Thank you for your kind words, Gerrit! Until a few weeks ago, my use of grouping was a groping one also in my case - until I learned from the spec that the FLWOR mechanics are only fully understood if we regard the clauses as a pipeline of stations, each one transforming an input tuple stream into an output tuple stream - that's the key. And that's at first weird, as we are used to think in terms of "assignments", "loops", sorting, grouping, filtering, counting devices. But when thinking of the different clauses as just variants of a single idea - "tuple stream transformer" - everything is decomposed into tiny bits - the details of how a particular kind of clause maps its input tuples to output tuples. No interaction whatsoever between the clauses, apart from output tuple streams becoming input tuple streams. Suddenly it is not amazing any more that you can have any number of any kind of clauses, e.g. several where clauses, and that the clauses can occur at any position within the expression (except for the constraint that the first clause must be let or for). And so, for example, I had always kept my fingers crossed when adding an order by after a group by. Now I understand that once the group by has done its thing and produced an output tuple stream according to its rules, you have a tuple stream without any memory of what has happened at earlier clauses.