xserver trace. This is the error message text that I am receiving (PA 1.14 build 870) - see the beginning of the message text in the subject line of this post.
Hi Experts!
Each time that I run algorithms (I use one of the most straightforward ones, like HANA K-Means) on on object (table or analytical view) that contains variables of INTEGER type, each time I add more than 2 or 3 such variables as inputs in the algorithm node, I receive the above-mentioned error message:
An error occurred while executing the query.Error details: SAP DBTech JDBC: [423]: AFL error: [423] "SYSTEM"."AFL_WRAPPER_GENERATOR": line 32 col 1 (at pos 1198): [423] (range 3) AFL error exception: AFL error: registration finished with errors, see indexserver trace
If I change all of my variables to DOUBLE type, the error message goes away.
Does that mean that PA requires that INTEGER type is avoided? If so, what is the recommendation as to how joins should be built in HANA in order to create the objects (views) consumed by PA? Using DOUBLE for this purpose would be cumbersome if not a showstopper.
And above all, why INTEGER is supported in such a limited way (i.e. only a couple of variables with a rather low "collective cartesian dimensionality", i.e. with a small number of values per variable)?
Any hints or knowledge would be highly appreciated.
Thanks,
Sergey