Re: RPG Blocked Writes

Kurt

That confirms what I saw in the Database Programming manual and presented in a response to Dennis (IIRC) when he mentioned a unique key causing non-blocking. The manual said that if you omit NBRRCDS when using SEQONLY(*YES), that it will not be blocking if there are ANY keyed LFs over the PF. Using NBRRCDS lets the system use blocking. in this case.

Cool!
Vern

On 6/8/2010 5:01 PM, Kurt Anderson wrote:

Thanks, Chuck. What you quoted I hadn't recalled reading, and that's because I was looking at an IBM technical document about OVRDBF and not actually the help. I've made a change and the result is success.

Earlier this week (oh, it's Monday, I guess last week) I was testing OVRDBF SEQONLY and couldn't see any impact (was doing time trials). Well, now that I was looking at the job's I/O (which was telling me my writes weren't blocked), and since this particular program is a simple in and out (from one file to another), both I/O's should have been the same when both blocked - so I changed it to use SEQONLY with a larger # of records (110,000 / rcdlen) and I could see the difference. Wonderful. (110,000 b/c the IBM document suggested that # when the file contains Varying length fields.)

So I consider myself edjumacated.

Thanks,
Kurt

-----Original Message-----
From: rpg400-l-bounces@xxxxxxxxxxxx [mailto:rpg400-l-bounces@xxxxxxxxxxxx] On Behalf Of CRPence
Sent: Tuesday, June 08, 2010 4:13 PM
To: rpg400-l@xxxxxxxxxxxx
Subject: Re: RPG Blocked Writes

The existence of "at least one dependent logical file with a
keyed sequence access path that does not share the access path of
the keyed physical file member" is a reason why blocking will not be
used. So... Why not just OVRDBF like the documentation says?
Unless there is a UNIQUE key, the effect of single row processing
for having non-unique keyed LFs effect can be overridden, just as
the quoted documentation from several messages back. So although
the attempted default normally would be for a 4K buffer for record
blocking, the existence of the keyed LF causes that default to
revert to single record I\O, and then an override can be used to set
the preferred or otherwise more /appropriate/ block size.

Regards, Chuck

On 08-Jun-2010 14:31, Charles Wilt wrote:

If the logical has keys, particularly unique keys or is set
MAINT(*IMMED), then I would expect your results.

It's not the logical per say, it's the existence of a
MAINT(*IMMED) access path; note MAINT(*IMMED) is required for
unique keys, either in the physical itself or the logical.

On Tue, Jun 8, 2010 at 3:09 PM, Kurt Anderson wrote:

I did some testing, and as it stands, my results show that
when the PF (keyed or not) has a LF, output writes to the PF
are not blocked. If the PF (keyed or not) does not have a
logical, the output writes are blocked.

<<SNIP>>

On Tue, Jun 8, 2010 at 9:44 AM, Kurt Anderson wrote:

<<SNIP>>

And the sentence after the quote I provided is: "The
programmer is expected to override the file or use larger
blocks if the default is not appropriate." But the default
_is_ appropriate. It's an output-only specification, and
thus should block (barring other reasons for the file not
blocking which isn't mentioned in the article).

<<SNIP>>