Skip to content

Explore unboxed parser representations #193

@Boarders

Description

@Boarders

There is a nice library by Ed Kmett called parsnip: https://hackage.haskell.org/package/parsnip-0/

This represents a parser as follows:

newtype Option a = Option# (# a | (##) #)
type Result s a = (# Option a, Addr#, State# s #)

newtype Parser s a = Parser
 { runParser :: Addr# -> State# s -> Result s a
 }

Here is how the Alex parser for monad-bytestring looks:

newtype Alex a = Alex { unAlex :: AlexState -> Either String (AlexState, a) }

-- monad-bytestring wrapper
data AlexState = AlexState {
       alex_pos :: !AlexPosn,  -- position at current input location
       alex_bpos:: !Int64,     -- bytes consumed so far
       alex_inp :: ByteString.ByteString,      -- the current input
       alex_chr :: !Char,      -- the character before the input
       alex_scd :: !Int        -- the current startcode
     , alex_ust :: AlexUserState -- AlexUserState will be defined in the user program
   }

It seems this sort of unboxing experiment is not so far out of reach for a wrapper that could produce potentially very fast code. On top of that, as it is user-generated code it seems reasonable that it might be quite horrid to actually read. How feasible would it be to add a wrapper inspired by parsnip?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions