Comments on: Testing RPGSQL for the number of records.

By: slateken

slateken — Mon, 20 Sep 2010 13:39:07 +0000

I've used Tom's approach and it works well. The following method also works for checking that a record exists without performing the "count".

D RecFound s 1a inz('N') 

/free 

reset RecFound; 

Exec SQL set :RecFound = (select 'Y' from sysibm/sysdummy1 
where exists (select zip1 from MyZips where zip1 = 15205));

By: tomliotta

tomliotta — Sat, 18 Sep 2010 01:36:51 +0000

I was wondering about indexing because I was hoping DB2 would be smart enough to check index-entry counts for an indexed column. IMO, that ought to make any physical I/O be a minimum. Just the page(s) with appropriate index tree elements should ever be hit.

Maybe a SET :var statement:

exec sql set :RcdCnt = ( select count(*) from MYFILE ) ;

Tom

By: murrayinfosys

murrayinfosys — Fri, 17 Sep 2010 22:52:36 +0000

Tom -
Slap-me-up-side-of-head! SMACK!

> SELECT COUNT(*)

Why didn’t I think of that.

Thanks

Phil

By: tomliotta

tomliotta — Thu, 16 Sep 2010 22:31:43 +0000

I need to update file A…

But not with SQL?

That is, a SQL UPDATE statement could restrict itself only to rows in A that had a match in B. But it seems that that isn’t exactly what you’re after. You want to “select” the number of matching rows in B, and, if not zero, then you want to move on to some later code that will eventually result in an update to A after some preparation is done over the values.

If that’s reasonably accurate, then a potential solution could be SELECT COUNT(*) FROM FILEB WHERE {fileb=filea}. In there, COUNT(*) would be selected into a host variable that could be tested for zero.

Of course, that’s different from SETLL since multiple possible I/Os could take place and you are trying to avoid that… assuming I’m understanding what you want.

Is the matching column indexed?

Tom