Stackless & String-processing

[Fernando Pereira]

In article <7mssne$m8r$1 at cronkite.cc.uga.edu>, Graham Matthews
<graham at sloth.math.uga.edu> wrote:
> You have totally misunderstood what I said. I said "you can have regular
> expression parsers as primitive parsers". There is no need for the compiler
> to "solve the halting problem". The programmer is free to use regular
> expression/automaton techniques where he/she wants. Such parsers take
> strings and return ASTs. They can then be combined with other parsers
> as if they were primitive parsers.
One last try. I understood your point from the beginning. However, you
did not understood *mine*. If you push the complex
pattern-matching/parsing machinery into black boxes, there's not much
for the monadic glue to do, is there? And even then, some
optimizations, those involving the various black boxes, will be
blocked. Here's a simple example. Assume that there are two black boxes
A and B, and that you use monadic glue to say "A or B". A and B may be
deterministic, but the result will not be, and exhaustive search will
be required in the glue. Sure, I can recognize such cases by hand and
create a new black box "A or B" which is optimized, but that is an
error-prone hand optimization. And one may not notice subtler
inefficiencies.

-- F