SEARCH DATA STRUCTURES FOR SKEWED STRINGS

We consider skewed distributions of strings, in which any two such strings share a common prefix much longer than that expected in uniformly distributed (random) strings. For instance, this is the case of URL addresses, IP addresses, or XML path strings, all representing paths in some hierarchical order. As strings sharing a portion of the path have a quite long common prefix, we need to avoid the time-consuming repeated examination of these common prefixes while handling the linked data structures storing them. For this purpose, we show how to implement search data structures that can operate on strings with long prefixes in common. Despite the simplicity and the generality of the method, our experimental study shows that it is quite competitive with several optimized and tuned implementations currently available in the literature.