Is it just me, or is PX warning happy?

mfavero · Post by **mfavero** » Sat Apr 19, 2008 9:29 am

I get a bunch of these warnings from my Parallel job and I don't have a clue which stage they are coming from:

APT_CombinedOperatorController(3),2: APT_Decimal::ErrorBase: From: the source decimal has even precision, but non-zero in the leading nybble, or is too large for the destination decimal..Record dropped.

what the heck is a nybble? any relation to Nygel?

and what ever happened to those good old Server days?

eostic · Post by **eostic** » Sat Apr 19, 2008 9:55 am

Hi Michael...

Don't you just love EE's strict datatyping? It has its advantages...although sometimes you long for Server's "ah, just let it go for now" attutude.

Ernie

mfavero · Post by **mfavero** » Sat Apr 19, 2008 10:31 am

the problem here is that I haven't any idea which stage much less which column is causing this error. We're getting more than 50 and the job aborts. I'm afraid to let it run with maybe a million warnings.

ray.wurlod · Post by **ray.wurlod** » Sat Apr 19, 2008 4:37 pm

The reason you don't know which stage is your own fault; you are allowing the operators to combine (to execute in the same process). For diagnosis at least you do not want this to occur.

This particular message results from the way in which decimal data are stored internally, and the first half-byte ("nybble", or four bits) can be used as a "sanity check" on the precision value.

Set APT_DISABLE_COMBINATION to true and run your job (it will still generate the 50 warnings and abort) so you can figure out which stage is throwing the error. Then investigate the Precision column in the Columns grid to see which Decimal data type might be too small or too even. Then you can adjust this, and run again. Once it's cool, you can re-enable operator combination if you wish.