I am mapping a one-to-many relationship between two tables. I have seen many examples of this with the collection in the parent object mapped using set, map, bag or list.

Can anyone describe the differences in using these?

A set is a collection in which no item occurs more than once. This is the most common persistent collection type, in my experience.

A bag is a colleciton in which items may occur more than once: they are very inefficient because hibernate cannot tell if items you put into it are the same as ones already in it (assuming that they are equal), so it has to delete the entire collection and re-save it from memory.

A list is an indexed bag. The index lets hibernate know whether or not a particular in-memory object is the same one as an equal on-DB object, so complete delete/re-insert is not needed.

A map is just like a list, except the index doesn't have to be a computable (usually sequential) integer, it can be anything, even another object.

Perfect. Thanks

is this answer also applicable to nHibernate?
I'm using bags more then sets and I'm not seeing complete deletion and resaves occuring.

It does apply to NHibernate too, yes. Bags don't need complete deletion when you're modifying existing members, but when you add or remove an item from the bag, the entire collection should be deleted/reinserted. If you're not seeing this behaviour, you should write a test to verify that you're seeing the expected results after these operations. If you are, I wouldn't worry about it: perhaps there's special cases when hibernate can avoid complete collection deleteion operations. I know that they're not needed for idbags, perhaps there are similar situations where they're not needed.

thnx

From your post it seems that bag does not have any advantage. But it surely has, as otherwise it would not be there. Can you tell us?

Thanks,
Viliam

I'm happy to learn, but as far as I can see you have it exactly the wrong way around.

A set is very inefficient if/when adding to a large collection, because NHibernate has to load the entire collection into memory in order to ensure that uniqueness is preserved.

A bag suffers no such thing. Why would it care? - it's allowed to have duplicates.

Set is inefficient with large collections, but bag is as well: it has to delete whole large collection and re-insert it to add/remove one single row. It seems that list is best at performance while set enforces uniqueness at the cost of performance, if you need it.

And again: where is bag more appropriate than these two?

Viliam

Hey Viliam,

What makes you think that? I've used a lot of bags in my time, and none of them have done that.

Unless you're thinking of the 'all-delete-orphan' mapping?