Re: ICFP Writeup

["John R. Hall" <overcode@cc.gatech.edu>]

My entry, the Midnight Markup Masher (#144 in the list), used MzScheme but 
bugged out (dropped some attributes for some reason). Grrr.
It was fun, though.

My biggest problems were a) starting on it about halfway into the contest,
  b) trying to write it in a language (Scheme) that I hadn't used for a 
long time (it proved a pleasant coding experience, but I was rusty), and c)
  almost no testing. The SML tree code turned out well, and might actually 
be useful in future projects (I'm not entirely sure why I didn't use PLT's 
existing XML code, though). Some of my initial optimization strategies 
didn't work out. I had two optimizers: a list optimizer (which would 
attempt to discard redundant tags, etc from a flat list rather than a tree)
, and a tree optimizer (which would attempt more complex optimizations). 
These two fed into a rewriter, which would generate SML output without 
extra whitespace, etc. The list optimizer basically worked, and I think I 
got color nesting optimizations to work, but a large chunk of the 
optimization code completely failed, so I had to leave it out. I'm sure it 
was a simple bug, but I was out of time at that point.

Live and learn (and debug next time), I guess. This is a really cool 
contest. I stopped by the Python channel on irc.openprojects.net just to 
see how their entries were going, and even though I don't know Python, we 
had a good chat about the evils of optimizing markup.

-John

On Tuesday, August 28, 2001, at 08:22 AM, Shriram Krishnamurthi wrote:

> Folks --
>
> It's really cool that four unrelated plt-scheme members got together
> to try out for the contest.  The usual suspects, yes, but still very
> cool.  Too bad it didn't work out.  Kick our butts next time!
>
> For what it's worth, one (PLT) Scheme entry did survive the
> elimination rounds: Brownian Code Motion.  See
>
>   http://www.cs.brown.edu/research/plt/Notes/icfp2001-contest.txt
>
> What we don't mention in that we had to CPS some of the code to handle
> the deep recursion.  We noticed, for what it's worth, that the
> Objective CAML people had to do this too!
>
> Shriram
>

--
John R. Hall |  Overfull \brain has occurred while \learn is active
CS Student - Georgia Tech  |  Author - Programming Linux Games
Unsolicited commerical e-mail will be considered an act of war.