Assert is just replaced with workingMemory.assertOjbect ( object )
anyway. It's just a bit of regexp we apply to make things easier for users.
BTW this is what mandarax has to say about JSR94:
A: Mandarax does not (yet) support JSR-94
<http://www.jcp.org/jsr/detail/094.jsp>. The reason is that this
proposal is so general that it is almost useless. We are monitoring the
JSR-94 efforts, and perhaps add support later.
We support JSR94, but my feelings are just the same - just say NO! :)
Mark
Juergen wrote:
Questions on assert in RHS:
1) is it the same to call "special" assert() in RHS vs. indirectly
calling JSR 94 Rule session method addObject() (or
WorkingMemory.assert()) in RHS? would the latter way of asserting
interfere with the active rule execution?
rule "A"
when ...
then
assert(new String("A"))
someMethodCallsThisEnginesWorkingMemoryAssert(new String("B"))
someMethodCallsThisEnginesJSR94SessionAddObject(new String("C"))
end
2) what other special names/objects/methods do exist, that can be used
in a RHS code block (like special drools object in drools 2.5)?
Haven't found much on that in current docu.
My questions relate to the idea to wrapping assert with code to check
for equality.
Michael Neale wrote:
if there is enough demand for it, we may be able to make it some working
memory wide configuration option - certainly. Submit a JIRA request
if you
like, may come in handy (certainly will for you). But not as the
default, as
not everyone implements equals() correctly !
On 4/11/06, Juergen <[EMAIL PROTECTED]> wrote:
Quick'n dirty I could make drools 3 do it the way I would prefer - in
the context of my current problem and datastructures involved - by
modifying:
org.drools.reteoo.WorkingMemoryImpl, line 95f (rev. 3509):
replace:
/** Object-to-handle mapping. */
private Map identityMap
= new IdentityMap();
by:
/** Object-to-handle mapping. */
private Map identityMap
= new java.util.HashMap();
(not considering side effects, identityMap type cast problems, ...)
Then org.drools.reteoo.WorkingMemoryImpl, assertObject (line 422, rev.
3509):
FactHandleImpl handle = (FactHandleImpl) this.identityMap.get(
object );
would return first "A"s handle when the second "A" is asserted,
refering
to the example in my former posting.
I expected some way to customize the behaviour of the working memory,
basically by choosing between either of the Maps.
But as mentioned, such behaviour might not be desirable by others (they
probably want "A" being asserted twice and handled as two different
objects) and even if not customizable in drools itself, it can most
likely be done by wrapping assertObject and managing your own
object-to-handle map plus taking care not to use assert directly in
RHS.
I first wrongly thought logical assertions had something to do with
this, but I guess that relates more to the 'logical' conditional
element
in jess and logical dependencies among facts.
Michael Neale wrote:
what is your expectations on how it should work? (everyone has
different
ideas it seems !)
On 4/10/06, Juergen <[EMAIL PROTECTED]> wrote:
I mean
workingMemory.assertObject(new String("A"));
workingMemory.assertObject(new String("A"));
would currently assert both strings into working memory, whereas
if it
would check not for identityHashCode but with equals, the second
assert
would have no effect.
When during experimentation with rule engines I switched from jess
(where it is done with equals I think) to drools 2 I first
experienced
problems due to this differences (also did not find much in both
engines
docu), leading to my question if I could customize drools behaviour.
As one can always write a wrapper for drools assertObject to make
such
checks, and lots of other users probably dont need, my request is not
important.
Juergen
Michael Neale wrote:
so you mean
p1 : Person()
p2 : Person()
eval ( p1.equals(p2) )
(but presumably a shorthand form?)
On 4/8/06, Mark Proctor <[EMAIL PROTECTED]> wrote:
We hadn't planned it, not promising anything - but I'll see how
things
pan out, may be we will get time.
Mark
Juergen wrote:
ad. Can object equality for assertion be customized?
Drools 3 still seems to use IdentityMap with
System.identityHashCode
().
Is it planned to be customizeable e.g. with equals() in 3.0?
Michael Neale wrote:
Short answer, not easily with Drools 2. But yes, with Drools 3.
On 3/22/06, Juergen <[EMAIL PROTECTED]> wrote:
I recently looked into drools after experimenting with JESS
and got
a
few questions I could not look up in the drools documentation:
- Exists a condition to test for non-existence of a matching
object/fact?
-- If yes, how to use it for own domain specific language
conditions
- How is sharing of conditions/nodes between productions
implemented
in
drools, one of the main benefits of the rete algorithm?
-- How is the equality of conditions defined for java smf? via
textual
equality of the condition's java code?
-- How can equality of conditions for sharing be defined for
domain
specific language conditions?
e.g. Conway's game of life (slightly modified dsl)
<rule name="kill the overcrowded">
<conway:cellIsAlive cellName="cell"/>
<conway:cellIsOverCrowded cellName="cell"/>
<conway:killCell cellName="cell"/>
</rule>
<rule name="kill the lonely">
<conway:cellIsAlive cellName="cell"/>
<conway:cellIsLonely cellName="cell"/>
<conway:killCell cellName="cell"/>
</rule>
The conway:cellIsAlive condition could be shared between
these two
productions. Would it?
- When will the properties map be supported in drools
implementation
of
JSR 94 javax.rules API?
-- e.g. to be able to set conflict resolver of a rule base and
other
settings that can be set via drools native API
- Can object equality for assertion be customized?
-- Currently, org.drools.util.IdentityMap
(org.drools.reteoo.WorkingMemoryImpl) makes use of
System.identityHashCode(). There seems to be no way to customize
that,
e.g. make drools use equals() instead, neither via native nor
JSR
94
API.
My apologies if some questions have already been answered in
other
postings, I made no thorough search in this newsgroup yet.
Thanks a lot, Juergen
