Prefix searches and range queries in a large corpus with a lot of shared prefixes - a file system listing cache - have been my primary reason for using tries.
You can answer questions like "how many files does this directory have, immediately in it" and also "under it recursively" relatively quickly, and if you use a Patricia trie with a little cleverness in the compression of the text, it needn't take up much space at all even for millions of paths. Whereas a binary tree isn't so great at the memory saving bit.
You can answer questions like "how many files does this directory have, immediately in it" and also "under it recursively" relatively quickly, and if you use a Patricia trie with a little cleverness in the compression of the text, it needn't take up much space at all even for millions of paths. Whereas a binary tree isn't so great at the memory saving bit.