Hello,
I am defining a node and a link and a “superclass” NodeOrLink. To lookup if a node or link exists i am using three predicates:
node_exists(Node) :- node(Node).
link_exists(Link) :- link(Link, _, _).
and the generic one:
node_or_link_exists(X):- node(X); link(X, _,_).
The latter is in the worst case two failed hash lookups in series. I am now wondering, what if I created a thread for each as so:
node_or_link_exists(X) :-
must_be(atom, X),
thread_create(node(X), Thread_1),
thread_create(link(X, _,_), Thread_2),
thread_join(Thread_1, Status1),
thread_join(Thread_2, Status2),
(Status1; Status2).
For a large kb with lots of nodes and links – would this be making sense to do to reduce hash lookup time?
Btw, if it does makes sense, then I am thinking that this could be improved – the thread_join wait for both to complete – but, if one completes with true, there is no need to wait for the other … not sure how to improve the code for that with the given thread API …
Edit:
Btw, another question is if there is anything gained - speed wise – if i unify node and link, into a new asserted fact – node_or_link/1 – creating a third lookup hash table – would the joined hash table lookups perform about the same as two run in parallel – i.e. anything significant to gain for the cost of another indexed fact.
Dan