From: Robert Dober <robert.dober@...>
Date: 2007-07-20T21:17:47+09:00
Subject: Re: The Pebble in the Ruby Shoe

On 7/20/07, Logan Capaldo <logancapaldo@gmail.com> wrote:
> On 7/19/07, M. Edward (Ed) Borasky <znmeb@cesmail.net> wrote:
> >
> > Logan Capaldo wrote:
> >
> > > You've never written code to traverse some kind of tree? And if you
> > have,
> > > you did it all iteratively? I would suggest _lots_ of programmers use
> > > recursion, even in C.
> >
> > There's a big difference between an explicit stack with iterative code
> > and recursion with an implicit stack. I've done a lot of explicit stack
> > coding for walking data structures -- indeed, the last time I did so was
> > a recent piece of code that generates the state space of a large finite
> > Markov chain. But the only time I've written recursive functions as a
> > *preferred* modus operandi has been in Lisp, Scheme and the Lisp-based
> > symbolic math package Derive. And there, I've always aimed for tail
> > recursion.
> > >
>
>
> Interesting. This is wild speculation of course, but I imagine you are
> probably atypical as far as always using explicit stacks. (Since I get the
> impression from your post history that you are heavily focused on
> performance.). I would guess (more speculation ;)) that if someone were to
> grab a set of programs that for instance walked an XML DOM (or a similar day
> to day tree structure, like a nested directory) that the vast majority of
> those programs would use recursion with implicit stacks. Now that I've
> contributed that bit of unsubstantiated nonsense, I'll resume eating my
> dinner. :)
>
I remember a funny use case of explicit stack handling now.
About 4 years ago I was writing an XML wrapper around the Python tupleparser.
I started recursively, of course and very soon got completely lost, I
tried to undo the recursion so that I had full control over some
details on the call stack (things like cheating by examining the call
stack) never got anywhere but I have to admit it was good to have that
technique at my disposal.
If I recall correctly I would probably have used a recursive approach
with yield in Ruby...

But I agree these are rare exceptions and no reason at all not to use
recursion, a better design might have allowed for a recursive solution
too :)

Cheers
Robert


-- 
I always knew that one day Smalltalk would replace Java.
I just didn't know it would be called Ruby
-- Kent Beck