There is only one object of class ArrayList<Cheese>. Bound to $cheeseList is the "live" list used in the rete machinery, and this reference is stored in the global "results" (List<ArrayList<Cheese>>, at 0). Due to more cheese being put on the table, the list fattens (small wonder!), and Java code "sees" the growing list.

The second set of assertions was updated by Mark to use 3 for the size of the list, and does not fail now.


On 27 May 2011 22:47, Michael Anstis <> wrote:
Where does it now fail?

I'd assume the first part of the test (where only 1 Cheese exists) would still pass, but the second part of the test where 3 Cheeses existing in total to now fail.

Is my understanding correct?



On 26 May 2011 05:44, Mark Proctor <> wrote:

Previously ALL collections where cloned when bound:

public Object getNonShadowedValue(InternalWorkingMemory workingMemory,
                                      final Object object) {
        Object result = this.readAccessor.getValue( workingMemory,
                                                    object );
        if ( this.isInternalFact() && result instanceof Collection ) {
            try {
                Collection newCol = (Collection) result.getClass().newInstance();
                for ( Iterator it = ((Collection) result).iterator(); it.hasNext(); ) {
                    Object element =;
                    newCol.add( element );
                return newCol;
            } catch ( InstantiationException e ) {
                // nothing we can do, so just return the resulting object
            } catch ( IllegalAccessException e ) {
                // TODO Auto-generated catch block
        return result;

This was left over from shadow facts and actually still existed today. I have removed the code but now we have a failing test in FirstOrderLogicTest.testCollectResultConstraints.

rule "Collect Test" salience 70
        $cheeseList  : ArrayList(size == 1) from collect( Cheese( ) );

The reason why this passed before was the collection was cloned, so the accumulate's evaulations would not impact it. Now it is no longer cloned and while the rule does not fire the Collection is updated. I think the later is the correct behaviour and I've updated the test as expected. Everyone agreed?

rules-dev mailing list

rules-dev mailing list