Tutorials - Java-calls-Prolog
Assume we have a testing Prolog file with this content:
child_of(joe, ralf).
child_of(mary, joe).
child_of(steve, joe).
descendent_of(X, Y) :-
child_of(X, Y).
descendent_of(X, Y) :-
child_of(Z, Y),
descendent_of(X, Z).
You may wish to load this database into an interactive Prolog session to experiment with the predicates in this database before experimenting with JPL.
Consulting the Prolog database from its text file
In an ordinary interactive Prolog session, we would load the above Prolog database using the Prolog consult/1 predicate, a built-in predicate in standard Prolog. Note, however, that as a Prolog predicate, “calling” consult/1 is just an example of making a Prolog query, and this is how we perform it with JPL.
First, we construct an instance of Query, whose name is consult and whose arguments (just one) comprise the atom 'test.pl':
Query q1 =
new Query(
"consult",
new Term[] {new Atom("test.pl")}
);
Then, we call the hasSolution() method of this Query object, which returns a boolean value indicating its success:
System.out.println( "consult " + (q1.hasSolution() ? "succeeded" : "failed"));
At this point, this process may seem a bit long-winded; however, you should soon see that the classes are sufficiently general that they provide a robust and powerful interface into the Prolog engine.
Querying the database
Using the same technique, we can query the Prolog database about inferences it can make. To ask whether the Prolog query child_of(joe,ralf) is true, given the above Prolog database, for example, we write:
Query q2 =
new Query(
"child_of",
new Term[] {new Atom("joe"),new Atom("ralf")}
);
System.out.println(
"child_of(joe,ralf) is " +
( q2.hasSolution() ? "provable" : "not provable" )
);
To take an example that requires a bit more work on the part of the Prolog engine, on the other hand, we can ask whether descendent_of(steve,ralf) is true:
Query q3 =
new Query(
"descendent_of",
new Term[] {new Atom("steve"),new Atom("ralf")}
);
System.out.println(
"descendent_of(joe,ralf) is " +
( q3.hasSolution() ? "provable" : "not provable" )
);
Querying with variables
Ground queries (those without variables) like the above are relatively straightforward; they are essentially either provable or not, and there is typically no point in backtracking.
Once we use variables, however, things get a bit more complicated. Using the Variable class, we can construct a non ground query; and using other methods of Query we can obtain a solution in the form of a java.util.Map. If the Query has one or more solutions, then its Query.oneSolution() method returns a Map<String,Term> representing the first solution, otherwise it returns null:
Variable X = new Variable("X");
Query q4 =
new Query(
"descendent_of",
new Term[] {X,new Atom("ralf")}
);
java.util.Map<String,Term> solution;
solution = q4.oneSolution();
System.out.println( "first solution of descendent_of(X, ralf)");
System.out.println( "X = " + solution.get("X"));
The Map contains bindings in the form of Terms, each of which is indexed by its corresponding Variable in the Query.
Finding all solutions
The previous query finds only the first solution. Often, however, one wants all solutions, or at least more than just the first.
The Query class also provides the allSolutions() method, which returns an array of zero or more Map, each of which represents a given solution.
In this example we reuse the query q4, which was reset to its initial state by the call of oneSolution(), and instead call allSolutions(), which returns an array of solutions:
java.util.Map<String,Term>[] solutions = q4.allSolutions();
for ( int i=0 ; i < solutions.length ; i++ ) {
System.out.println( "X = " + solutions[i].get('X"));
}
If one wants not all but, say, at most n solutions, one could use method nSolutions(n). Again it will return an array of at most n solution bindings:
java.util.Map<String,Term>[] solutions = q4.nSolutions(20);
for ( int i=0 ; i < solutions.length ; i++ ) {
System.out.println( "X = " + solutions[i].get('X"));
}
Observe that even though we asked for at most twenty solutions, less could be returned since there may be less than such number of solutions.
Iterating through solutions
Equivalently, one can obtain each solution by exploiting the Iterator interface, which the Query class implements. This will allow a Java program to iteratively go through solutions one-by-one.
In this example, we iteratively call hasMoreSolutions() and nextSolution() to exhaustion:
System.out.println( "each solution of descendent_of(X, ralf)");
while ( q4.hasMoreSolutions()) {
solution = q4.nextSolution();
System.out.println( "X = " + solution.get("X"));
}
The hasMoreSolutions() method of the Query class returns a boolean, indicating whether there are any solutions “left” in the query. If the answer to this is ‘yes’, then the solution can be obtained in the form of a Map<String,Term> by the nextSolution() method.
As with any iterator in Java, it is safest to call nextSolution() only when it has been checked that there is indeed a “next” element in the iteration (via hasMoreSolutions()). Otherwise, we would run the risk of getting a java.util.NoSuchElementException exception.
In this final example, we reuse the previous variable X with a new variable Y in a new query q5:
Variable Y = new Variable();
Query q5 =
new Query(
"descendent_of",
new Term[] {X,Y}
);
while ( q5.hasMoreSolutions() ){
solution = q5.nextSolution();
System.out.println( "X = " + solution.get("X") + ", Y = " + solution.get("Y"));
}
Note: when using an iterator query, the query stays open until all solutions have been retrieved or the query is closed explicitly via the close() method. When open, the query is attached to a Prolog engine, and there are a finite number of them, so care must be taken in a multi-threaded application to avoid a deadlock situation. Please refer to the Multi-Threaded-Queries for details.
Also, check to check all the high-level API provided to access Prolog from Java, see the entry Types-of-Queries:-One-shot-vs-Iterative.