Re: SQL efficiency vs. OPNQRYF

[Vernon Hamberg <vhamberg@xxxxxxxxxxx>]

OPNQRYF & SQL, in one way of thinking, are just different interfaces to the 
query engine of the database on iSeries. There are now (since v5r2, at 
least) 2 paths, and SQL can sometimes take the newer one, with potential 
benefit for complex joins of certain types. There are several things that 
determine whether the new engine is used - there is an info APAR for that, 
go to www.iseries.ibm.com/db2 and dig around to get more info.

There are ways to optimize either OPNQRYF or SQL - a valuable tool is to 
run STRDBG with no program and then run the statement you are testing. 
There will be several statements that say what the optimizer did. Some will 
say that an index was built. Pay attention to the recommended index 
columns, create such an index, and see whether the performance improves. 
The same index will help both OPNQRYF and SQL - it might even solve you old 
problems.

The join order can make a difference, as well, and there are some settings 
that can force certain things using SQL, just as there are with OPNQRYF.

Long answer, sorry. Short answer, it depends. With no new indexes, it is 
possible you will have the same problem. Maybe show us the commands and 
statements and let us take a shot at adjusting them.

Cheers
Vern

At 05:50 PM 9/13/2005, you wrote:

> Does anybody have any sort of educated opinion on relative efficiency of 
> SQL vs. OPNQRYF, when doing complex (and mostly joined) queries on 
> physical files?
>
> In the project I'm working on that involves taking a rather Rube Goldberg 
> contrivance of OPNQRYF calls, and re-implementing it in SQL, part of the 
> complexity of the old OPNQRYF implementation was that with a complex query 
> involving 2 levels of AND and OR across a bunch of files, trying to do it 
> in a single OPNQRYF would take far longer (order of FOREVER) than breaking 
> it down into a bunch of simpler OPNQRYFs. Am I likely to encounter the 
> same problem with SQL?
>
> --
> JHHL
> --