[gecode-users] access to Space during branching?

[Luis Quesada]

Christian Schulte wrote:

>Of course you can do that! However: how do you know what are the views?
>Views are very cheap (just a pointer) so there should be no need to put
>anything globally into the space.
>  
>
OK. I see. The point is that we want the propagator to be re-usable. Of 
course, in the current case, I know what the views are. In general I 
don't know.
Thanks for the explanation.
Luis

>Christian
>
>-----Original Message-----
>From: Luis Quesada [mailto:luque at info.ucl.ac.be] 
>Sent: Monday, June 05, 2006 12:32 PM
>To: Christian Schulte
>Cc: 'gecode user list'
>Subject: Re: [gecode-users] access to Space during branching?
>
>
>Christian Schulte wrote:
>
>  
>
>>No, you always have to update views! If a propagator stores a view it 
>>must be updated. In the situation you describe it will be that the 
>>variable implementations to which the views stored in the space have 
>>already been updated (that is, a copy has been created). However, also 
>>the views in the propagators must be updated to point to the copied 
>>variable implementations.
>>
>> 
>>
>>    
>>
>OK, I see your point! However, do the propagators (in the particular 
>situation I am describing) need to store the views (taking into account 
>that the space is an argument (and that I can have access to the views 
>through the space))?
>
>Luis
>
>  
>
>>Christian
>>
>>-----Original Message-----
>>From: Luis Quesada [mailto:luque at info.ucl.ac.be]
>>Sent: Monday, June 05, 2006 12:01 PM
>>To: Christian Schulte
>>Cc: 'gecode user list'
>>Subject: Re: [gecode-users] access to Space during branching?
>>
>>
>>Christian Schulte wrote:
>>
>> 
>>
>>    
>>
>>>Yes, you got it right. It has been really a mistake to not pass a 
>>>space
>>>argument to branch: everything is constructed in such a way that one 
>>>never has to store a space anywhere: no Gecode data structure stores a 
>>>space, all is passed as arguments.
>>>
>>>
>>>
>>>   
>>>
>>>      
>>>
>>Thanks for your confirmation.
>>
>>In fact, this made me remember another question: suppose that in my
>>program all the propagators are using the same views of the variables 
>>(i.e., there is only one view per variable) and theses views are updated 
>>in the constructor for cloning of the space. Is it correct to say that 
>>there is no need of updating these views in the constructor for cloning 
>>of the propagators (since these views are already updated in the 
>>constructor for cloning of the space)?
>>
>>Luis
>>
>> 
>>
>>    
>>
>
>
>  
>


-- 
Luis QUESADA
Catholic University of Louvain 
Department of Computing Science and Engineering 
Place Sainte Barbe, 2 
B-1348 Louvain-la-Neuve, Belgium 
Phone:	(++32) (10) 47 90 13 
Fax: 	(++32) (10) 45 03 45 
Web: 	http://www.info.ucl.ac.be/~luque