How does the Alteryx engine handle multiple multirow calculations?

[andreeroos]

Hi.

I have a workflow that needs several multirow calculations to be performed (one per column).

How are these translated to the Alteryx engine? In the example below, are these calculations optimized or are table scans (or the Alteryx equivalent) performed several times (one for each node)?

I am thankful for any help on this issue.

[jdunkerley79 · Accepted solution]

(I don't have any access to the code so this my best guess based on how the SDK works.)

For simplicity, I am going to assume you have calculation depending on last 3 rows for each column.

Each calculation will be performed separately but not as a table scan. The calculations are more like stream processors.

Each multi-row formula tool will keep a small collection of the number of rows it needs.

The upstream tool pushes a record into the first calculation.

Assuming it already has 3 rows in its internal cache it will drop the oldest and store the new one.

It will then do the computation and push the new record into the next MR formula.

The second multi-row tool will act in the same way acting on the new row and then pushing downstream tools.

If you have a grouping in the multi-row this will cause a sort to be applied before the tools. Alteryx is very smart and I think if both have the same grouping it recognises this and hence only does the sort once.