与我的多对多关系，获取核心数据非常慢？

Question

I have a Core Data database with 3 entity types, Unit/Department/Document..

Unit/Department has a one-to many on the Department side, that is, a Unit can have multiple Departments, but each Department can only have one Unit.

Each Document can be a part of one or more Units and-or Departments, and i need to be able to find document matching a selected Unit or Department. The Unit id could be "1234" and the Department id for that Unit could be "123499". Notice that the Department id is always the 4 figures of the Unit, and then 2 more describing the Department.

So much for the data model, the problem is that when i select a Unit(and no Department), and need to list all the Documents under that Unit, the search should return all documents that belongs to that Unit AND all the documents that that belongs to ALL the departments under that Unit, and that search is taking over 4 seconds with 5000 Documents, I must do something wrong.

I am using a NSFetchedResultsController and here are my Predicate.

NSPredicate *predicate;
NSFetchRequest *req = [NSFetchRequest fetchRequestWithEntityName:@"Document"];   
[req setPropertiesToFetch:[NSArray arrayWithObjects:@"name",@"publisher",@"isFavourite", nil]];
 req.sortDescriptors = [NSArray arrayWithObject:[NSSortDescriptor sortDescriptorWithKey:@"name" ascending:YES]]; 
[req setFetchBatchSize:15];
predicate = [NSPredicate predicateWithFormat:@"ANY units.id == %@ || ANY departments.id BEGINSWITH[cd] %@",self.selectedUnit.id,self.selectedUnit.id];

Can I do this in another way ? I was wondering if I could select the Unit instead, and then go through Unit.documents instead, but then i think my FetchController is not needed, and i need to put the results in an Array, and sort them first(which fetches all Entitites, and takes even longer)..

To make matters worse, i also need to be able to search within the Name attribute of the documents later, and that CONTAINS[cd] makes the search even slower, I unfortunately cannot split the words into a seperate table, because I need to be able to mach words WITHIN words, that is, the user could type "result" and the the search should return all documents which has "result" somewhere in the title, an example could be "NSFetchedResultControllers are cool"

How can I make that CONTAINS search perform better, right now i have only 5000 docs, but it could also be 20000 or more.

Any good ideas, thoughts ? I have watched a lot of WWDC Core Data videos, read Core Data performance on the Apple developer site, I really think i have done it the right way, but maybe my model is all wrong, I can see in the SQL debug, that there is a lot of time used on joining tables, and I have tried to figure out how to make that SQL cleaner.

Thank you very much /Jacob

Answer 1

You are right in assuming that CONTAINS and BEGINSWITH are slowing the query down significantly. I have two ideas for you:

1. Use indexing (as I am sure you know from the WWDC videos). Especially for your text search with CONTAINS this should be a big performance boost.

2. With the specific predicate you mention, I would try using the number property of the id. In your scheme you have already conveniently ordered the ids by unit.

The predicate could then be something like this:

NSNumber *bottom = [NSNumber numberWithInt:[self.selectedUnit.id intValue]*100];
NSNumber *top = [NSNumber numberWithInt:([self.selectedUnit.id intValue]+1)*100];
[NSPredicate predicateWithFormat:
   @"ANY units.id == %@ || ANY (departments.id > %@ AND departments.id < %@)",
   self.selectedUnit.id, bottom, top];