Cheers prem !
Actually, as an initial thing to try, I would suggest keeping calling
.newWorkingMemory each time you want a clean WM - rather then clearing it.
WMs are lightweight objects and very quick to create. I think in your case,
as you are clearing the WM, it may actually be quicker to create it fresh
and discard the old one.
Give that a try, it may solve your problems (unless I have mis-interpreted
your requirements). Generally the reason to keep working memories is if you
want to keep the state in them (and keep adding facts over time), but in
your case, you are clearing it, so my suggestion is to get a fresh one (and
you don't really need a pool then - so less contention for the app server).
Michael.
On 1/22/06, prem <[EMAIL PROTECTED]> wrote:
>
>
> Hi drools team,
>
> First of all, drools is a great product and saved us lots of time, effort
> and
> money. Thanks.
>
> We have a web application in which drools is a key component. I have
> created a
> pool of WorkingMemories, since given any time, there could be multiple
> users
> requesting for the rules service. Whenever rules need to be run, a
> WorkingMemory
> is pulled out of the pool used and returned back to the pool. After
> allocating
> the WorkingMemory, before using it, old facts are cleared. It is done by
> the
> following method:
>
> public static void clearWorkingMemory(WorkingMemory wMem) {
> List factHandleList;
>
> factHahdleList=wMem.getFactHandles();
> for(FactHandle fh : factHandleList) {
> wMem.retractObject(fh);
> }
> }
>
>
> Now, when I assert a null fact the above method throws a
> NoSuchFactObjectException.
>
>
>
> Besides, some of my conditions actually check for the fact to be null and
> act
> appropriately. This is because a fact being null has a meaning by itself.
> These
> conditions are not even evaluated when the fact is asserted as null.
> Without
> understanding this behaviour, I got around it by creating FactWrapper
> bean.
>
> After pondering little while I realized that when a null fact is asserted
> into
> the working memory, there is no information to ascertain which fact is
> being
> asserted. [when a fact is non-null, by reflection the class and the
> properties
> can be ascertained. This can be compared against parameter declaration].
> Which
> actually means that asserting a null fact can have no effect on the rule
> execution. If that is the case, I feel that it will be better to prevent
> asserting null facts. This will save time for so many. Do I have a point
> here
> or am I missing something??
>
>
>
> Regarding the NoSuchFactObjectException, I dug into the code and found
> that
> org.drools.reteoo.WorkingMemoryImpl.retractObject calls
> WorkingMemoryImpl.getObject(). getObject() when it gets a null from the
> PrimitiveLongMap assumes that the object is not found and throws the
> exception.
> This should have been looking for objects.containsKey() and if this
> returns
> false, then throw NoSuchFactObjectException. I checked PrimitiveLongMap
> and
> found that even containsKey does [return get(key)!=null]. This can be
> fixed by
> adding a getInternal() private method like this:
>
> <<<<<PrimitiveLongMap change start>>>>>
> Existing code: org.drools.util.PrimitiveLongMap.java
> public Object get(long key)
> {
> if ( key > this.maxKey || key < 0 )
> {
> return null;
> }
>
> Object value = findPage( key ).get( key );
>
> // NULL means the key exists, so return a real null
> if ( value == NULL )
> {
> value = null;
> }
> return value;
> }
>
> public boolean containsKey(long key)
> {
> if ( key < 0 ) return false;
>
> return get( key ) != null;
> }
>
> Can be changed to:
>
> public Object get(long key)
> {
> Object value = getInternal(key);
>
> // NULL means the key exists, so return a real null
> if ( value == NULL )
> {
> value = null;
> }
> return value;
> }
>
> private Object getInternal(long key)
> {
> if ( key > this.maxKey || key < 0 )
> {
> return null;
> }
>
> return findPage( key ).get( key );
> }
>
> public boolean containsKey(long key)
> {
> if ( key < 0 ) return false;
>
> return getInternal( key ) != null;
> }
> <<<<<PrimitiveLongMap change end>>>>>
>
>
> <<<<<WorkingMemoryImpl change start>>>>>
> Existing code: org.drools.reteoo.WorkingMemoryImpl.java
> public Object getObject(FactHandle handle) throws
> NoSuchFactObjectException
> {
> Object object = this.objects.get( ((FactHandleImpl) handle).getId(
> ) );
>
> if ( object == null )
> {
> throw new NoSuchFactObjectException( handle );
> }
>
> return object;
> }
>
> can be changed to:
> public Object getObject(FactHandle handle) throws
> NoSuchFactObjectException
> {
> long key = ((FactHandleImpl) handle).getId( );
>
> if(! objects.containsKey(key))
> {
> throw new NoSuchFactObjectException( handle );
> }
>
> return this.objects.get(key);
> }
> <<<<<WorkingMemoryImpl change end>>>>>
>
>
>
> To put it different way, the above fix will prevent
> NoSuchFactObjectException
> being thrown in the code below;
>
> factHandle=wMem.assertObject(someFactObject); // someFactObject is null
> wMem.retractObject(factHandle); // will throw NoSuchFactObjectException
>
>
> Please check if this is making sense...
>
> Thanks again and keep up the great work..
> Cheers
> prem
>
>
>
