From: Adam Shelly <adam.shelly@...>
Date: 2008-04-20T03:04:36+09:00
Subject: Re: [SUMMARY] Word Search Generator (#159)

On Fri, Apr 18, 2008 at 2:38 PM, Matthew Moss <matthew.moss@gmail.com> wrote:

Thanks for the nice writeup.
>
>  My initial thought was that Adam's overlap code would provide more
>  interesting puzzles than my very rough generator, as mine had no
>  metrics or heuristics beyond "Does it fit?" But there was very little
>  difference between Adam's output and my own. I have two suspicions.
>
>  First, Adam checks only the first character of one word against all of
>  the characters of another. I think that might have been two limiting,
>  and to my recollection, word search puzzles tend to have intersections
>  in the interior of words, and not at the ends.
>
>  Second, I suspect that a simple two-word intersection test like Adam's
>  won't be sufficient for interesting puzzles.

I intended to do a more thorough intersection test, but that complicated
the code for finding a valid start point, and I ran out of time...
But your suspicion is right, I had a chance to try the deeper test today,
and it didn't significantly improve the number of intersections or
reduce the parallelism.

I think three-word
>  (triangle), and perhaps four-word (rectangle), overlap tests might be
>  required. These are common patterns in word search puzzles, and the
>  triangle pattern would better fend off the uninteresting parallel
>  groupings.
>

Sounds like an interesting algorithm to research... someday.

-Adam