Question

I have documents in RavenDb that may look something like this:

{ "Id": "obj/1", "Version": 1 },
{ "Id": "obj/1", "Version": 2 },
{ "Id": "obj/1", "Version": 3 },
{ "Id": "obj/1", "Version": 4 },
{ "Id": "obj/2", "Version": 1 },
{ "Id": "obj/2", "Version": 2 },
{ "Id": "obj/2", "Version": 3 },
{ "Id": "obj/3", "Version": 1 },
{ "Id": "obj/3", "Version": 3 }

I’m trying to create an index that would give me:

• The sequences “obj/1” and “obj/2”, preferably grouped by Id.
• Not the sequence “obj/3”, since its not yet complete

How would I do this?

Answer 1

I managed to solve it. I’m not sure it is the optimal solution but it seems to work.

class SequenceIndex : AbstractIndexCreationTask<MyObject>
{
    public EventSequenceIndex()
    {
        Map = objects => from d in docs
                         orderby d.Version
                         select new
                         {
                             Id = d.Id,
                             Version = d.Version
                         };

        TransformResults = (database, results) => 
            from result in results
            group result by result.Id into g
            where g.Select(d => d.Version + 1)
               .Except(g.Select(d => d.Version))
               .Count() == 1
            select new
            {
                Id = g.Key,
                Objects = g
            };
    }
}

With query:

var events = session.Query<MyObject, SequenceIndex>()
    .As<MySequenceObjectView>()
    .ToArray();

I group the documents by Id and then take all version + 1 except all version, which for an original sequence of 1, 2, 3 will be 2, 3, 4 except 1, 2, 3, which gives me 4 (which is why I use the Count() == 1. But if there is a hole in the sequence, the count would be greater than 1, and therefore excluded from the results.