Hi,
I am still not sure I understand the semantics -- the constraint is
posted with domain consistency (the default). My understanding of domain
consistency is that any values remaining in the domain is consistent
with the constraint, i.e. the constraint should not fail if the variable
is instantiated to one of the values, as long as this is the only change
made. For
Y :: [3,4], count([3,0,3,3,4], Y, =, 2)
Y is the only domain variable with more than one value, so if I
understand domain consistency correctly, then instantiating it to either
3 or 4 should succeed, but doing so fails for both values.
The documentation states that:
Supports domain consistent propagation (icl = ICL_DOM, default) and
"almost" domain consistent propagation (all other values for icl), where
y is not pruned.
(I assume by other values of icl, you exclude ICL_DEF, which behaves as
ICL_DOM as that is the default)
It seems to me the behaviour is that y is not pruned in the above, even
when the constraint is posted with ICL_DOM and ICL_DEF -- I have
explicitly posted the constraint with ICL_DOM, to check if it makes any
difference, and it does not, i.e. the example with Y::[3,4] succeeds.
Cheers,
Kish
On 22/11/2011 08:49, Christian Schulte wrote:
This is a fundamental missunderstanding of what an integer set and an
integer variable denotes.
It might be that the current domain of an integer variable is a set, however
the variable denotes a single integer (which you do not know yet). That's
fundamental and very different from a set.
With that in mind, count behaves exactly like one would expect.
In your example, the constraint does not succeed (as in being entailed). It
just does not fail yet. Let's take the more general example (2 is replaced
by an integer variable Z):
Y :: [3,4], count([3,0,3,3,4], Y, =, Z)
Then it will propagate that Z has the domain [0..4] as there are four
elements in the array that still could be equal to Y with domain 3 or 4 and
one is already known to be different (the element 0) from 3 or 4.
Christian
--
Christian Schulte, www.ict.kth.se/~cschulte/
-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf
Of Kish Shen
Sent: Tuesday, November 22, 2011 8:09 AM
To: [email protected]
Subject: [gecode-users] semantics for count(Home home, const IntVarArgs&x,
IntVar y, IntRelType r, IntVar z)
Hi,
I have been looking at the behaviour of
count(Home home, const IntVarArgs&x, IntVar y, IntRelType r, IntVar z)
and cannot understand what is happening when IntVar y has more than one
value in its domain. Until now, I have only tested this constraint with y
having a singleton value in its domain, and in this case, the constraint
behaves the same as the case where y is Int, i.e. the number of occurrences
of the value y in x is counted.
However, one of my user wanted to use the constraint where y has multiple
values (i.e. the counts constraint in the global constraint catalog). I know
this constraint is provided with y being an IntSet, but my user did not know
about this constraint, so he used the version where y is a IntVar instead.
I had expected that mulitple domain values in y would behave like multiple
values in the IntSet for this constraint, but this does not seem to be the
case. For example, the constraint succeeds with the following (in ECLiPSe
syntax, but the argument order of Gecode's
constraint):
Y :: [3,4], count([3,0,3,3,4], Y, =, 2)
where [3,0,3,3,4] is x, IntRelType is IRT_EQ.
I don't understand why this succeeds -- the count for 3 is 3, and for 4 is
1, and for both 3 and 4 is 4, none of which is equal to 2. So what is the
semantics?
The equivalent call using y as an IntSet [3,4] fails as expected. Also, for
the same count constraint, if y has a single domain value (e.g. 3 or 4),
then the constraint behave as I expected, i.e. z = 3 for y = 3, and z = 1
for y = 1.
Thanks and cheers,
Kish
--
This e-mail may contain confidential and privileged material for the sole
use of the intended recipient. Any review, use, distribution or disclosure
by others is strictly prohibited. If you are not the intended recipient (or
authorized to receive for the recipient), please contact the sender by reply
e-mail and delete all copies of this message.
Cisco Systems Limited (Company Number: 02558939), is registered in England
and Wales with its registered office at 1 Callaghan Square, Cardiff, South
Glamorgan CF10 5BT.
_______________________________________________
Gecode users mailing list
[email protected]
https://www.gecode.org/mailman/listinfo/gecode-users
--
This e-mail may contain confidential and privileged material for the
sole use of the intended recipient. Any review, use, distribution or
disclosure by others is strictly prohibited. If you are not the intended
recipient (or authorized to receive for the recipient), please contact
the sender by reply e-mail and delete all copies of this message.
Cisco Systems Limited (Company Number: 02558939), is registered in
England and Wales with its registered office at 1 Callaghan Square,
Cardiff, South Glamorgan CF10 5BT.
_______________________________________________
Gecode users mailing list
[email protected]
https://www.gecode.org/mailman/listinfo/gecode-users
