The WM is thread safe, on all insert/update/delete we apply a lock
making sure that not more than one operation can happen. However we do
not apply this lock when the iteraters are used externally via users -
so while iterating any working memroy actions should be avoided. For the
minute I think you would have to do it in two stages, iterate and store
the objects you wish to change in a list. Then iterate that list and
remove/modiy them via the normal working memory api. You can't remove a
fact via unsafe code, the workingmemory.retract(...) is thread safe.
Mark
The iterators do not have "remove" so cannot damage the internal "assert
table" but if you change an underlying object and don't notify
Michael Anstis wrote:
Excuse me if this would be best posted to the developer list...
...but let me undertstand: The Fact table maintained for a
WorkingMemory is not thread safe(?); however WorkingMemories are
thread safe - suggesting Shadow Facts and (excuse my vague terminolgy)
Node Memories are the thread safe aspect of working memories - the
Fact table containing references to the original inserted objects not
the Shadows. So if a Fact is removed using my "unsafe" code it will
cause Shadow Facts to become disconnected from their original
non-shadowed Fact? Would using ThreadLocal storage for the Fact table
(together with node memories and shadow facts) mean a complete
thread-safe Drools? I don't mean this to become a tutorial in either
Drools threading or general java thread safety!! (I don't think
there's a category on the "how to have your emails ignored" listing
for this). I just want to increase my awareness of (Drools) threading
issues... and this proves a good example.
On 05/10/2007, *Mark Proctor* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
Anstis, Michael (M.) wrote:
Wouldn't the best approach be to get the FactHandles iterator and retact
them from working memory rather than removing them through the iterator?
Iterator itr = wm.iterateFactHandles();
While(itr.next()) {
FactHandle h = itr.next();
wm.retract(h);
}
This would ensure truth maintenance is preserved.
Actually I'm not sure that would be safe.... the objects and the
handles are in the same hashtable. Those internal data structures
where built for performance and lightweightness, not thread
safeness and mutability. If you actually look we have an internal,
fast, iterator which we simple adapt to a slower
java.util.Iterator. At the moment none of our iterators are thread
safe, but I do see a valid use case here, we will have to think on
it for the next major release - cleam implementation patch welcome
with unit tests :) I'm less concerned about the iterator adapter
performance, but I cannot compromise on the performance of our
internal iterators.
public static class HashTableIterator
implements
Iterator {
private static final long serialVersionUID = 400L;
private AbstractHashTable hashTable;
private Entry[] table;
private int row;
private int length;
private Entry entry;
public HashTableIterator(final AbstractHashTable hashTable) {
this.hashTable = hashTable;
}
/* (non-Javadoc)
* @see org.drools.util.Iterator#next()
*/
public Object next() {
if ( this.entry == null ) {
// keep skipping rows until we come to the end, or
find one that is populated
while ( this.entry == null ) {
this.row++;
if ( this.row == this.length ) {
return null;
}
this.entry = this.table[this.row];
}
} else {
this.entry = this.entry.getNext();
if ( this.entry == null ) {
this.entry = (Entry) next();
}
}
return this.entry;
}
/* (non-Javadoc)
* @see org.drools.util.Iterator#reset()
*/
public void reset() {
this.table = this.hashTable.getTable();
this.length = this.table.length;
this.row = -1;
this.entry = null;
}
}
-----Original Message-----
From: [EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>
[mailto:[EMAIL PROTECTED] On Behalf Of Godmar Back
Sent: 05 October 2007 16:40
To: Rules Users List
Subject: Re: [rules-users] does WorkingMemory.iterator support remove()?
Thanks - consider supporting it for efficiency. (Otherwise, removing a
set of facts from working memory requires a temporary container to
hold the facts to be removed.)
- Godmar
On 10/5/07, Mark Proctor <[EMAIL PROTECTED]> <mailto:[EMAIL PROTECTED]>
wrote:
Godmar Back wrote:
Does the iterator returned by WorkingMemory.iterator support remove()?
I checked the javadoc and the Drools manual, but may have missed it.
If you try it you'll get an exception thrown
"UnsupportedOperationException"
http://anonsvn.labs.jboss.com/labs/jbossrules/trunk/drools-core/src/main/jav
<http://anonsvn.labs.jboss.com/labs/jbossrules/trunk/drools-core/src/main/jav>
a/org/drools/util/JavaIteratorAdapter.java
Please answer and augment documentation.
- Godmar
_______________________________________________
rules-users mailing list
[email protected] <mailto:[email protected]>
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
[email protected] <mailto:[email protected]>
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
[email protected] <mailto:[email protected]>
https://lists.jboss.org/mailman/listinfo/rules-users
------------------------------------------------------------------------
_______________________________________________
rules-users mailing list
[email protected] <mailto:[email protected]>
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
[email protected] <mailto:[email protected]>
https://lists.jboss.org/mailman/listinfo/rules-users
<https://lists.jboss.org/mailman/listinfo/rules-users>
------------------------------------------------------------------------
_______________________________________________
rules-users mailing list
[email protected]
https://lists.jboss.org/mailman/listinfo/rules-users
_______________________________________________
rules-users mailing list
[email protected]
https://lists.jboss.org/mailman/listinfo/rules-users
