RE: Astonished by embedded sql

[Terrence Enger <tenger@xxxxxxxxxxxxxxxx>]

On Tue, 2010-10-12 at 07:41 -0500, Kevin Bucknum wrote:
> Since the original programmer is reading one record with an sql
> statement with no ordering and then exiting, then a chain should get the
> same record as he is getting anyways. Unless sql doesn't respect
> LIFO/FIFO/FCFO.

I believe that all the query interfaces (SQL, OPNQRYF, QUERY/400) 
carefully refrain from guaranteeing any sequence not specified 
explicitly.  It is just by luck--good luck, or bad luck, you can choose 
which you like--that we often get the sequence we were thinking about.

<aside attitude="rueful">
  I remember, long ago, that I keyed the very first record with a date 
  in 2000 into a file, and then a downstream report suddenly looked 
  really, grossly different.  It took entirely too long to verify that 
  the changed result had nothing to do with date; it was merely that
  one additional record made the query optimizer choose an entirely 
  different strategy.  For decades, the report had been delivering the 
  right result *by accident*.  (Thank you for letting me give this hobby
  horse some exercise.)
</aside>

HTH,
Terry.