Re: [OWLAPI-developer] reasoners and ontologies

[<phi...@ne...>]

Ignazio Palmisano <ipa...@gm...> writes:
>> I want to store a reasoner object so that I can reuse it everytime I
>> query for consistency, to avoid re-reasoning everytime.
>>
>> My initial idea was to put the reasoner into a WeakHashMap keyed on the
>> ontology. This seemed like a reasonable idea to my, but it's not
>> working.
>
> You need to deregister the reasoner from the ontology change listeners,
> otherwise your key will be referred by its value and then never garbage
> collected.


Ah. Okay, that's a problem. The reasoner is strongly referenced from the
hash, so if it's strongly referencing the ontology, then we are stuck.  

Had a go at fixing this, by putting the reasoner on a weak reference
also, but it doesn't help, the reasoner still will not GC. Perhaps it
being held onto by the GUI. 

It looks like this is not a profitable way forward; I guess I was just
being lazy, and I will have to hook into the new ontology machinary. 




>> The problem is, I think, from two things. Firstly, something, somewhere
>> appears to be hanging onto a reference for the ontology. I'd hard for me
>> to be sure that this is happening, but once in the hashmap, the ontology
>> never disappears.
>>
>
> Yep that's the symptom. Maybe the change listeners should be held with weak
> references.

This isn't normal practice, although, perhaps it should be. 

>
>> The second problem is that the hashCode for an ontology is keyed on the
>> identifier for the ontology...
>> @Override
>>
>>         public int  hashCode() {
>>                 return ontologyID.hashCode();
>>         }
>>
>>
>> So, if you have two ontologies with the same IRI they will appear to be
>> equal, even if they have different axioms in them.
>>
>> The use case here is that, periodically, I want to throw an ontology
>> object away and start from fresh, to make sure that everything is up to
>> date with my source. I do this by just creating a new ontology object.
>>
>
> Wouldn't this replace the previous ontology in the map?


No. I check the map first to see if it already exists. Unfortunately,
according to hashCode and .equals it does already exist, if it it has
totally different numbers of axioms. 



> I have my doubts that the way hashcode and equals work for ontologies is
> the best one, but I think there is a lot of code relying on that behavior,
> so I'm a bit reluctant to change it.
>
>> The other alternative would be empty the Ontology of all axioms, but I
>> can't see how to do this.
>>
>
> Would make sense to have a clear() method.

Yes. At the moment, I think, I have to remove all the axioms one by one;
this can't be a good way of dumping data structures. 

Phil