Under the Hood - Core Data

Swift Bengaluru

Transcript

1. CORE DATA UNDER THE HOOD Swift Bengaluru - 14/04/2018 JATIN ARORA (@arorajatin)
2. WHAT WILL WE COVER TODAY? ▸ How does a Fetch Request work? ▸ How does Fetch Batch Size work? ▸ How does a Fetch Batch Request work? ▸ What happens when you ‘Fault’ an object? ▸ Extra Special - Relationship faults ▸ Performance characteristics of Core Data ▸ Tips for High Performance with Core Data
3. WHAT IS A FETCH REQUEST?
4. WHAT HAPPENS WHEN YOU EXECUTE A FETCH REQUEST?
5. FETCH BATCH SIZE
6. Context Coordinator Persistent Store SQLite execute(_ request:) with batch size executes query to fetch all objectIDs Loads all objectIDs into memory Special array backed by objectIDs fetchedObjects Array
7. Context Coordinator Persistent Store SQLite Executes query to fetch the objects Loads all objects into memory Object at index & surrounding objects retrieved objectAtIndexPath(_ indexPath:) Loads fetchBatchSize objects around the accessed index
8. WHAT IS A FETCH BATCH REQUEST?
9. CAVEATS OF USING BATCH REQUESTS ▸ The data in contexts and Persistent Store’s row-cache are out of date. ▸ Just refreshing won’t help because the objects would be turned into faults but the data in row-cache will still be older. ▸ Solution: ▸ Use mergeChanges(fromRemoteContextSave:into:) on the MOC with the [NSUpdatedObjectsKey: objectIDs]. ▸ Under the hood, this method updates the row-cache as well. :)
10. WHAT IS A FAULT?
11. WHAT HAPPENS WHEN YOU FAULT AN OBJECT?
12. RELATIONSHIP FAULTS ▸ To-one relationship act like normal faults. The relationships have objectIDs through which faults are fulfilled. ▸ To-many relationships work on two-level faulting. First when you access the relationship, the first level fault is fired and objectIDs for the relationship are loaded into the memory. ▸ Second-level faulting occurs like normal fault when any relationship object’s property is being accessed.
13. PERFORMANCE
15. PERFORMANCE TIERS
16. TIPS TO IMPROVE PERFORMANCE ▸ Use materialised objects method. ▸ Use indexing and compound indexing. ▸ It’s okay to do de-normalization in order to make your UI faster. ▸ Put simpler and more performant predicates first as it reduces the sample set. ▸ Use batch requests. It’s a little difficult to get right but once you get a hang of it. It just works. :)
17. DEBUGGING
18. DEBUGGING ▸ -com.apple.CoreData.SQLDebug 1 ▸ -com.apple.CoreData.ConcurrencyDebug 1 ▸ -com.apple.CoreData.MigrationDebug 1
19. REFERENCES ▸ https://www.objc.io/books/core-data/ ▸ https://github.com/arorajatin/CoreDataDemo
20. THANK YOU