[iPhone SDK] Database/SQLite Best Practices

Question

Taking a look at the SQLiteBooks example, the idea of "hydrating" objects on-the-fly doesn't seem like a best practice to me. If you have a large number of books, for example, and they all need to be hydrated at the same time, I would want to lump the hydrating statements into one large query instead of each Book doing a separate query for it's own data.
I'm imagining these two scenarios:
1. A large query that loops through the rows of it's result set and creates Book objects from them, adding them to an array.
2. A query to get the primary keys, creating Book objects with only that data. When the book hydrates, it queries the rest of the data specifically for itself. (SQLiteBooks example)
I can see how the hydrating of method 2 would be easier to manage the memory stored in each Book object, but I'm really concerned with the speed it would take hundreds of Books to hydrate themselves.

Answer 1

If you know you're going to need many of the objects hydrated at the same time you may want to create some mechanism that allows you to do a single query and hydrate everything at once. In the case of SQLiteBooks, the hydrate method is only called when the user clicks through to the details view for a particular item, so hydrating objects en masse is not necessary.
I'd be very careful to avoid premature optimization here. We're working with very limited resources on these devices, and it is definitely important to optimize your code, but if you're dealing with a large number of items you could easily shoot yourself in the foot by hydrating a bunch of items at once and using up all your memory. If you dehydrate your objects when you receive a memory warning (as I believe the SQLiteBooks example does) you could end up thrashing - reading all your data from the DB, hydrating your objects, receiving a memory warning, dehydrating, and repeating. As the previous reply states, indexed lookups in SQLite are extremely fast. Even on the iPhone hardware you can probably do hundreds of lookups per second (I haven't run a test, but that'd be my guesstimate).