Full agreement here.
No, there is currently no option available to limit resource usage on connection/user level.
However, even if it was, this wouldn't improve or fix bad model design.
The data type conversion in the example above is done twice (? to be really sure?) as CAST () plus as TO_DATE().
Also it happens outside the analytic view, again, why?
Not sure what happens *inside* the analytic view, but I got a feeling that
"_SYS_BIC"."fi-gl-lineitems/AN_BKPF_BSEG"."HKONT_SUBSTR
requires to perform a SUBSTR operation on each and every one of the 1.3B line items.
If that's the case, long runtimes are to be expected.
Finally, and I might be wrong on that, the GROUPING done here is likely to produce a group for rather few original records. This translates into a vast result set with lots of groups.
And the question here clearly is: what is the use case for delivering thousands of grouped records back?
- Lars