penntreebank-megaparsec-0.2.0: src/Data/Tree/Parser/Penn/Megaparsec/Char.hs
{-|
Module : Data.Tree.Parser.Penn.Megaparsec
Description : Parser combinators that parse 'Data.Text'
or 'String' streams to generate Penn Treebank trees.
Copyright : (c) 2020 Nori Hayashi
License : BSD3
Maintainer : Nori Hayashi <net@hayashi-lin.net>
Stability : experimental
Portability : portable
Language : Haskell2010
-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE UndecidableInstances #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE ScopedTypeVariables #-}
module Data.Tree.Parser.Penn.Megaparsec.Char (
-- * Type Classes for Node Parsers
ParsableAsTerm(..),
UnsafelyParsableAsTerm(..),
-- * Parsers
pTree,
pUnsafeTree,
-- * Parser Type Synonyms
PennTreeParserT,
PennTreeParser,
-- * Parser Runners
-- | Re-imported from "Text.Megaparsec".
parse,
parseMaybe,
parseTest,
runParser,
runParser',
runParserT,
runParserT'
) where
import Data.Char (isSpace)
import Data.Tree
import Data.Void (Void)
import Data.Proxy (Proxy(..))
import Text.Megaparsec
import qualified Text.Megaparsec.Char as MC
import qualified Text.Megaparsec.Char.Lexer as Lex
import Control.Monad (forM_)
import Control.Monad.Trans.Class (lift)
import Control.Monad.Identity (Identity)
import Data.Tree.Parser.Penn.Megaparsec.Internal
{-|
A type class for node label types @term@
data of which can be obtained by parsing a stream of type @str@
that is supposed to be embedded into the parser
'Data.Tree.Parser.Penn.Megaparsec.Char.pUnsafeTree'.
It is your own responsibility to avoid these parsers to crash
by letting them unintendedly consume
symbols preserved for tree node demarcation
(e.g. parentheses and spaces).
@since 0.1.1
-}
class (Stream str) => UnsafelyParsableAsTerm str term where
{-|
A parser for non-terminal node labels.
-}
pUnsafeNonTerm :: (Ord err) => ParsecT err str m term
{-|
A parser for terminal node labels.
This parser may not comsume empty inputs,
for otherwise parsing will fall into infinte recursion.
-}
pUnsafeTerm :: (Ord err) => ParsecT err str m term
instance (Stream str, Tokens str ~ term, Token str ~ Char)
=> UnsafelyParsableAsTerm str term where
pUnsafeNonTerm
= takeWhileP (Just "Non-Terminal Label")
(\c -> c /= '(' && c /= ')' && not (isSpace c))
pUnsafeTerm
= takeWhile1P (Just "Terminal Label")
(\c -> c /= '(' && c /= ')' && not (isSpace c))
-- | A parser (monad transformer) that consumes spaces.
spaceConsumer :: (MonadParsec err str m, Token str ~ Char) => m ()
spaceConsumer
= Lex.space
MC.space1 -- Space
empty -- LineComment
empty -- BlockComment
-- | A parser (monad transformer) wrapper to make lexemes separated by spaces.
lexer :: (MonadParsec err str m, Token str ~ Char)
=> m term -- ^ A lexeme parser to be wrapped
-> m term -- ^ The parser wrapped by spaces
lexer = Lex.lexeme spaceConsumer
-- | A parser (monad transformer) wrapper that parenthesize the given parser.
pParens :: (
MonadParsec err str m,
Token str ~ Char,
Tokens str ~ str
) => m term -- ^ A parser to be wrapped
-> m term -- ^ The parser wrapped by parentheses
pParens =
between
(lexer $ single '(')
(lexer $ single ')')
-- | A parser for raw node labels.
invokeLabelParserRaw :: (MonadParsec err str m, Token str ~ Char)
=> m (Tokens str)
invokeLabelParserRaw
= takeWhile1P
(Just "Literal String")
(\x -> x /= '(' && x /= ')' && not (isSpace x))
{-|
A vertial parser (monad transformer) compositor that
make raw node labels get a secondary parse
by another parser.
-}
invokeLabelParser :: (
Ord err,
ParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m term -- ^ A secondary parser for node labels
-> State str err -- ^ The state given by the main parsing thread
-> ParsecT err str m term -- ^ The resulting parser that is going to be embedded in the main thread
invokeLabelParser labelParser substate = do
(_, res) <- lift $ runParserT' (labelParser <* eof) substate
case res of
Left b@(ParseErrorBundle errors _) -> do
forM_ errors registerParseError
return undefined
Right label -> return label
{-|
A parser (parser monad transformer) for trees in the Penn Treebank format,
where @err@ is the type of custom errors,
@str@ is the type of the stream,
@m@ is the type of the undelying monad and
@term@ is the type of node labels.
This parser will do a preliminary parse for
node labels as raw strings of type @str@.
The carved strings will contain no spaces or parentheses.
A seconsary parsing will take place on the spot
the way of parsing of which is designated by you
specifying the type @term@ as an instance of 'ParsableAsTerm'.
This parser accepts various types of text stream,
including 'String', 'Data.Text.Text' and 'Data.Text.Lazy.Text'.
You might need to manually annotate the type of this parser
to specify what type of stream you target at in the following way:
> (pTree :: ParsecT Void Text Identity (Tree Text))
-}
pTree ::
(
Ord err,
ParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m (Tree term)
pTree
= pParens pTreeInside <|> pTerminalNode <?> "Parsed Tree"
where
pTreeInside :: forall str. forall err. forall m. forall term. (
Ord err,
ParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m (Tree term)
pTreeInside = do
state <- getParserState
maybeLabelRaw <- lexer $ optional invokeLabelParserRaw
let substate = state {
stateInput
= case maybeLabelRaw of
Just labelRaw -> labelRaw
Nothing -> tokensToChunk pxy []
,
stateOffset = 0
}
labelParsed <- invokeLabelParser pNonTerm substate
children <- many $ lexer pTree -- Recursion
return $ Node labelParsed children
where
pxy :: Proxy str
pxy = Proxy
pTerminalNode :: (
Ord err,
ParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m (Tree term)
pTerminalNode = do
state <- getParserState
labelRaw <- lexer $ invokeLabelParserRaw
labelParsed <- do
let substate = state {
stateInput = labelRaw,
stateOffset = 0
}
invokeLabelParser pTerm substate
return $ Node labelParsed []
{-|
Another parser for trees in the Penn Treebank format,
where @err@ is the type of custom errors,
@str@ is the type of the stream,
@m@ is the type of the undelying monad and
@term@ is the type of node labels.
Apart from 'pTree', you can customize node label parsers
in a more liberal way
by specifying the type @term@
as an instance of 'UnsafelyParsableAsTerm'.
You can let parsers consume spaces and parentheses
(perhaps with quotation markers)
unless they do not broke the parsing of the whole trees.
It is all your responsibility to make sure of it.
This parser accepts various types of text stream,
including 'String', 'Data.Text.Text' and 'Data.Text.Lazy.Text'.
You might need to manually annotate the type of this parser
to specify what type of stream you target at in the following way:
> (pUnsafeTree :: ParsecT Void Text Identity (Tree Text))
@since 0.1.1
-}
pUnsafeTree ::
(
Ord err,
UnsafelyParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m (Tree term)
pUnsafeTree
= pParens pTreeInside <|> pTerminalNode <?> "Parsed Tree"
where
pTreeInside :: forall str. forall err. forall m. forall term. (
Ord err,
UnsafelyParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m (Tree term)
pTreeInside
= Node
<$> lexer pUnsafeNonTerm
<*> (many $ lexer pUnsafeTree)
pTerminalNode :: (
Ord err,
UnsafelyParsableAsTerm str term,
Monad m,
Token str ~ Char,
Tokens str ~ str
) => ParsecT err str m (Tree term)
pTerminalNode
= Node <$> lexer pUnsafeTerm <*> pure []
type PennTreeParserT str m term = ParsecT Void str m (Tree term)
type PennTreeParser str term = PennTreeParserT str Identity term