsec" and "syntax-printer" override it with an
140− -- efficient implementation that works on a vector directly.
141− --
142− -- @ivecN n e@ describes a size @n@ vector with elements @e@. Each element
143− -- gets its index and should output a value and the index unchanged.
144− --
145− -- Also see 'Data.Syntax.Combinator.ivec'.
146− uivecN :: VU.Unbox a => Int -> syn Int (Int, a) -> syn () (VU.Vector a)
147− uivecN n e = (packed /$/)
148− $ sisequence
149− $ map (\(i, e') -> constant i ^>> e'
150− >>> first (sipure (constant i))
151− >># unit . swapped)
152− $ zip [0 .. n-1] (replicate n e)
153−
15496 {-# MINIMAL anyChar #-}
155−
156−instance Syntax syn => Syntax (ReaderCT env syn) where
157− type Seq (ReaderCT env syn) = Seq syn
158− anyChar = clift anyChar
159− char = clift . char
160− notChar = clift . notChar
161− satisfy = clift . satisfy
162− satisfyWith ai = clift . satisfyWith ai
163− string = clift . string
164− take = clift . take
165− takeWhile = clift . takeWhile
166− takeWhile1 = clift . takeWhile1
167− takeTill = clift . takeTill
168− takeTill1 = clift . takeTill1
169− vecN n e = ReaderCT $ vecN n . runReaderCT e
170− ivecN n e = ReaderCT $ ivecN n . runReaderCT e
171− uvecN n e = ReaderCT $ uvecN n . runReaderCT e
172− uivecN n e = ReaderCT $ uivecN n . runReaderCT e
173−
174−-- | Execute a computation in an isolated context.
175−--
176−-- The motivating example: you want to write a function
177−--
178−-- > serializeList :: Syntax syn => syn () a -> syn () [a]
179−--
180−-- Notice that we cannot just use simany, because the first @syn () a@
181−-- could eat the entire sequence (even though we printed more than 1 value!), so
182−-- we have to insert some kind of separators between the element. But how can we
183−-- be sure that @syn () a@ will not eat our separators? We can't! Thats why we
184−-- have to do the parsing in two stages: first extract the sequence between separators,
185−-- then run the @syn () a@ on this sequence.
186−class Syntax syn => Isolable syn where
187− -- | Turns a computation into one executed in an isolated context.
188− isolate :: syn () b -> syn (Seq syn) b
Data/Syntax/Char.hs−8
@@ -25,8 +25,6 @@ module Data.Syntax.Char (
2525 digitHex
2626 ) where
2727
28−import Control.Category.Reader
29−import Control.Category.Structures
3028import Control.Lens.SemiIso
3129import Control.SIArrow
3230import Data.Bits
@@ -56,12 +54,6 @@ class (Syntax syn, Element (Seq syn) ~ Char) => SyntaxChar syn where
5654
5755 {-# MINIMAL decimal, hexadecimal, realFloat, scientific #-}
5856
59−instance SyntaxChar syn => SyntaxChar (ReaderCT env syn) where
60− decimal = clift decimal
61− hexadecimal = clift hexadecimal
62− scientific = clift scientific
63− realFloat = clift realFloat
64−
6557-- | An useful synonym for SyntaxChars with Text sequences.
6658type SyntaxText syn = (SyntaxChar syn, Seq syn ~ Text)
6759
Data/Syntax/Combinator.hs+11−149
@@ -1,6 +1,3 @@
1−{-# LANGUAGE RankNTypes #-}
2−{-# LANGUAGE ScopedTypeVariables #-}
3−{-# LANGUAGE FlexibleContexts #-}
41{- |
52Module : Data.Syntax.Combinator
63Description : Combinators that work with any sequence type.
@@ -12,50 +9,11 @@ Stability : experimental
129
1310Combinators that work with any sequence type.
1411-}
15−module Data.Syntax.Combinator
16− (
17− -- * Combinators.
18− optional,
19− opt,
20− opt_,
21− choice,
22− eitherOf,
12+module Data.Syntax.Combinator where
2313
24− -- * Lists.
25− manyTill,
26− sepBy,
27− sepBy1,
28−
29− -- * Arrowized combinators.
30− takeArr,
31−
32− -- * Vectors.
33− vecNSepBy,
34− ivecNSepBy,
35− vec,
36− vecSepBy,
37− ivec,
38− ivecSepBy,
39−
40− -- * Unboxed vectors.
41− uvecNSepBy,
42− uivecNSepBy,
43− uvec,
44− uvecSepBy,
45− uivec,
46− uivecSepBy
47− ) where
48−
49−import Control.Category
50−import Control.Category.Structures
51−import Control.Lens
52−import Control.Lens.SemiIso
53−import Control.SIArrow
54−import qualified Data.MonoTraversable as Seq
55−import Data.Syntax
56−import qualified Data.Vector as V
57−import qualified Data.Vector.Unboxed as VU
58−import Prelude hiding (id, (.), take)
14+import Control.Lens
15+import Control.Lens.SemiIso
16+import Control.SIArrow
5917
6018-- | One or zero occurences of @f@.
6119optional :: SIArrow cat => cat () a -> cat () (Maybe a)
@@ -70,15 +28,6 @@ opt_ :: SIArrow cat => cat () () -> cat () ()
7028opt_ f = semiIso (const (Left "opt_")) Right /$/ f
7129 /+/ sipure id
7230
73−-- | Tries to apply the actions in the list in order, until one of
74−-- them succeeds. Returns the value of the succeeding action.
75−choice :: SIArrow cat => [cat () a] -> cat () a
76−choice = foldr (/+/) (sifail "choice: all alternatives failed")
77−
78−-- | Combine two alternatives.
79−eitherOf :: SIArrow cat => cat () a -> cat () b -> cat () (Either a b)
80−eitherOf a b = _Left /$/ a /+/ _Right /$/ b
81−
8231-- | @manyTill p end@ applies action p zero or more times until action
8332-- end succeeds, and returns the list of values returned by p.
8433manyTill :: SIArrow cat => cat () a -> cat () () -> cat () [a]
@@ -93,98 +42,11 @@ sepBy v s = sepBy1 v s /+/ sipure _Empty
9342sepBy1 :: SIArrow cat => cat () a -> cat () () -> cat () [a]
9443sepBy1 v s = _Cons /$/ v /*/ (s */ sepBy1 v s /+/ sipure _Empty)
9544
96−-- | A string of given length.
97−takeArr :: Syntax syn => syn Int (Seq syn)
98−takeArr = sibind $ iso (\n -> constant n #>> take n)
99− (\s -> constant (Seq.olength s) #>> take (Seq.olength s))
100−
101−-- | Constant size vector with separators.
102−--
103−-- @vecNSepBy n e sep@ describes a size @n@ vector with elements @e@ separated by @sep@.
104−vecNSepBy :: Syntax syn => Int -> syn () a -> syn () () -> syn () (V.Vector a)
105−vecNSepBy n e sep = ivecNSepBy n (unit ^>> second e) sep
106−
107−-- | Constant size vector with separators and index-aware elements.
108−--
109−-- @ivecNSepBy n e sep@ describes a size @n@ vector with elements @e@ separated by @sep@.
110−-- Each element gets its index and should output a value and the index unchanged.
111−ivecNSepBy :: Syntax syn => Int -> syn Int (Int, a) -> syn () () -> syn () (V.Vector a)
112−ivecNSepBy n e sep = ivecN n $ sibind $ iso el (el . fst)
113− where
114− el k | k < n - 1 = unit ^>> e *** sep >># unit
115− | otherwise = e
116−
117−-- | Runtime sized vector. The size can depend on the result of some computation.
118−--
119−-- @vec e@ describes a vector with elements @e@.
120−vec :: Syntax syn => syn () a -> syn Int (V.Vector a)
121−vec e = vecSepBy e (siarr id)
122−
123−-- | Runtime sized vector with separators. The size can depend on the result of some
124−-- computation.
125−--
126−-- @vecSepBy e sep@ describes a vector with elements @e@ separated by @sep@.
127−vecSepBy :: Syntax syn => syn () a -> syn () () -> syn Int (V.Vector a)
128−vecSepBy e sep = ivecSepBy (unit ^>> second e) sep
129−
130−-- | Runtime sized vector with index-aware elements. The size can depend on the result
131−-- of some computation.
132−--
133−-- @ivec e@ describes a vector with elements @e@.
134−ivec :: Syntax syn => syn Int (Int, a) -> syn Int (V.Vector a)
135−ivec e = ivecSepBy e (siarr id)
136−
137−-- | Runtime sized vector with index-aware elements and separators. The size can depend
138−-- on the result of some computation.
139−--
140−-- @ivecSepBy e sep@ describes a vector with elements @e@ separated by @sep@.
141−ivecSepBy :: Syntax syn => syn Int (Int, a) -> syn () () -> syn Int (V.Vector a)
142−ivecSepBy e sep = sibind $ iso (\n -> constant n #>> ivecNSepBy n e sep)
143− (\v -> constant (V.length v) #>> ivecNSepBy (V.length v) e sep)
144−
145−-- | Constant size unboxed vector with separators.
146−--
147−-- @uvecNSepBy n e sep@ describes a size @n@ vector with elements @e@ separated by @sep@.
148−uvecNSepBy :: (Syntax syn, VU.Unbox a) => Int -> syn () a -> syn () () -> syn () (VU.Vector a)
149−uvecNSepBy n e sep = uivecNSepBy n (unit ^>> second e) sep
150−
151−-- | Constant size unboxed vector with separators and index-aware elements.
152−--
153−-- @uivecNSepBy n e sep@ describes a size @n@ vector with elements @e@ separated by @sep@.
154−-- Each element gets its index and should output a value and the index unchanged.
155−uivecNSepBy :: (Syntax syn, VU.Unbox a) => Int -> syn Int (Int, a)
156− -> syn () () -> syn () (VU.Vector a)
157−uivecNSepBy n e sep = uivecN n $ sibind $ iso el (el . fst)
158− where
159− el k | k < n - 1 = unit ^>> e *** sep >># unit
160− | otherwise = e
161−
162−-- | Runtime sized unboxed vector. The size can depend on the result of some computation.
163−--
164−-- @uvec e@ describes a vector with elements @e@.
165−uvec :: (Syntax syn, VU.Unbox a) => syn () a -> syn Int (VU.Vector a)
166−uvec e = uvecSepBy e (siarr id)
167−
168−-- | Runtime sized unboxed vector with separators. The size can depend on the result of some
169−-- computation.
170−--
171−-- @uvecSepBy e sep@ describes a vector with elements @e@ separated by @sep@.
172−uvecSepBy :: (Syntax syn, VU.Unbox a) => syn () a -> syn () () -> syn Int (VU.Vector a)
173−uvecSepBy e sep = uivecSepBy (unit ^>> second e) sep
174−
175−-- | Runtime sized unboxed vector with index-aware elements. The size can depend on the result
176−-- of some computation.
177−--
178−-- @uivec e@ describes a vector with elements @e@.
179−uivec :: (Syntax syn, VU.Unbox a) => syn Int (Int, a) -> syn Int (VU.Vector a)
180−uivec e = uivecSepBy e (siarr id)
45+-- | Tries to apply the actions in the list in order, until one of
46+-- them succeeds. Returns the value of the succeeding action.
47+choice :: SIArrow cat => [cat () a] -> cat () a
48+choice = foldr (/+/) (sifail "choice: all alternatives failed")
18149
182−-- | Runtime sized unboxed vector with index-aware elements and separators. The size can depend
183−-- on the result of some computation.
184−--
185−-- @uivecSepBy e sep@ describes a vector with elements @e@ separated by @sep@.
186−uivecSepBy :: (Syntax syn, VU.Unbox a) => syn Int (Int, a)
187− -> syn () () -> syn Int (VU.Vector a)
188−uivecSepBy e sep = sibind $ iso (\n -> constant n #>> uivecNSepBy n e sep)
189− (\v -> constant (VU.length v)
190− #>> uivecNSepBy (VU.length v) e sep)
50+-- | Combine two alternatives.
51+eitherOf :: SIArrow cat => cat () a -> cat () b -> cat () (Either a b)
52+eitherOf a b = _Left /$/ a /+/ _Right /$/ b
Data/Syntax/Indent.hs+19−22
@@ -10,7 +10,7 @@ License : MIT
1010Maintainer : Paweł Nowak <pawel834@gmail.com>
1111Stability : experimental
1212
13−Provides a very simple indentation as a category transformer.
13+Provides a very simple indentation as a category transformer (a functor from Cat to Cat).
1414-}
1515module Data.Syntax.Indent (
1616 Indent,
@@ -21,7 +21,6 @@ module Data.Syntax.Indent (
2121
2222import Control.Category
2323import Control.Category.Reader
24−import Control.Category.Structures
2524import Control.SIArrow
2625import Data.Syntax
2726import Data.Syntax.Char
@@ -29,33 +28,31 @@ import Data.Syntax.Combinator
2928import Prelude hiding (takeWhile, take, id, (.))
3029
3130-- | Adds indentation to a syntax description.
32−newtype Indent cat a b = Indent { getIndent :: ReaderCT (Int, cat () ()) cat a b }
33− deriving (Category, Products, Coproducts, CatPlus, SIArrow)
31+newtype Indent cat a b = Indent { unIndent :: ReaderCT (Int, cat () ()) cat a b }
32+ deriving (Category, Products, Coproducts, CategoryPlus, SIArrow)
3433
35−instance CatTrans Indent where
34+instance CategoryTrans Indent where
3635 clift = Indent . clift
3736
38−-- Generalized newtype deriving cannot derive this :(
39−
4037instance Syntax syn => Syntax (Indent syn) where
4138 type Seq (Indent syn) = Seq syn
42− anyChar = Indent anyChar
43− char = Indent . char
44− notChar = Indent . notChar
45− satisfy = Indent . satisfy
46− satisfyWith ai = Indent . satisfyWith ai
47− string = Indent . string
48− take = Indent . take
49− takeWhile = Indent . takeWhile
50− takeWhile1 = Indent . takeWhile1
51− takeTill = Indent . takeTill
52− takeTill1 = Indent . takeTill1
39+ anyChar = clift anyChar
40+ char = clift . char
41+ notChar = clift . notChar
42+ satisfy = clift . satisfy
43+ satisfyWith ai = clift . satisfyWith ai
44+ string = clift . string
45+ take = clift . take
46+ takeWhile = clift . takeWhile
47+ takeWhile1 = clift . takeWhile1
48+ takeTill = clift . takeTill
49+ takeTill1 = clift . takeTill1
5350
5451instance SyntaxChar syn => SyntaxChar (Indent syn) where
55− decimal = Indent decimal
56− hexadecimal = Indent hexadecimal
57− scientific = Indent scientific
58− realFloat = Indent realFloat
52+ decimal = clift decimal
53+ hexadecimal = clift hexadecimal
54+ scientific = clift scientific
55+ realFloat = clift realFloat
5956
6057-- | @runIndent m tab@ runs the 'Indent' transformer using @tab@ once for each
6158-- level of indentation.
syntax.cabal+20−22
@@ -1,35 +1,33 @@
11name: syntax
22version: 1.0.0.0
3−synopsis: Reversible parsing and pretty-printing.
3+synopsis: Syntax descriptions for unified parsing and pretty-printing.
44description:
5− "syntax" allows you to write a single syntax description and instantiate is both as a parser and a pretty printer.
5+ 'syntax' allows you to write a single syntax description and instantiate is both as a parser and a pretty printer.
66 .
7− Syntax descriptions are written in applicative or arrow style. The library uses a custom typeclass hierarchy, provided
8− by the "semi-iso" package. Most of the time you will be using operators like '/$/', '/*/' and '/+/' (= '<|>'), just
9− like parser combinators. When more power is needed - e.g. when the syntax depends on the parsed or printed value -
10− you turn to arrows.
7+ The interface is based on a custom Functor\/Applicative\/Monad hierarchy, provided by the 'semi-iso' package. You fmap using
8+ a semi-isomorphism instead of function. A semi-isomorphism is a isomorphism that can fail in both directions, with slightly
9+ weakened laws. It is worth to note that @Iso@s and @Prism@s from 'lens' are valid semi-isomorphisms :)
1110 .
12− Semi-isomorphisms from "semi-iso" are the basic building block of syntax descriptions. I recommend reading
13− the hackage page of "semi-iso" first, as it contains much more information.
11+ Once you write a description you can, for example turn it into an Attoparsec parser, or into a 'Data.Syntax.Printer.Text.Printer'.
1412 .
15− Once you write a syntax description (polymorphic in the syntax category) you can instantiate it both as a parser or as
16− a pretty-printer. The library "syntax-attoparsec" gives you the ability to extract an Attoparsec parser. Pretty-printing
17− is implemented by the "syntax-printer" library, which uses Text and ByteString builders. (Note that formatting is handled
18− by "syntax" itself, not by the printer library)
13+ See @syntax-example@ and @syntax-example-json@ for examples, @syntax-attoparsec@ and @syntax-printer@ for a parser/printer implementation.
1914 .
20− Advanced formatting and parsing (for example indentation, haskell layout rule) is implemented as category transformers
21− (similar to monad transformers). Currently only simple indentation is implemented (in "Data.Syntax.Indent") - basically
22− a reader category transformer that tracks current indentation level. I plan on implementing Haskell layout rule in the
23− future.
15+ The library was inspired by:
2416 .
25− The library can work with both text and binary data. Alas, there are no binary combinators implemented yet.
26− I will implement them when i have the time (but these category transformers look so much more interesting for now ;).
17+ * Rendel, Tillmann, and Klaus Ostermann. "Invertible syntax descriptions: unifying parsing and pretty printing." ACM Sigplan Notices. Vol. 45. No. 11. ACM, 2010.
2718 .
28− EXAMPLES! See @syntax-example@ and @syntax-example-json@ for examples.
19+ TODO:
2920 .
30− * "syntax-example" implements a simple lambda calculus.
21+ * Research relative monads and relative monad transformers. Indent is basically a Reader monad over a syntax. How would a State monad look?
3122 .
32− * "syntax-example-json" implements a json parser and pretty printer.
23+ * Try to implement Haskell layout rule.
24+ .
25+ * Combinators for binary data formats, vectors.
26+ .
27+ * Better error messages.
28+ .
29+ * Maybe an implementation of do notation for SemiIsoMonad with QuasiQuoters, like the codo notation for comonads.
30+
3331license: MIT
3432license-file: LICENSE
3533author: Paweł Nowak
@@ -48,5 +46,5 @@ library
4846 Data.Syntax.Char
4947 Data.Syntax.Combinator
5048 Data.Syntax.Indent
51− build-depends: base >= 4 && < 5, mono-traversable, lens >= 4, semi-iso >= 1, scientific >= 0.3, text, vector
49+ build-depends: base >= 4 && < 5, mono-traversable, lens >= 4, semi-iso >= 1.0.0.0, scientific >= 0.3, text
5250 default-language: Haskell2010