Autonumberhash256 limit?

[#1]

We have a QVW that contains near 300 million rows.  The keys were all concatenated strings but we were experiencing performance issues so I decided to autonumberhash256 each key.  Our build completed successfully but it appears as though QlikView stopped created keys somewhere around 66.5 million.  Is there an upper limit to the number of values autonumberhash256 will create?

[johnw]

I can't imagine why there would be any limit.  The autonumberhash256() function should be stateless.  Each invocation should be completely independent of all the others.  It's hard to predict the behavior of bugs, of course, but a limit here, even from a bug, seems unlikely.  I suspect there's some other explanation, even if I have none to offer.

[#3]

That is what I assumed... but during the load script the autonumberhash256() just stopped creating keys.  The weird part is that it stopped creating them for my large 100m+ row table, but subsequent statements where I am loading the dimensions, it created them...

Also, I thought that they were independent of each other but the output of autonumberhash256 seems to generate sequential numbers, up to the 66.5 million limit I reached.

[#4]

Ahh, you know one thing i thought of... These concatenated keys were already generated.. could it be that i hit a limit on the length of the field?  Maybe it would be best to go back one step in the ETL and do the autonumberhash256(field1,field2,field3,etc), rather than doing it on (synkey)?