GLL.Combinators.Interface

This module provides a combinator library in which combinators act as an interface to a back-end parser. This technique is inspired by Tom Ridge (2014) and Peter Ljunglöf (2002). The back-end parser is provided in GLL.Parser. The library is fully general: it enables writing parsers for arbitrary context-free grammars.

The user writes a combinator expression representing a grammar. The represented grammar is extracted and given, together with an input string, to a back-end parser. The derivations constructed by the parser act as a guide in the "semantic phase" in which the combinator expressions are evaluated to produce semantic results for all derivations. Infinitely many derivations would result in a loop. This problem is avoided by discarding the derivations that would arise from such a loop.

This library uses GLL.Parser as the back-end parser and provides Control.Applicative-like parser combinators: <**> for sequencing, <||> for choice, <$$> for application, satisfy instead of pure and derived combinators <**, **>, <$$, many, some and optional.

The semantic phase might benefit from memoisation (see memo). Using memoisation voids pureness waranty".

Example usage

This library differs from parser combinator libraries in that combinator expressions are used to describe a grammar rather than a parser.

A rule is considered to be of the form X ::= a | .. | z, and represented by the combinator expression.

pX = "X" <::=> altA <||> ... <||> altZ

Alternates ('a' ... 'z') start with the application of a semantic action using <$$> (or variants <$$ and satisfy). The alternate is extended with <**> (or variants **>, <**).

altA = action1 <$$> keychar 'a' <**> pX
altZ = action2 <$$> keychar 'z'

Usability is improved by automatic lifting between expressions that represent symbols and alternates. The main difference with Control.Applicative style parser combinator libraries is that the user must use <:=> (or <::=>) to represent all recursive nonterminals and must use <::=> to represent all nonterminals that potentially lead to an infinite number of derivations. It is, however, possible to represent left-recursive nonterminals.

Example

In this example we define expressions for parsing (and evaluating) simple arithmetic expressions for single digit integers.

The library is capable of parsing arbitrary token types that are Parseable, orderable and have a Show instance. This example demonstrates the usage of the builtin Token datatype and uses the elementary parsers keychar and int_lit to create parsers for character- and integer-terminals.

We define a very simpler lexer first.

lexer :: String -> [Token]
lexer [] = []
lexer (x:xs)
    | isDigit x = IntLit (Just (read [x])) : lexer xs
    | otherwise = Char x                   : lexer xs

Note that the Char constructor of the Token type is used for character-level parsing. Char contains no lexeme, unlike the Int constructor.

Consider the following (highly ambiguous and left-recursive) grammar:

Expr ::= Expr '-' Expr
       | Expr '+' Expr
       | Expr '*' Expr
       | Expr '/' Expr
       | INT
       | '(' Expr ')'

The grammar is translated to the following combinator expression, adding the expected evaluation functions as semantic actions.

pExpr :: BNF Token Int
pExpr = "Expr" <::=> (-) <$$> pExpr <** keychar '-' <**> pExpr
                 <||> (+) <$$> pExpr <** keychar '+' <**> pExpr
                 <||> (*) <$$> pExpr <** keychar '*' <**> pExpr
                 <||> div <$$> pExpr <** keychar '/' <**> pExpr
                 <||> int_lit
                 <||> parens pExpr

Note that <** is used to ignore the parse result of the second argument and that **> is used to ignore the parse result of the first argument. These combinators help us to define the derived combinators within and parens.

within :: BNF Token a -> BNF Token b -> BNF Token a -> BNF Token b
within l p r = mkRule $ l **> p <** r

parens :: BNF Token a -> BNF Token a
parens p = within (keychar '(') p (keychar ')')

All possible evaluations are obtained by invoking the parse function.

run1 = parse pExpr (lexer "1+2*2-5")            -- [0,1,0,-5,-9] 
run2 = parse pExpr (lexer "((1+(2*2))-3)-5")    -- [-3]

With every introduction of an operator +, -, * or / the number of ambiguities is multiplied. The number of ambiguities behaves like the sequence https://oeis.org/A000108.

Simple disambiguation

This library offers simple disambiguation strategies that are applied post-parse (the parser still faces the ambiguity, but the semantic evaluation only yields the results according to the strategy). The disambiguations strategies are still in the experimental phase.

We group the operators according to their priorities and use <::= to turn the choice operator <||> into a left-biased operator locally (use leftBiased for the same effect globally).

pExpr1 :: BNF Token Int
pExpr1 = "Expr" <::=  (      (-) <$$> pExpr1 <** keychar '-' <**> pExpr1
                        <||> (+) <$$> pExpr1 <** keychar '+' <**> pExpr1 )
                 <||> (      (*) <$$> pExpr1 <** keychar '*' <**> pExpr1
                        <||> div <$$> pExpr1 <** keychar '/' <**> pExpr1 )
                 <||> (      int_lit
                        <||> parens pExpr1 )

run3 = parseWithOptions [maximumPivotAtNt] pExpr1 (lexer "1+2*2-5") -- [0]

The option maximumPivotAtNt enables the 'longest-match' disambiguation strategy and makes the arithmetic operators left-associative.

Grammar rewrites

To deal with the particular ambiguities associated with operators we can rewrite the grammar to disambiguate pre-parse.

We define the chainl combinator for parsing chains of left-associative operators.

chainl :: BNF Token a -> BNF Token (a -> a -> a) -> BNF Token a
chainl p s = mkRule $
    foldl (flip ($)) <$$> p <**> many (flip <$$> s <**> p)

The expression parser is written with chainl as follows:

pExpr2 :: BNF Token Int
pExpr2 = pE1
 where  pE1 = chainl pE2 ("E1" <::=> (+) <$$ keychar '+' <||> (-) <$$ keychar '-')
        pE2 = chainl pE3 ("E2" <::=> (*) <$$ keychar '*' <||> div <$$ keychar '/')
        pE3 = "E3" <::=> int_lit <||> parens pExpr2

run4 = parse pExpr2 (lexer "1+2*2-5")       -- [0]

Pre-parse disambiguation is desirable, as the parsing process could speed up dramatically. In general however, it is not always possible to find the appropriate grammar rewrite and implement it in a high-level combinator such as chainl, motivating the existence of this library.

More simple examples can be found in GLL.Combinators.Test.Interface.

Elementary parsers

term_parser :: t -> (t -> a) -> SymbExpr t a

Create a symbol-parse for a terminal given:

• The Parseable token represented by the terminal.
• A function from that Parseable to a semantic result.

satisfy :: (Show t, Ord t) => a -> AltExpr t a

The empty right-hand side that yields its first argument as a semantic result.

Elementary parsers using the Token datatype

keychar :: SubsumesToken t => Char -> SymbExpr t Char

Parse a single character, using a SubsumesToken type.

keyword :: SubsumesToken t => String -> SymbExpr t String

Parse a single character, using a SubsumesToken type.

int_lit :: SubsumesToken t => SymbExpr t Int

Parse a single integer, using a SubsumesToken type. Returns the lexeme interpreted as an integer.

bool_lit :: SubsumesToken t => SymbExpr t Bool

Parse a single Boolean, using a SubsumesToken type. Returns the lexeme interpreter as a Boolean.

char_lit :: SubsumesToken t => SymbExpr t Char

Parse a single Character literal, using a SubsumesToken type. Returns the lexeme interpreted as a Character literal.

string_lit :: SubsumesToken t => SymbExpr t String

Parse a single String literal, using a SubsumesToken type. Returns the lexeme interpreted as a String literal.

alt_id_lit :: SubsumesToken t => SymbExpr t String

Parse a single alternative identifier, using a SubsumesToken type. Returns the lexeme as a String.

id_lit :: SubsumesToken t => SymbExpr t String

Parse a single identifier, using a SubsumesToken type. Returns the lexeme as a String.

token :: SubsumesToken t => String -> SymbExpr t String

Parse a single arbitrary token, using a SubsumesToken type. Returns the lexeme.

Elementary character-level parsers

char :: Char -> SymbExpr Char Char

Parse a single character.

char c = term_parser c id

Currently, this is the only character-level combinator exported by this module. Please use token-level combinators for practical parsing. Might change in the future.

Elementary combinators

Sequencing

(<**>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t (a -> b) -> s t a -> AltExpr t b infixl 4

Add a SymbExpr to the right-hand side represented by an AltExpr creating a new AltExpr. The semantic result of the first argument is applied to the second as a cross-product.

Choice

(<||>) :: (Show t, Ord t, IsAltExpr i, HasAlts b) => i t a -> b t a -> AltExprs t a infixr 3

Add an AltExpr to a list of AltExpr The resuling '[] :. AltExpr' forms the right-hand side of a rule.

Semantic actions

(<$$>) :: (Show t, Ord t, IsSymbExpr s) => (a -> b) -> s t a -> AltExpr t b infixl 4

Form an AltExpr by mapping some semantic action overy the result of the second argument.

Nonterminal introduction

(<:=>) :: (Show t, Ord t, HasAlts b) => String -> b t a -> SymbExpr t a infixl 2

Form a rule by giving the name of the left-hand side of the new rule. Use this combinator on recursive non-terminals.

(<::=>) :: (Show t, Ord t, HasAlts b) => String -> b t a -> SymbExpr t a infixl 2

Variant of <:=> for recursive non-terminals that have a potentially infinite number of derivations for some input string.

A non-terminal yields infinitely many derivations if and only if it is left-recursive and would be left-recursive if all the right-hand sides of the productions of the grammar are reversed.

chooses :: (Show t, Ord t) => String -> [AltExpr t a] -> SymbExpr t a

Variant of ::= that can be supplied with a list of alternates

Types

Grammar (combinator expression) types

type BNF t a = SymbExpr t a

A combinator expression representing a BNF-grammar. The terminals of the grammar are of type t. When used to parse, the expression yields semantic results of type a.

data SymbExpr t a

A combinator expression representing a symbol. A SymbExpr either represents a terminal or a nonterminal. In the latter case it is constructed with (a variant of) <:=> and adds a rule to the grammar of which the represented symbol is the left-hand side.

Instances

IsAltExpr SymbExpr

Methods

toAlt :: (Show t, Ord t) => SymbExpr t b -> AltExpr t b

HasAlts SymbExpr

Methods

altsOf :: (Show t, Ord t) => SymbExpr t b -> [AltExpr t b]

IsSymbExpr SymbExpr

Methods

toSymb :: (Show t, Ord t) => SymbExpr t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => SymbExpr t b -> BNF t b

data AltExpr t a

A combinator expression representing an alternative: the right-hand side of a production.

Instances

IsAltExpr AltExprs

Methods

toAlt :: (Show t, Ord t) => AltExprs t b -> AltExpr t b

IsAltExpr AltExpr

Methods

toAlt :: (Show t, Ord t) => AltExpr t b -> AltExpr t b

HasAlts AltExprs

Methods

altsOf :: (Show t, Ord t) => AltExprs t b -> [AltExpr t b]

HasAlts AltExpr

Methods

altsOf :: (Show t, Ord t) => AltExpr t b -> [AltExpr t b]

IsSymbExpr AltExprs

Methods

toSymb :: (Show t, Ord t) => AltExprs t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => AltExprs t b -> BNF t b

IsSymbExpr AltExpr

Methods

toSymb :: (Show t, Ord t) => AltExpr t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => AltExpr t b -> BNF t b

type AltExprs = OO [] AltExpr

A list of alternatives represents the right-hand side of a rule.

Parseable token types

data Token

A datatype for representing tokens with some builtins and an aribitrary Token constructor. This datatype stores (optional) lexemes.

Constructors

Char Char

Keyword String

EOS

Epsilon

IntLit (Maybe Int)

BoolLit (Maybe Bool)

StringLit (Maybe String)

CharLit (Maybe Char)

IDLit (Maybe String)

AltIDLit (Maybe String)

alternative identifiers, for example functions vs. constructors (as in Haskell).

Token String (Maybe String)

Instances

Eq Token

Methods

(==) :: Token -> Token -> Bool

(/=) :: Token -> Token -> Bool

Ord Token

Methods

compare :: Token -> Token -> Ordering

(<) :: Token -> Token -> Bool

(<=) :: Token -> Token -> Bool

(>) :: Token -> Token -> Bool

(>=) :: Token -> Token -> Bool

max :: Token -> Token -> Token

min :: Token -> Token -> Token

Show Token

Methods

showsPrec :: Int -> Token -> ShowS

show :: Token -> String

showList :: [Token] -> ShowS

SubsumesToken Token

Methods

upcast :: Token -> Token

downcast :: Token -> Maybe Token

Parseable Token

Methods

eos :: Token

eps :: Token

matches :: Token -> Token -> Bool

unlex :: Token -> String

class (Ord a, Eq a, Show a) => Parseable a where

Class that captures elements of an input string (tokens).

• eos is the end-of-string symbol
• eps is the empty-string symbol

Both eos and eps must be distinct from eachother and from all tokens in the input string. The show instance is required to throw error messages.

Minimal complete definition

eos, eps, matches

Methods

eos :: a

eps :: a

matches :: a -> a -> Bool

This function is used for matching grammar tokens and input tokens. Override this method if, for example, your input tokens store lexemes while the grammar tokens do not

unlex :: a -> String

This function pretty-prints the Parseable type by displaying its lexeme. Default implementation is show, which should be replaced for prettier error messages.

Instances

Parseable Token

Methods

eos :: Token

eps :: Token

matches :: Token -> Token -> Bool

unlex :: Token -> String

class SubsumesToken a where

Class whose members are super-types of Token.

Minimal complete definition

upcast, downcast

Methods

upcast :: Token -> a

downcast :: a -> Maybe Token

Instances

SubsumesToken Token

Methods

upcast :: Token -> Token

downcast :: Token -> Maybe Token

unlexTokens :: [Token] -> String

Pretty-prints a list of Tokens as a concatenation of their lexemes.

unlexToken :: Token -> String

Running a parser

parse :: (Show t, Parseable t, IsSymbExpr s) => s t a -> [t] -> [a]

Runs a parser given a string of Parseables and returns a list of semantic results, corresponding to all finitely many derivations.

Running a parser with options

parseWithOptions :: (Show t, Parseable t, IsSymbExpr s) => CombinatorOptions -> s t a -> [t] -> [a]

Run the parser with some CombinatorOptions.

parseWithParseOptions :: (Show t, Parseable t, IsSymbExpr s) => ParseOptions -> CombinatorOptions -> s t a -> [t] -> [a]

Run the parser with some ParseOptions and CombinatorOptions.

Possible options

type CombinatorOptions = [CombinatorOption]

A list of CombinatorOptions for evaluating combinator expressions.

type CombinatorOption = PCOptions -> PCOptions

A single option.

maximumErrors :: Int -> CombinatorOption

Set the maximum number of errors shown in case of an unsuccessful parse.

throwErrors :: CombinatorOption

If there are no parse results, the default behaviour is to return an empty list. If this option is used, a runtime error will be reported, with debugging information.

maximumPivot :: CombinatorOption

Enables a 'longest-match' at production level.

maximumPivotAtNt :: CombinatorOption

Enables 'longest-match' at non-terminal level.

Parser options

fullSPPF :: ParseOption

Create the SPPF with all nodes and edges, not necessarily strictly binarised.

allNodes :: ParseOption

Create all nodes, but no edges between nodes.

packedNodesOnly :: ParseOption

Create packed-nodes only.

strictBinarisation :: ParseOption

Fully binarise the SPPF, resulting in a larger SPPF and possibly slower runtimes. When this flag is on, packed nodes can only have a single symbol node child or one intermediate node child and one symbol node child. With the flag disabled a packed node can have two symbol node children.

noSelectTest :: ParseOption

Turn of select tests. Disables lookahead.

Running a parser with options and explicit failure

parseWithOptionsAndError :: (Show t, Parseable t, IsSymbExpr s) => CombinatorOptions -> s t a -> [t] -> Either String [a]

Run the parser with some CombinatorOptions and return either an error or the results. Any returned results will be a list of length greater than 0.

parseWithParseOptionsAndError :: (Show t, Parseable t, IsSymbExpr s) => ParseOptions -> CombinatorOptions -> s t a -> [t] -> Either String [a]

Run the parser with some ParseOptions and CombinatorOptions. Returns either an error or the results. Any returned results will be a list of length greater than 0.

Runing a parser to obtain ParseResult.

parseResult :: (Show t, Parseable t, IsSymbExpr s) => s t a -> [t] -> ParseResult t

Get the ParseResult, containing an SPPF, produced by parsing the given input with the given parser.

parseResultWithOptions :: (Show t, Parseable t, IsSymbExpr s) => ParseOptions -> CombinatorOptions -> s t a -> [t] -> ParseResult t

Get the ParseResult given some ParseOptions and CombinatorOptions.

data ParseResult t

The ParseResult datatype contains the SPPF and some other information about the parse:

• SPPF
• Whether the parse was successful
• The number of descriptors that have been processed
• The number of symbol nodes (nonterminal and terminal)
• The number of intermediate noes
• The number of packed nodes
• The number of GSS nodes
• The number of GSS edges

Constructors

ParseResult

Fields

• sppf_result :: SPPF t
• res_success :: Bool
• nr_descriptors :: Int
• nr_nterm_nodes :: Int
• nr_term_nodes :: Int
• nr_intermediate_nodes :: Int
• nr_packed_nodes :: Int
• nr_sppf_edges :: Int
• nr_gss_nodes :: Int
• nr_gss_edges :: Int
• error_message :: String

Instances

Show (ParseResult t)

Methods

showsPrec :: Int -> ParseResult t -> ShowS

show :: ParseResult t -> String

showList :: [ParseResult t] -> ShowS

Builtin lexers.

default_lexer :: SubsumesToken t => String -> [t]

A lexer using the default LexerSettings.

Lexer settings

lexer :: SubsumesToken t => LexerSettings -> String -> [t]

A lexer parameterised by LexerSettings.

data LexerSettings

Settings for changing the behaviour of the builtin lexer lexer. Lexers are built using Text.Regex.Applicative.

Constructors

LexerSettings

Fields

• keychars :: [Char]

  Which keychars to recognise? Default: none.

• keywords :: [String]

  Which keywords to recognise? Default: none.

• whitespace :: Char -> Bool

  What is considered a whitespace character? Default: isSpace.

• lineComment :: String

  How does a line comment start? Default: '"//"'.

• blockCommentOpen :: String

  How does a block comment open? Default: '"'{-'"'.

• blockCommentClose :: String

  How does a block comment close? Default: '"'-}'"'.

• identifiers :: RE Char String

  How to recognise identifiers? Default alphanumerical with lowercase alpha start.

• altIdentifiers :: RE Char String

  How to recognise alternative identifiers? Default alphanumerical with uppercase alpha start.

• tokens :: [(String, RE Char String)]

  Arbitrary tokens (a,b). a is the token name, b is a regular expression.

emptyLanguage :: LexerSettings

The default LexerSettings.

Derived combinators

mkNt :: (Show t, Ord t, IsSymbExpr s) => s t a -> String -> String

Helper function for defining new combinators. Use mkNt to form a new unique non-terminal name based on the symbol of a given SymbExpr and a String that is unique to the newly defined combinator.

Ignoring semantic results

(<$$) :: (Show t, Ord t, IsSymbExpr s) => b -> s t a -> AltExpr t b infixl 4

Variant of <$$> that ignores the semantic result of its second argument.

(**>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t a -> s t b -> AltExpr t b infixl 4

Variant of <**> that ignores the semantic result of the first argument.

(<**) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t a -> s t b -> AltExpr t a infixl 4

Variant of <**> that ignores the semantic result of the second argument.

EBNF patterns

optional :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t (Maybe a)

Make a piece of BNF optional, but proceed if unsuccessful (yielding Nothing as a result in that case).

multiple :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t [a]

Try to apply a parser multiple times (0 or more). The results are returned in a list. In the case of ambiguity the largest list is returned.

multiple1 :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t [a]

Try to apply a parser multiple times (1 or more). The results are returned in a list. In the case of ambiguity the largest list is returned.

multipleSepBy :: (Show t, Ord t, IsSymbExpr s, IsSymbExpr s2, IsAltExpr s2) => s t a -> s2 t b -> SymbExpr t [a]

Same as multiple but with an additional separator.

multipleSepBy1 :: (Show t, Ord t, IsSymbExpr s, IsSymbExpr s2, IsAltExpr s2) => s t a -> s2 t b -> SymbExpr t [a]

Same as multiple1 but with an additional separator.

multipleSepBy2 :: (IsSymbExpr s2, IsSymbExpr s, IsAltExpr s2, Show t, Ord t) => s t a -> s2 t b -> BNF t [a]

Like multipleSepBy1 but matching at least two occurrences of the first argument. The returned list is therefore always of at least length 2. At least one separator will be consumed.

within :: IsSymbExpr s => BNF Token a -> s Token b -> BNF Token c -> BNF Token b

Place a piece of BNF within two other BNF fragments, ignoring their semantics.

parens :: IsSymbExpr s => s Token b -> BNF Token b

Place a piece of BNF between the characters '(' and ')'.

braces :: IsSymbExpr s => s Token b -> BNF Token b

Place a piece of BNF between the characters '{' and '}'.

brackets :: IsSymbExpr s => s Token b -> BNF Token b

Place a piece of BNF between the characters '[' and ']'.

angles :: IsSymbExpr s => s Token b -> BNF Token b

Place a piece of BNF between the characters < and >.

Disambiguation

(<:=) :: (HasAlts b, Show t, Ord t) => String -> b t a -> SymbExpr t a infixl 2

Variant of <:=> that prioritises productions from left-to-right (or top-to-bottom).

(<::=) :: (HasAlts b, Show t, Ord t) => String -> b t a -> SymbExpr t a infixl 2

Variant of <::=> that prioritises productions from left-to-right (or top-to-bottom).

(<<<**>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t (a -> b) -> s t a -> AltExpr t b infixl 4

Variant of <**> that applies shortest match on the left operand.

(<**>>>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t (a -> b) -> s t a -> AltExpr t b infixl 4

Variant of <**> that applies longest match on the left operand.

(<<**>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t a -> s t b -> AltExpr t b infixl 4

(<<<**) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t a -> s t b -> AltExpr t a infixl 4

Variant <** that applies shortest match on its left operand

(**>>>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t a -> s t b -> AltExpr t b infixl 4

(<**>>) :: (Show t, Ord t, IsAltExpr i, IsSymbExpr s) => i t a -> s t b -> AltExpr t a infixl 4

Variant of <** that applies longest match on its left operand.

rassoc :: AltExpr t a -> AltExpr t a

Apply this combinator to an alternative to indicate that the alternative is for a right-associative operator. Used for top-down disambiguation.

lassoc :: AltExpr t a -> AltExpr t a

Apply this combinator to an alternative to indicate that the alternative is for a left-associative operator. Used for top-down disambiguation.

assoc :: AltExpr t a -> AltExpr t a

Apply this combinator to an alternative to indicate that the alternative is for an associative operator. Used for top-down disambiguation.

many :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t [a]

Try to apply a parser multiple times (0 or more) with shortest match applied to each occurrence of the parser.

many1 :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t [a]

Try to apply a parser multiple times (1 or more) with shortest match applied to each occurrence of the parser.

some :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t [a]

Try to apply a parser multiple times (0 or more) with longest match applied to each occurrence of the parser.

some1 :: (Show t, Ord t, IsSymbExpr s) => s t a -> SymbExpr t [a]

Try to apply a parser multiple times (1 or more) with longest match applied to each occurrence of the parser.

manySepBy :: (Show t, Ord t, IsSymbExpr s, IsSymbExpr s2, IsAltExpr s2) => s t a -> s2 t b -> SymbExpr t [a]

Same as many but with an additional separator.

manySepBy1 :: (Show t, Ord t, IsSymbExpr s, IsSymbExpr s2, IsAltExpr s2) => s t a -> s2 t b -> SymbExpr t [a]

Same as many1 but with an additional separator.

manySepBy2 :: (IsSymbExpr s2, IsSymbExpr s, IsAltExpr s2, Show t, Ord t) => s t a -> s2 t b -> BNF t [a]

Like multipleSepBy2 but matching the maximum number of occurrences of the first argument as possible (at least 2).

someSepBy :: (Show t, Ord t, IsSymbExpr s, IsSymbExpr s2, IsAltExpr s2) => s t a -> s2 t b -> SymbExpr t [a]

Same as some1 but with an additional separator.

someSepBy1 :: (Show t, Ord t, IsSymbExpr s, IsSymbExpr s2, IsAltExpr s2) => s t a -> s2 t b -> SymbExpr t [a]

Same as some1 but with an additional separator.

someSepBy2 :: (IsSymbExpr s2, IsSymbExpr s, IsAltExpr s2, Show t, Ord t) => s t a -> s2 t b -> BNF t [a]

Like multipleSepBy2 but matching the minimum number of occurrences of the first argument as possible (at least 2).

Lifting

class HasAlts a where

Class for lifting to AltExprs.

Minimal complete definition

altsOf

Methods

altsOf :: (Show t, Ord t) => a t b -> [AltExpr t b]

Instances

HasAlts AltExprs

Methods

altsOf :: (Show t, Ord t) => AltExprs t b -> [AltExpr t b]

HasAlts AltExpr

Methods

altsOf :: (Show t, Ord t) => AltExpr t b -> [AltExpr t b]

HasAlts SymbExpr

Methods

altsOf :: (Show t, Ord t) => SymbExpr t b -> [AltExpr t b]

class IsSymbExpr a where

Class for lifting to SymbExpr.

Minimal complete definition

toSymb

Methods

toSymb :: (Show t, Ord t) => a t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => a t b -> BNF t b

Synonym of toSymb for creating derived combinators.

Instances

IsSymbExpr AltExprs

Methods

toSymb :: (Show t, Ord t) => AltExprs t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => AltExprs t b -> BNF t b

IsSymbExpr AltExpr

Methods

toSymb :: (Show t, Ord t) => AltExpr t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => AltExpr t b -> BNF t b

IsSymbExpr SymbExpr

Methods

toSymb :: (Show t, Ord t) => SymbExpr t b -> SymbExpr t b

mkRule :: (Show t, Ord t) => SymbExpr t b -> BNF t b

class IsAltExpr a where

Class for lifting to AltExpr.

Minimal complete definition

toAlt

Methods

toAlt :: (Show t, Ord t) => a t b -> AltExpr t b

Instances

IsAltExpr AltExprs

Methods

toAlt :: (Show t, Ord t) => AltExprs t b -> AltExpr t b

IsAltExpr AltExpr

Methods

toAlt :: (Show t, Ord t) => AltExpr t b -> AltExpr t b

IsAltExpr SymbExpr

Methods

toAlt :: (Show t, Ord t) => SymbExpr t b -> AltExpr t b

Memoisation

memo :: (Ord t, Show t, IsSymbExpr s) => MemoRef [a] -> s t a -> SymbExpr t a

This function memoises a parser, given:

• A MemoRef pointing to a fresh MemoTable, created using newMemoTable.
• The SymbExpr to memoise.

Use memo on those parsers that are expected to derive the same substring multiple times. If the same combinator expression is used to parse multiple times the MemoRef needs to be cleared using memClear.

memo relies on unsafePerformIO and is therefore potentially unsafe. The option useMemoisation enables memoisation. It is off by default, even if memo is used in a combinator expression.

newMemoTable :: MemoRef a

Create a reference to a fresh MemoTable.

memClear :: MemoRef a -> IO ()

Clears the MemoTable to which the given reference refers.

type MemoTable a = IntMap (IntMap a)

A MemoTable maps left-extent l to right-extent r to some results a indicating the the substring ranging from l to r is derived with parse result a.

type MemoRef a = IORef (MemoTable a)

An impure reference to a MemoTable.

useMemoisation :: CombinatorOption

Whether to use unsafe memoisation to speed up the enumeration of parse results.