Re: Bison and C++

[Robert Anisko]

> >...STL stack might NOT be a good idea if
> >the bison algorithm has to look forward and backward in the stack; I
dunno
> >if it allows that...
>
> The standard LR(1) state machine does not look deeper into the stack, but
> the DR  does (I take it from the description given in the Bison lists).
> Then just use say std::vector instead. The std::stack is just an adaptor
> class.
>
> Or it should be easy to merely require the use of an abstract class A,
> which the user specifies. For example, the LR(1) algorithm would require A
> to have A::push_back(), A::pop_back(), etc., and some other parser
> algorithm may require extra class functions

This is true if you only consider parsing, but we still need a data
structure with random access for semantic actions.