upload many files and save in CoreData - ios

Question

Using the API I get 500 pictures, upload them asynchronously. Then I want to keep all these pictures in CoreData, but the application crashes due to insufficient memory.
When upload finished i call method createFromBlock
+(id)createFromBlock:(MRBlock *)block{
ManagedBlock *item = [ManagedBlock MR_createInContext:DefaultContext];
item.id = #(block.id);
item.name = block.name;
item.slidesInBlock = #(block.slidesInBlock);
item.sizeBlock = block.sizeBlock;
item.desc = block.desc;
item.imagePath = block.imagePath;
item.image = [MRUtils transformedValue:block.image];
item.price = block.price;
int i = 0;
ManagedItem *new = nil;
for (MRItem *lol in block.items){
NSLog(#"%i", i);
new = [ManagedItem createFromItem:lol];
new.block = item;
[item addItemsObject:new];
new = nil;
i++;
}
[DefaultContext MR_saveWithOptions:MRSaveSynchronously completion:nil];
return item;
}
In foreach block.items app is crashed. approximately after 150-160 positions.
If i comment new = [ManagedItem createFromItem:lol]; - app dont crash
+(id)createFromItem:(MRItem *)object{
ManagedItem *item = [ManagedItem MR_createInContext:DefaultContext];
item.id = #(object.id);
item.title = object.title;
item.detail = object.detail;
item.imagePath = object.imagePath;
item.image = [MRUtils transformedValue:object.image];
return item;
}

Answer 1

First, you should not load all your data and then save it. You should load in small batches, and save each batch.
However, for your specific example, I believe you can get away with turning each object into a fault after saving.
I have never used magical record, and have no idea if it will allow you do to contexty things without calling its methods. I looked at it once, but it hides way too many details for me... Unless you are doing the most basic things, Core Data can be quite complex, and I want to know everything that is going on with my core data code.
+(id)createFromBlock:(MRBlock *)block{
#autoreleasepool {
ManagedBlock *item = [ManagedBlock MR_createInContext:DefaultContext];
item.id = #(block.id);
item.name = block.name;
item.slidesInBlock = #(block.slidesInBlock);
item.sizeBlock = block.sizeBlock;
item.desc = block.desc;
item.imagePath = block.imagePath;
item.image = [MRUtils transformedValue:block.image];
item.price = block.price;
NSUInteger const batchSize = 10;
NSUInteger count = block.items.count;
NSUInteger index = 0;
while (index < count) {
#autoreleasepool {
NSMutableArray *newObjects = [NSMutableArray array];
for (NSUInteger batchIndex = 0;
index < count && batchIndex < batchSize;
++index, ++batchIndex) {
MRItem *lol = [block.items objectAtIndex:index];
ManagedItem *new = [ManagedItem createFromItem:lol];
new.block = item;
[item addItemsObject:new];
[newObjects addObject:new];
}
[DefaultContext MR_saveWithOptions:MRSaveSynchronously completion:nil];
for (NSManagedObject *object in newObjects) {
// Don't know if MagicalRecord will like this or not...
[DefaultContext refreshObject:object mergeChanges:NO];
}
}
}
return item;
}
Basically, that code processes 10 objects at a time. When that batch is done, the context is saved, and then turns those 10 objects into faults, releasing the memory they are holding. The auto-release-pool makes sure that any object created in the containing scope are released when the scope exits. If no other object holds a retain on the objects, then they are deallocated.
This is a key concept when dealing with large numbers or just large objects in Core Data. Also, understanding the #autoreleasepool is extremely important, though people who have never used MRR do not appreciate its benefit.
BTW, I just typed that into the editor, so don't expect it to compile and run with a simple copy/paste.