uu-parsinglib-2.4.1: Online, error-correcting parser combinators; monadic and applicative interfaces




The module Core contains the basic functionality of the parser library. It uses the breadth-first module to realise online generation of results, the error correction administration, dealing with ambigous grammars; it defines the types of the elementary parsers and recognisers involved.For typical use cases of the libray see the module Text.ParserCombinators.UU.Examples


The Classes Defining the Interface


class (Applicative p, ExtApplicative p, Alternative p) => IsParser p Source

This class collects a number of classes which together defines what a Parser should provide. Since it is just a predicate we have prefixed the name by the phrase Is


class ExtApplicative p whereSource

The module Control.Applicative contains the definition for <$ which cannot be changed . Since we want to give optimised implementations of this combinator, we hide its definition, and define a class containing its signature.


(<$) :: a -> p b -> p aSource


ExtApplicative (P st)

In the new module Control.Applicative the operator <$ is still hard coded. We hide this import and provide an implementation using a class, which can be redfined when needed.


class Symbol p symbol token | p symbol -> token whereSource

Many parsing libraries do not make a distinction between the terminal symbols of the language recognised and the tokens actually constructed from the input. This happens e.g. if we want to recognise an integer or an identifier: we are also interested in which integer occurred in the input, or which identifier.

The function pSym takes as argument a value of some type symbol, and returns a value of type token. The parser will in general depend on some state which is maintained holding the input. The functional dependency fixes the token type, based on the symbol type and the type of the parser p. Since pSym is overloaded both the type and the value of symbol determine how to decompose the input in a token and the remaining input. pSymExt is the actual function, which takes two extra parameters: one describing the minimal numer of tokens recognised, and the second whether the symbol can recognise the empty string and the value which is to be returned in that case


pSymExt :: Nat -> Maybe token -> symbol -> p tokenSource

The first parameter to pSymExt is a Nat which describes the minimal numer of tokens accepted by this parser. It is used in the abstract interpretation which computes this property for each parser. It's main use is in choosinga non-recursive alternative in case a non-terminal has to be inserted. The second parameter indicates whether this parser can also skip recognising anything and just return a value of type a, hence a `Maybe a`

pSym :: symbol -> p tokenSource


Provides state symbol token => Symbol (P state) symbol token 


class Provides state symbol token | state symbol -> token whereSource

The function splitStae playes a crucial role in splitting up the state. The symbol parameter tells us what kind of thing, and even which value of that kind, is expected from the input. The state and and the symbol type together determine what kind of token has to be returned. Since the function is overloaded we do not have to invent all kind of different names for our elementary parsers.


splitState :: symbol -> (token -> state -> Steps a) -> state -> Steps aSource


(Eq a, Show a) => Provides (Str a) a a 
(Show a, Eq a) => Provides (Str a) (Token a) [a] 
Show a => Provides (Str a) (Munch a) [a] 
(Ord a, Show a) => Provides (Str a) (a, a) a 
Show a => Provides (Str a) (a -> Bool, String, a) a 


class Eof state whereSource


eof :: state -> BoolSource

deleteAtEnd :: state -> Maybe (Cost, state)Source


Show a => Eof (Str a) 

The type describing parsers: P

data P st a Source


P (forall r. (a -> st -> Steps r) -> st -> Steps r) (forall r. (st -> Steps r) -> st -> Steps (a, r)) (forall r. (st -> Steps r) -> st -> Steps r) Nat (Maybe a) 


Monad (P st) 
Functor (P state) 
MonadPlus (P st) 
Applicative (P state) 
Alternative (P state) 
ExtApplicative (P st)

In the new module Control.Applicative the operator <$ is still hard coded. We hide this import and provide an implementation using a class, which can be redfined when needed.

Provides state symbol token => Symbol (P state) symbol token 

Parsers are functors: fmap

Parsers are Applicative: <*>, <*, *> and pure

Parsers are Alternative: <|> and empty

Parsers can recognise single tokens: pSym and pSymExt

Parsers are Monads: >>= and return

Additional useful combinators

Controlling the text of error reporting: <?>

(<?>) :: P state a -> String -> P state aSource

The parsers build a list of symbols which are expected at a specific point. This list is used to report errors. Quite often it is more informative to get e.g. the name of the non-terminal. The <?> combinator replaces this list of symbols by it's righ-hand side argument.

An alternative for the Alternative, which is greedy: <<|>

Parsers can be disambiguated using micro-steps: micro

Dealing with (non-empty) Ambigous parsers: amb

amb :: P st a -> P st [a]Source

Parse errors can be retreived from the state: pErrors

class Stores state error | state -> error whereSource

getErrors retreives the correcting steps made since the last time the function was called. The result can, using a monad, be used to control how to-- proceed with the parsing process.


getErrors :: state -> ([error], state)Source


pErrors :: Stores st error => P st [error]Source

Starting and finalising the parsing process: pEnd and parse

pEnd :: (Stores st error, Eof st) => P st [error]Source

The function pEnd should be called at the end of the parsing process. It deletes any unsonsumed input, and reports its preence as an eror.

parse :: Eof t => P t a -> t -> aSource

The state may be temporarily change type: pSwitch

pSwitch :: (st1 -> (st2, st2 -> st1)) -> P st2 a -> P st1 aSource

pSwitch takes the current state and modifies it to a different type of state to which its argument parser is applied. The second component of the result is a function which converts the remaining state of this parser back into a valuee of the original type.

A more efficient version for <$ from the module Control.Applicative

Maintaining Progress Information

type Cost = IntSource

The data type Steps is the core data type around which the parsers are constructed. It is a describes a tree structure of streams containing (in an interleaved way) both the online result of the parsing process, and progress information. Recognising an input token should correspond to a certain amount of Progress, which tells how much of the input state was consumed. The Progress is used to implement the breadth-first search process, in which alternatives are examined in a more-or-less synchonised way. The meaning of the various Step constructors is as follows:

A token was succesfully recognised, and as a result the input was advanced by the distance Progress
The type of value represented by the Steps changes by applying the function parameter.
A correcting step has to made to the input; the first parameter contains information about what was expected in the input, and the second parameter describes the various corrected alternatives, each with an associated Cost
A small cost is inserted in the sequence, which is used to disambiguate. Use with care!

The last two alternatives play a role in recognising ambigous non-terminals. For a full description see the technical report referred to from the README file..

data Steps a whereSource


Step :: Progress -> Steps a -> Steps a 
Apply :: forall a b. (b -> a) -> Steps b -> Steps a 
Fail :: Strings -> [Strings -> (Cost, Steps a)] -> Steps a 
Micro :: Cost -> Steps a -> Steps a 
End_h :: ([a], [a] -> Steps r) -> Steps (a, r) -> Steps (a, r) 
End_f :: [Steps a] -> Steps a -> Steps a 

eval :: Steps a -> aSource

push :: v -> Steps r -> Steps (v, r)Source

apply :: Steps (b -> a, (b, r)) -> Steps (a, r)Source

pushapply :: (b -> a) -> Steps (b, r) -> Steps (a, r)Source

best :: Steps a -> Steps a -> Steps aSource

best' :: Steps b -> Steps b -> Steps bSource

getCheapest :: Int -> [(Int, Steps a)] -> Steps aSource

traverse :: Int -> Steps a -> Int -> Int -> IntSource

Auxiliary functions and types

Checking for non-sensical combinations: must_be_non_empty and must_be_non_empties

must_be_non_empty :: [Char] -> P t t1 -> t2 -> t2Source

The function checks wehther its second argument is a parser which can recognise the mety sequence. If so an error message is given using the name of the context. If not then the third argument is returned. This is useful in testing for loogical combinations. For its use see the module Text>parserCombinators.UU.Derived

must_be_non_empties :: [Char] -> P t1 t -> P t3 t2 -> t4 -> t4Source

This function is similar to the above, but can be used in situations where we recognise a sequence of elements separated by other elements. This does not make sense if both parsers can recognise the empty string. Your grammar is then highly ambiguous.

The type Nat for describing the minimal number of tokens consumed

data Nat Source

The data type Nat is used to represent the minimal length of a parser. Care should be taken in order to not evaluate the right hand side of the binary functions nat_min and `nat-add` more than necesssary.


Succ Nat