Question

I am working on an app that imports a (very) large csv file into Core Data for catching purposes. The process goes something like this:

1. Parse the file line-by-line into an NSMutableArray full of NSStrings
2. After X number of lines, import the array into Core Data (this involves manipulating and in some cases creating new NSStrings)
3. [NSMutableArray removeAllObjects];
4. Rinse and repeat

At first glance it looks like the memory should be freed up at the conclusion of each cycle. With large files, however, I am finding that the app crashes after signaling a couple low memory warnings. Running the leaks tool tells me that most of the memory is being used up by CFString objects, which I understand are related to NSString objects (although I don’t know how)

I understand that NSString are reused whenever possible, and that they don’t act quite the same as other objects where memory is concerned, although I don’t understand any of the details. How can I reclaim the memory that my NSString objects are using?

Answer 1

Running the leaks tool tells me that most of the memory is being used up by CFString objects, which I understand are related to NSString objects (although I don’t know how)

NSString is actually a class cluster. Although you think you are working with NSStrings, you are almost certainly really working with one of its subclasses. The Cocoa framework chooses which subclass to use depending on circumstances.

CFString is actually not really an NSString at all, it is the pure C string object used by Core Foundation. However, you’ll find it is “toll free bridged” to NSString. This means that, to Cocoa, it looks like an NSString. The reason you are seeing lots of CFString usage is because whatever Cocoa API you are using to obtain these strings ultimately performs its work in Core Foundation.

Anyway, all that is irrelevant to your problem except for the fact, that lots of CFStrings more or less means the same as lots of NSStrings. What you need to reduce your memory footprint is nested autorelease pools as Girish has already said. As a first step, modify your algorithm like this:

1. Create a new autorelease pool.
2. Parse the file line-by-line into an NSMutableArray full of NSStrings
3. After X number of lines, import the array into Core Data (this involves manipulating and in some cases creating new NSStrings)
4. [NSMutableArray removeAllObjects];
5. drain the autorelease pool
6. Rinse and repeat (start at 0)

If this doesn’t help, or only helps a bit, consider bracketing just the parsing with an autorelease pool.