Re: Question about opening and closing files

["Charles Wilt" <charles.wilt@xxxxxxxxx>]

Crispin,

Sorry typo on my part,  that should have said "But a trigger program
should never be ended with *LR = *ON anyway."

In one word, performance.

Reactivating the program, reintializing all the variables, and
reopening the files will kill the performance.

The REDBOOK - Stored Procedures, Triggers and User-Defined Functions
on DB2 UDB for iSeries says:
Performance Considerations (p435)
– Try to exit a trigger program the "soft" way. Avoid, if you can,
SETON LR in RPG, STOP
RUN in COBOL, and exit() in C. This allows you to leave some files
open and avoid the
overhead of opening them again when you get back into the trigger.
This technique is
broadly used in our trigger examples. Use a static variable to
determine whether the
file needs to be opened. In the C language, define the file pointer as
static and check
for the NULL value. In terms of application logic, if you open a file
to append record at
the end or for reading with random positioning, you can avoid closing it.

Since I happened to be doing some benchmarking anyway, I ran a quick quick.
An a 515 running v5r4, with no other users, and having a program
submitted to batch that updates every row of a 2.9 million record
table. (Table is journaled and update program commits every 1000
records, for why look for my next benchmark post :)

                             CPU         Clock
no trigger                   41         1:58
LR=*ON no files        260         7:30
LR=*ON 1 file            519       16:25
LR=*OFF 1 file          215         6:11

Charles


On Tue, Sep 9, 2008 at 11:25 AM, Crispin Bates <cbates@xxxxxxxxxxx> wrote:
> Charles,
>
> Why should trigger programs end with *INLR=*ON?
>
> Thanks,
>
> Crispin.
>
> ----- Original Message -----
> From: "Charles Wilt" <charles.wilt@xxxxxxxxx>
> To: "RPG programming on the AS400 / iSeries" <rpg400-l@xxxxxxxxxxxx>
> Sent: Tuesday, September 09, 2008 10:59 AM
> Subject: Re: Question about opening and closing files
>
>
> > No of course not.
> >
> > But a trigger program should never be ended with *LR = *OFF anyway.
> >
> > So alarm() might work.
> >
> > Charles
> >
> > On Tue, Sep 9, 2008 at 10:03 AM, Lim Hock-Chai
> > <Lim.Hock-Chai@xxxxxxxxxxxxxxx> wrote:
> > > Are you saying that the alarm() api can actually wake up an in-active
> > > program (ended with *inlr = *off)?  That would be cool if this is true
> > > :).
> > >
> > >
> > >
> > >
> > > "Charles Wilt" <charles.wilt@xxxxxxxxx> wrote in message
> > > news:<mailman.9454.1220926197.2545.rpg400-l@xxxxxxxxxxxx>...
> > > > Pretty big I'd guess....but it should be easy enough to test for
> > > yourself..
> > > > (Maybe) a better idea:
> > > > Come up with a way to have the trigger program wake up after a few
> > > > minutes and close its files.  This way, if your changing 100s of
> > > > records at a time, you'll only have to pay for one open/close.
> > > >
> > > > Question is how to do the wake up?
> > > > First thought is to use alarm() API along with the appropriate signal
> > > > handling APIs.
> > > >
> > > > BUT I don't know that trigger's play well with signals...don't know
> > > > they don't either. <grin>
> > > >
> > > > Another possibly easier option:
> > > > Look at why is the file open an issue in the first place.  Can you
> > > > make it not an issue?  For instance, once place I've had trouble is
> > > > with a process that uses CLRPFM to clear a work file when that process
> > > > is called by an ODBC/JDBC client.  Instead of a CLRPFM, I sometimes
> > > > will use an SQL DELETE.  Or I'll change the program reading the file
> > > > to delete the record after it reads it.  Consider adding
> > > > REUSEDLT(*YES) to the file.
> > > >
> > > > HTH,
> > > > Charles
> > > >
> > > >
> > > >
> > > >
> > > >
> > > > On Mon, Sep 8, 2008 at 8:47 PM, James Lampert
> > > <jamesl@xxxxxxxxxxxxxxxxx> wrote:
> > > > > How much overhead would you say is involved in making files *USROPN,
> > > and
> > > > > keeping them open only while actually performing I/O on them, as
> > > opposed
> > > > > to doing things the usual way?
> > > > >
> > > > > We're thinking of doing this in a trigger program, in order to cut
> > > back
> > > > > on the number of file locks.
> > > > >
> > > > > --
> > > > > James H. H. Lampert
> > > > > Touchtone Corporation
> > > > >
> > > > > --
> > > > > This is the RPG programming on the AS400 / iSeries (RPG400-L)
> > > mailing list
> > > > > To post a message email: RPG400-L@xxxxxxxxxxxx
> > > > > To subscribe, unsubscribe, or change list options,
> > > > > visit: http://lists.midrange.com/mailman/listinfo/rpg400-l
> > > > > or email: RPG400-L-request@xxxxxxxxxxxx
> > > > > Before posting, please take a moment to review the archives
> > > > > at http://archive.midrange.com/rpg400-l.
> > > --
> > > This is the RPG programming on the AS400 / iSeries (RPG400-L) mailing
> > > list
> > > To post a message email: RPG400-L@xxxxxxxxxxxx
> > > To subscribe, unsubscribe, or change list options,
> > > visit: http://lists.midrange.com/mailman/listinfo/rpg400-l
> > > or email: RPG400-L-request@xxxxxxxxxxxx
> > > Before posting, please take a moment to review the archives
> > > at http://archive.midrange.com/rpg400-l.
> > --
> > This is the RPG programming on the AS400 / iSeries (RPG400-L) mailing list
> > To post a message email: RPG400-L@xxxxxxxxxxxx
> > To subscribe, unsubscribe, or change list options,
> > visit: http://lists.midrange.com/mailman/listinfo/rpg400-l
> > or email: RPG400-L-request@xxxxxxxxxxxx
> > Before posting, please take a moment to review the archives
> > at http://archive.midrange.com/rpg400-l.
>
>
> --
> This is the RPG programming on the AS400 / iSeries (RPG400-L) mailing list
> To post a message email: RPG400-L@xxxxxxxxxxxx
> To subscribe, unsubscribe, or change list options,
> visit: http://lists.midrange.com/mailman/listinfo/rpg400-l
> or email: RPG400-L-request@xxxxxxxxxxxx
> Before posting, please take a moment to review the archives
> at http://archive.midrange.com/rpg400-l.