Coincidentally, just today I implemented the greedy packed trie algorithm from Liang's thesis on LaTeX hyphenation [1]. The idea is to store the trie into a bi

Coincidentally, just today I implemented the greedy packed trie algorithm from Liang's thesis on LaTeX hyphenation [1]. The idea is to store the trie into a big array, overlapping the storage of your child nodes with one another where possible. This reduces the memory footprint of the trie (when compared to using a separate hash table for each node) and increases the lookup speed as you don't need a hash table, you just index directly for each character. You can also collapse common suffixes into the same state, saving a bunch more space. Would be a nice enhancement to this algorithm; probably more so if it were written in C/C++. [1] http://www.tug.org/docs/liang/liang-thesis.pdf http://www.tug.org/docs/liang/liang-thesis.pdf