[rules-users] Tuples discrimination

[Eymeric VIEUILLE]

Hi !

After solving the problem of Out of Memory, I come back with another 
question :

I currently implemented rules that give me tuples like this (in the 
agenda, before firing rules) : M1xM2xM3xM4.


But now, I want to select THE tuple that as M1xM2xM3xM4 with a property 
p as little as possible to execute the right hand of the rule.

For sample, I have the two following tuples activated in the agenda :

M'1(date = "09/25/2006")      M'1(date = "09/25/2006")
M'2(date = "09/26/2006")      M'2(date = "09/26/2006")
M'3(date = "09/29/2006")      M"3(date = "09/27/2006")
M'4(date = "09/31/2006")      M"4(date = "09/30/2006")

I would like the second tuples to be selected for executing the right 
hand of the rule because dates of M'2, M"3, M"4 are the closest dates 
compared to M'1.

The problem is that I do not know on what layer this selection must be 
made : in the Left Hand of the rule ? Or in a AgendaFilter (selecting 
the better tuple to use) ?

For the LH, I thought to the "accumate" CE, but I am affraid to have 
unreadle rules (impossible to maintain). And I don't see quite well how 
to implement this with "accumulate" CE (knowing that I use a DSL).

For the AgendaFilter, I saw that only the current activation is 
accessible...so impossible to compare whith other tuples 
values....Moreover, I am affraid to make performance fall.

Is this selection possible ? And what would be the best practice to 
implement it ?

Regards,
Eymeric.