This project let's you generate dense word boards.

First run:

./test.sh

Wait for "Done reading dictionary" info report

Then type this in the console:

wb_generator:print_boards([
  {rows, 5},
  {columns, 5},
  {grid_count, 5},
  {rounds, 300},
  {target_score, 500},
  {repeats, 2}
]).

This will keep the top 5 (grid_count) grids between each round
from ten boards generated using a random perturbation.

If after 300 rounds the target score has not been reached
(ie one of the boards does not have over 500 words on it)
then the attempt is repeated.

If after two attempts a board has still not been generated,
the algorithm gives up.

See: http://ai.stanford.edu/~chuongdo/boggle/index.html
For a better description of the algorithm.

I am new to erlang, and this algorithm runs slower than the c version
it also does not produce boggle boards as dense, but it can get into
the thousands of words for 5x5 boards, whichs is quite dense.

The reason for the slowness is because ets table lookups for trie
are much slower than native c trie structures.

The reason for less density in the boards is most likely because
I have not implemented recording results of perturbations either,
and each perturbation is truly random instead of being weighted on
past success rates.