Predicate in Map or Reduce (RavenDb)? - ravendb

Question

If I want to apply a predicate to a document before I aggregate in a Reduce function, do I want to place that predicate in the Map function, or in the Reduce function?
So for example putting the predicate in the Map function would look like this:
Map = orders => orders
.Where(order => order.Status != OrderStatus.Cancelled)
.Select(order => new
{
Name = order.Firstname + ' ' + order.Lastname,
TotalSpent = order.Total,
NumberOfOrders = 1
});
Reduce = results => results
.GroupBy(result => result.Email)
.Select(customer => new
{
Name = customer.Select(c => c.Name).FirstOrDefault(),
TotalSpent = customer.Sum(c => c.TotalSpent),
NumberOfOrders = customer.Sum(c => c.NumberOfOrders)
});
And putting it in the Reduce function would look like this:
Map = orders => orders
.Select(order => new
{
Name = order.Firstname + ' ' + order.Lastname,
TotalSpent = order.Total,
NumberOfOrders = 1,
Status = order.Status
});
Reduce = results => results
.Where(order => order.Status != OrderStatus.Cancelled)
.GroupBy(result => result.Email)
.Select(customer => new
{
Name = customer.Select(c => c.Name).FirstOrDefault(),
TotalSpent = customer.Sum(c => c.TotalSpent),
NumberOfOrders = customer.Sum(c => c.NumberOfOrders),
Status = (OrderStatus)0
});
The latter obviously makes more sense, however it means that I have to add the Status property to the class of the Reduce result and then just set it to some unknown value in the Reduce as it doesn't actually mean anything there.

Answer 1

Only the first approach works for map/reduce. And no, the order will be ignored and you can't do something like FirstOrDefault in the result.
You need to think of map/reduce as two independent functions whereas the reduce function can be run multiple times on the same input, that's why the format of the input must match the format of the output. This can also happen on different server in parallel and asynchronously, thus new documents can be saved while the indexing is running.