Hello,
Apart from unicity, PK "not enforced" usage is disapointing regarding statistic lacks and I wonder what db2 optimizer can do with it really.
I would have expected a "virtual index" being inserted in SYSCAT.INDEXES, therefore a PK on (col1, col2, col3) would have firstkeycard, first2keycards, etc ... available to the optimizer.
But this is not implemented this way and we have to create a statistics profile with colgroups ourself.
This is really annoying for card estimation.
Best regards, JMB
------------------------------
Jean-Marc BLAISE
------------------------------
#Db2