
In the uu-parsinglib we actually have two versions of parsers: lazy ones and strict ones, which have different types. So by giving a type annotation you can select the one you want. Notice that in the left- hand side of a monadic construct it does not make sense to use a lazy parser, since its result will be used as a parameter to the right-hand side operator, so in case of a monad our library system automagically selects the strict version for the left hand side. For the right hand side it depends on the type of the overall expression. Unfortunately in Haskell both the left and right hand side of a bind need the to be elements of the same monad, whereas in the case of a lazy oevrall parser this is not the case. We solve this problem by tupling the two parsers (NOT the parsing results), so still the do-notation can be used. The use of the library is free of any trickery! Doaitse Swierstra On 28 mei 2009, at 11:41, Malcolm Wallace wrote:
Henning Thielemann
wrote: I don't think that it is in general possible to use the same parser for lazy and strict parsing, just because of the handling of parser failure.
Polyparse demonstrates that you can mix-and-match lazy parsers with strict parsers in the different parts of a grammar (by choosing whether to use applicative or monadic style). You can also switch between lazy or strict interpretations of the applicative parts of your grammar (by changing the import that decides which version of the parser primitives is in scope).
I also used polyparse for lazy parsing, but I found it unintuitive how to make a parser lazy.
It can certainly be tricky, and requires a certain amount of experimentation. I think the difficulties are mainly due to the mix of lazy (applicative) and strict (monadic) styles in different non-terminals. A parser that you intend to be lazy, may turn out to be stricter than you hope, because of the strictness of another parser that it depends upon.
Regards, Malcolm _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe