I am looking for a content addressing data structure

Question

I'm trying to design a data structure that allows efficient extraction of entries from a part of their content.

Let's say I am looking for an entry that matches this: [ x 2 3 xx ]

If [ 0 2 3 4 5 ] or [ 3 2 3 7 8 ] are in my data structure, they should be returned by my find function.

I wrote such a data structure where I compare the "pattern" against all entries of the data structure, but of course it takes way too much time. I have few ideas about how to do this a faster way but they are quite heavy to implement. Does something like this already exists? If not, how would you do this?

Answer 1

The first thing that comes to mind is to have a hash table for each position in the tuple. For searching you intersect the results for all the positions with a specified value.

Answer 2

Well, a Suffix tree could do it in O(1), but it would take a LOT of memory.

Answer 3

What you're trying to implement looks a lot like tuple space .

Answer 4

You might take a look at the RETE algorithm. This general problem came up in the AI systems of the 70s; Chapter 14 in Paradigms of AI Programming covers it. (IIRC there's mainly elaborations of starblue's answer and decision trees.)