I am now trying to use the assoc library in sw-prolog to create a relational prolog program. I need a dictionary that can store multiple values. This can be easily accomplished using a list as value.
However, i often need to look up across all keys, single values -- i.e. identify all keys that have included one particular value item. (However, the number of value is typically limited to at most tens of items).
If, i were to only use a association store that maps from keys to multiple values, then i would need to go through all keys, fetch the values and then check if a value is its member.
This is not efficient, nor does it seem scalable with large numbers of keys.
Is there a more efficient way to do this?
I could manage for each value item its own "inverse" key-value store, but this seems pretty inefficient. Also, if i were to use prolog fact store (with assert/ and retracts), to store each key value pair as an individual fact, i would probably get these lookup "indexed" for free.
What is the right design choice for a relational approach here?
[edit-1]: btw, one "solution" could be to allow in the assoc library to have multiple keys, in case values are then internally "indexed" -- but, the assoc library doesn't support this (as far as i know).
[edit-2]: look at additional libraries in swi-prolog i notice unweighted graphs. Perhaps using a graph structure would be more efficient for handing such a many-to-many relation between keys and values. And, in keeping with a relational approach.
[edit-3): or perhaps, ugraph, is not a good idea, since nodes (and links) are not typed. I would need to structurally indicate which node is a key and which are values.
thank you,
Daniel