Question

Let’s say I’ve got several lists of objects and need to union/intersect them. Currently I simply do something like this:

List result = lists[0];
for( int i = 1; i < lists.Count; i++ )
  result = (op == 'and') ? result.Union(lists[i]).ToList() 
    : result.Intersect(lists[i]).ToList(); 

I believe this is working very slow, even though lists are sorted by one of their fields. How would I make it faster? Hashsets, trees, etc.? Preferable provided by .NET 4.

These objects are actually cached DataRows, as relying on DB to make these operations seems to be much slower.

Answer 1

Well using a HashSet<T> would seem to make sense, yes – there’s no point converting everything to a list back and forth, assuming you really don’t care about the order:

var result = new HashSet<Foo>(lists[0]);
foreach (var list in lists.Skip(1))
{
    if (op == "and")
    {
        result.UnionWith(list);
    }
    else
    {
        result.IntersectWith(list);
    }
}

It’s somewhat ugly having the “if” there, mind. You might want:

var result = new HashSet<Foo>(lists[0]);
Action<IEnumerable<Foo>> action = op == "and"
     ? result.UnionWith : result.IntersectWith;
foreach (var list in lists.Skip(1))
{
    action(list);
}