Question

I would like to use Parsec’s makeTokenParser to build my parser, but I want to use my own definition of whiteSpace. Doing the following replaces whiteSpace with my definition, but all the lexeme parsers still use the old definition (e.g. P.identifier lexer will use the old whiteSpace).

...
lexer :: P.TokenParser ()
lexer      = l { P.whiteSpace = myWhiteSpace }
   where l = P.makeTokenParser myLanguageDef
...

Looking at the code for makeTokenParser I think I understand why it works this way. I want to know if there are any workarounds to avoid completely duplicating the code for makeTokenParser?

Answer 1

Sadly, I don’t think there is a way. The local definitions used in makeTokenParser refer recursively to themselves, and so, as you’ve noted, lexeme uses whiteSpace as defined there, rather than the whiteSpace record member you replace in your lexer object.

The code is taunting because it uses the same names as both local functions in makeTokenParser and as record members of the TokenParser constructor. They are in fact totally distinct entities.