You use it the same way you would use NSMutableDictionary. The difference is that when NSCache detects excessive memory pressure (i.e. it’s caching too many values) it will release some of those values to make room.
If you can recreate those values at runtime (by downloading from the Internet, by doing calculations, whatever) then NSCache may suit your needs. If the data cannot be recreated (e.g. it’s user input, it is time-sensitive, etc.) then you should not store it in an NSCache because it will be destroyed there.
Example, not taking thread safety into account:
// Your cache should have a lifetime beyond the method or handful of methods
// that use it. For example, you could make it a field of your application
// delegate, or of your view controller, or something like that. Up to you.
NSCache *myCache = ...;
NSAssert(myCache != nil, @"cache object is missing");
// Try to get the existing object out of the cache, if it's there.
Widget *myWidget = [myCache objectForKey: @"Important Widget"];
if (!myWidget) {
// It's not in the cache yet, or has been removed. We have to
// create it. Presumably, creation is an expensive operation,
// which is why we cache the results. If creation is cheap, we
// probably don't need to bother caching it. That's a design
// decision you'll have to make yourself.
myWidget = [[[Widget alloc] initExpensively] autorelease];
// Put it in the cache. It will stay there as long as the OS
// has room for it. It may be removed at any time, however,
// at which point we'll have to create it again on next use.
[myCache setObject: myWidget forKey: @"Important Widget"];
}
// myWidget should exist now either way. Use it here.
if (myWidget) {
[myWidget runOrWhatever];
}