Re: Embedded SQL in a Service Program versus in a Program

I can confirm Chuck's statement about speed with shared-weight - IF you have a logical created that way.

So in one implementation, I had a pair of logicals - one with default sort sequence, the other with the shared-weight sequence.

One thing about shared weight - hope I'm right on this - the order looks as if it is ASCII, not EBCDIC - Chuck, can you confirm? IIRC, that's what I saw - for example, numbers came first, which is like ASCII, not last, as in EBCDIC.

Vern

On 7/23/2014 8:03 AM, CRPence wrote:

On 23-Jul-2014 07:51 -0500, Eileen@xxxxxxxxxx wrote:

Thank you all for your responses. I found the issue. In order to
show you all an example of the program/service program I decided to
create a stripped down version of them. In doing so I realized I was
missing one thing in the program that was in the service program. For
these types of procedures we add an SQL option so that sorting will
be case insensitive. When I add this to the program I end up with the
same results as the service program. And if I remove it from the
service program I get the faster results.

C/EXEC SQL
C+ SET OPTION SRTSEQ=*LANGIDSHR
C/END-EXEC

FWiW, comparing the spooled data produced by Print SQL Information (PRTSQLINF), specifically the record with 'SRTSEQ(', would have revealed the difference between the SQL attributes for the two objects.

The SQL using the Sort Sequence of the Language Identifier Shared-weight collation, for implementations depending [heavily] on Keyed Access Paths can operate just as quickly as with another sort sequence, *if* the INDEX definitions were created with that matching Language Identifier (LANGID) and Sort Sequence (SRTSEQ) settings.

Note: For proper operation, some query requests will required a shared-weight, so be sure that the removal of that option does not _break_ the query; i.e. ensure the query does not depend on shared-weight comparisons\ordering.