Hi Thanks for taking the time to answer.
Indeed I assume it cannot use the index to do the where close, yet it
should be faster because it should only have to fetch one article on
compare the @rid (limit 1), the first article in the index has a different
@rid
Is it possible to create an index on @rid even if it not a property but a
metadata? so something like Article.@rid?
I think to use composite index, you need to query on both fields in order,
so in this case i would have to query on publishDate and @rid, yet I only
query on @rid and only use publishDate for the sorting.
Le vendredi 17 avril 2015 23:19:18 UTC-4, nagaraja sosale ramaswamy a
écrit :
>
> i could be wrong, but i suspect it is because when you are querying the
> index, the where clause is not filtering on the indexed field publishDate
> itself instead on @rid which is not part of the index.
> maybe you could try composite index on publishDate and @rid?
>
>
> On Saturday, April 18, 2015 at 12:52:41 AM UTC+5:30, Herve Eichwald wrote:
>>
>> In fact the problem can be reproduced with a simpler case:
>>
>> I have an article cluster with 400 000 records
>>
>> select from Article where @rid not in [#13:250000] limit 1 takes 30 ms
>>
>> while select from indexvaluesasc:Article.publishDate where @rid not in
>> [#13:250000] limit 1 takes 400ms
>>
>>
>> Why? It should just have to examine the first article and see that @rid
>> are different, so why is it so slow?
>>
>>
>> Thanks
>>
>>
>> Le vendredi 17 avril 2015 11:57:34 UTC-4, Herve Eichwald a écrit :
>>>
>>> Hello guys,
>>>
>>> I would like to speed up the following use case, not sure how to do it:
>>> I have a Vertex of Users and a Vertex of Articles
>>> Users visit articles.
>>> I want to get the list of non yet visited articles for a given user.
>>> So I do:
>>> select @rid from indexvaluesdesc:Article.publishDate where @rid =
>>> (select out('Visit') from User where guid =
>>> 'f60c1880-4c45-5efc-88cb-a54ed9a030bb') limit 1
>>> which is rather slow (500ms) ti give me #13:0
>>>
>>> select out('Visit') from User where guid =
>>> 'f60c1880-4c45-5efc-88cb-a54ed9a030bb' gives me 15 @rid and is fast
>>>
>>>
>>> Note that the article has been inserted in the cluster in the
>>> publishDate desc order, so #13:0 is the latest.
>>> When I do:
>>> select @rid from Article where @rid not in (select out('Visit') from
>>> User where guid = 'f60c1880-4c45-5efc-88cb-a54ed9a030bb') limit 1
>>> it gives me the response #13:0 in 17ms
>>>
>>> Yet I need the index as my articles won't always be inserted in this
>>> order and the cluster can grow pretty big, (right now around 500 000
>>> records) and most articles won't have been visited for a user
>>>
>>>
>>> Any idea how I could do that?
>>>
>>>
>>> Thanks
>>>
>>
--
---
You received this message because you are subscribed to the Google Groups
"OrientDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
