I've recently discovered the moving aggregation node and it's absolutely brilliant!
But, would it be very difficult to add partitioning to the node? (are there any plans to do so?) That would be super helpful, especially with cumulative computation, like cumsum groupped by columns.
SUM(SALE) OVER (PARTITION BY DEPTNO) DEPTTOTAL
Group loop + Moving Aggregation is not efficient, and Java snippet I'm using now is fine, and fast, but would be quite ugly if I wanted to run multiple functions on multiple columns, especially partitioned by different subset of columns each.
Your Rank node it's great, but it only provides ranking, no cumsum, or cumavg.