Stackless & String-processing

[Greg Ewing]

Tim Peters wrote:
> 
> In a way deeper than it is in a functional program not using monads?  IOW,
> is the clarity of "control flow" due to monads specifically or just to
> functional semantics in general?

Monads are really only a big deal in the context of
functional languages, where there is no such thing
as control flow. Monads provide an astoundingly
elegant way of writing what is effectively procedural
code in a functional language. If you've ever used
one of the pre-monad systems for doing I/O in a
functional language, you'll appreciate what a
massive achievement this is.

However, Python is already a procedural language,
so there's no need to use monads there to fake
procedural code. It would be as pointless as
putting training wheels on a tricycle.

With regard to parser combinators, there are various
things at work in the functional language context that
don't apply to Python.

Using monads and parser combinators in a functional
language, you can write parsers that look very much
like a BNF grammar or whatever your favourite notation
is -- *without* needing any special tool to compile
and execute them.

This is largely due to the general sparseness of the
syntax, and the ability to define new infix operators,
neither of which apply to Python. You can't translate
the code into Python without losing most of the notational
convenience which accounts for a large part of the
technique's attractiveness.

The idea of composing parsing functions can still
be useful, however. For instance you might have a
function with the signature

  def p_separated_list(p_item, p_separator):

where p_item and p_separator are parsers for
list items and separator tokens. Which incidentally
could be an answer to

> What I haven't seen is an example of something clearly useful
> (wrt parsing specifically) *beyond* those.

Note that the implementation of p_separated_list
need not be quite the standard version you might
find in a library. In a parser for Python code, for
example, it would take care of that quirky comma-at-
the-end-of-a-sequence thing.

> I will give parser combinators a try; they do sound interesting to me.  More
> interesting would be to hear about that "new programming language"!

If you're interested in examples of monad use for
parsing and compiling, I had a fit of insanity last
year in which I implemented a compiler in Hugs using
monads for several different things, including parsing.
If you'd like to see it, I could dig it out and put it
on a web page somewhere.

Greg