Core Data: Subquery Performance Problem - iphone

Question

I've been trying to see if there is any way I can improve on the performance of the following Predicate:
[NSPredicate predicateWithFormat:#"A=%# and SUBQUERY($self.words,$k,$k.keyword IN %#).#count==%d",
<value for A>,
keywordList,
[keywordList count]];
What I'm trying to do is return all the records of A that have keywords that are ALL contained in the supplied array (keywordList). I have a small database of about 2000 records. However, the keywords for each Entity ranges from 40-300 keywords. The keywords are represented as a to-Many relationship from A to an entity called Keywords. The above query works but takes about 7 seconds to run on my iPhone4. I want to see what I can do to shrink that down to a sub-second response.
Thanks,
Michael

Answer 1

This doesn't quite solve the problem as the results that come back from intersection of the fetchedObjs is not correct. It doesn't guarantee that all the 'A' objects contain all the supplied keywords. In fact, what comes back is an empty list and the entire process actually takes longer.
Something must be wrong with my model or the previous answer. I'm not getting the results I'm expecting. One alternative to returning a list of 'A' would be to return a list of ManagedObjectIDs of 'A'. [mo valueForKey: (NSString *)] returns the object of the relationship.
Note: I originally posted this question anonymously thinking I was logged in, so i can't seem to comment on anybody's answer.

Answer 2

I think you are approaching the problem backwards. If you have the keywords, you should search for the Keyword objects and then walk their A relationships to find the related A objects. Then check for overlap in the relationship sets.
So something like:
NSFetchRequest *fetch=//...set up fetch
[fetch setEntity:Keyword_entity];
NSPredicate *p=[NSPredicate predicateWithFormat:#"keyword IN %#",keywords];
[fetch setPredicate:p];
NSArray *fetchedObj=//... execute fetch
NSMutableSet *inCommon=[NSMutableSet setWithCapacity:[fetchedObjs count]];
for (NSManagedObject *mo in fetchedObjs){
if ([inCommon count]==0){
[inCommon addObjects:[mo valueForKey:#"aRelationshipName"]];
}else{
[inCommon intersectSet:[mo valueForKey:#"aRelationshipName"]];
}
}
... inCommon would then contain a set of all unique A objects that shared all the keywords.
Update:
From OP author:
This doesn't quite solve the problem
as the results that come back from
intersection of the fetchedObjs is not
correct. It doesn't guarantee that all
the 'A' objects contain all the
supplied keywords.
Okay, lets take another tack. Assuming you have a data model that looks something like this:
A {
keywords<<-->>Keyword.as
}
Keyword{
keyword:string
as<<-->>A.keywords
}
Then something like this should work:
NSFetchRequest *keyWordFetch=//...set up fetch
[keyWordFetch setEntity:Keyword_entity];
NSPredicate *p=[NSPredicate predicateWithFormat:#"keyword IN %#",keywords];
[keyWordFetch setPredicate:p];
NSArray *fetchedKeywords=//... execute keyWordFetch
NSFetchRequest *entityAFetch=//...set up fetch
[entityAFetch setEntity:A_entity];
NSPredicate *pred=[NSPredicate predicateWithFormat:#"ALL keywords IN %#", fetchedKeywords);
[entityAFetch setPredicate:pred];
NSArray *fetchedAs=//... execute fetch