On 2017-09-08 20:18, Andy Seaborne wrote:
> Yes, Chris - MultiUnion is linking with a distinguished that is updated.
> (Changing that would be a BIG change!)
> 
> Datasets like TIM (transactions in memory) and TDB can provide the union
> graph of their named graphs and more efficiently than the general purpose
> MultiUnion.

Previously I was using the "urn:x-arq:UnionGraph" named graph, but now I
need to be less generic and only include some of them. This is why I
have to work with the other options :(


>     Andy
> 
> On 8 September 2017 at 13:52, Chris Dollin <[email protected]>
> wrote:
> 
>> On 8 September 2017 at 10:11, George News <[email protected]> wrote:
>>
>>> Hi,
>>>
>>> Is multiunion only linking the graphs or copying them in a new graph?
>>>
>>> I'm thinking on splitting a huge graph on many simple ones and them
>>> making the union depending on the part of the main graph that has to be
>>> requested, therefore limiting the scope of the sparql and speeding
>>> things up. But if multiunion copies, then I don't think I will speed
>>> things.
>>>
>>
>> If things haven't changed too much ...
>>
>> (a) it links the graphs. They are not copied. Hooray!
>>
>> (b)ut when you run an operation over the graph that delivers
>> a set of triples (which is how queries work over union graphs)
>> it has to remember all the triples it has already seen so that it
>> doesn't deliver duplicates. Hence (depending on the triple
>> pattern being used) may mean lots of store being held onto.
>> Alas!
>>
>> Chris
>>
>> --
>> "What I don't understand is this ..."   Trevor Chaplin, /The Beiderbeck
>> Affair/
>>
>> Epimorphics Ltd, http://www.epimorphics.com
>> Registered address: Court Lodge, 105 High Street, Portishead, Bristol BS20
>> 6PT
>> Epimorphics Ltd. is a limited company registered in England (number
>> 7016688)
>>
> 

Reply via email to