Re: Another Question about variable length fields

Vern Hamberg wrote:

I'll check the data - I thought I had some that was longer than
the allocation, but the fields were defined as only 10 bytes -
maybe it keeps things that short in the main space - that'd be an
interesting optimization.

There is a minimum amount of varlen storage required to effect creation of an auxiliary SID for a column. I do not recall the minimal length, but perhaps nearly eighteen bytes to account both the data length and the effective address within the aux piece. With an explicit versus defaulted ALLOCATE(0), that may force to the aux regardless. The /porting/ documents suggest character fields under forty bytes should be CHAR instead of VARCHAR with no allocation, to get more similar performance results to other SQL databases. Creating additional fields and\or creating a longer varlen\varchar field will likely be required to get the auxiliary segment created for the file after at least one row requires the overflow to the aux storage area.

IIRC someone made a change to ensure that after two RGZPFM with no data in the aux SID then it is either truncated or destroyed; after one reorganize, the auxiliary storage area may persist to maintain performance accommodations for update & insert activity.

Regards, Chuck