on 10-09-2013 12:19 AM
Hi,
Background:
We setup a DSiM system and already loaded some sales data from the DSiM RDS into it. If we start SAP Predictive Analysis and open one of the SAP HANA Online Views (i.e. DS11) SAP PA shows the raw data and we can drag and drop the different columns in order to generate different visualizations.
Problem:
However if we go to the "Predict" tab, add i.e. a "HANA Triple Exponential Smoothing", configure the properties of the Algorithem and execute it we run into an error regarding the AFL wrapper generator.
The indexserver trace says:
AFLPM_SQLDriverObj.cpp(02936) : Registration of AFLLANG wrapper D failed at describeFunction with error: invalid parameter description!
cePopCustomLjit.cpp(00591) : Llang Runtime Error: Exception::SQLException423: AFL error: registration finished with errors, see indexserver trace
But nothing solved the problem.
Does anybody came across this error before?
Thanks,
Erik
Hi Eric,
what system do you use and how do your parameters look like?
Seems like there is a problem with the parameter description (did you check names an datatypes)?
Regards,
Sabrina
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Sabrina,
Thanks for your tip. We took a look at the PAL DEV guide pdf and created a new calculation view with only two rows: ID and RAWDATA. So we selected the output of the data source table.
First we created an ID column (basis was the date column of the source table). Then we renamed the data column to RAWDATA. Both columns have the data type INTEGER. When we tested this setup we received once again an error.
But this time it was a "SAP DBTech JDBC:[2048]: column store error: search table error: [2620] _SYS_AFL.AFLPAL:TRIPLESMOOTH: [135] (range 2) AFLFunctionFatal exception: PAL error[73001060]:Internal error"
We tested the new setup also with the Single and Double Exponential Smoothing. This actually worked pretty fine (no errors).
After we contacted some guys and product owner of the PAL and SAP Predictive Analysis one of them told us that currently INTEGER values aren't working and that we should switch to DOUBLE. And indeed this helped.
Conclusion:
Use correct data types. In our case we used INTEGER for the ID column and DOUBLE for the RAWDATA.
Thanks for your help,
Erik
Hi Andreas,
Due to the guide (http://help.sap.com/hana/SAP_HANA_Predictive_Analysis_Library_PAL_en.pdf), our knowledge and our experience you need a table/view with two columns.
One ID column of type INTEGER and one RAWDATA column of type DOUBLE. Cause DECIMAL is currently not supported by the PAL (contrary to the guide).
It may be that you do not need to name the columns exactly the way we did.
Does this help to solve your problem?
Best,
Erik
Erik,
thanks for the link. Unfortunately this does not solve my problem. I have to say that I'm using the Predictive Analysis Client which does not require an ID column. I've changed to DOUBLE but still the same problem.
Well, the real problem is that the error messages and the indexserver trace files are only of litte help. Just to state that there is "something wrong with the parameters" is not what I expect from an enterprise application...
However, thanks for your help.
Andreas
Hi Guys,
if you still have this issue you can have a look at the Note below:
1930665 - Labor Demand Planning - determination of planned execution times unsuccessful
https://service.sap.com/sap/support/notes/1930665
The same error is also resolved in this blog:
http://scn.sap.com/thread/3657025
And also resolved here: https://scn.sap.com/thread/3311619
Test this construction :
call SYSTEM.afl_wrapper_generator ('PAL_ANOMALY_DETECTION1', 'AFLPAL', 'ANOMALYDETECTION', PDATA)
and check PDATA(only distinct values in it)
Hope that helps
Cheers
Kingsley
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
8 | |
4 | |
3 | |
2 | |
2 | |
1 | |
1 | |
1 | |
1 | |
1 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.