RE: Sequence error in MR

[Neil Palmer <neilp@xxxxxxxxxxx>]

Of course you're right.  The program I had was updating every record in a 
current year file with data from a previous years version of the same file 
- so pretty much a 100% match between primary & secondary files.

...Neil




G Armour <garmour400m@xxxxxxxxx> 
Sent by: midrange-l-bounces@xxxxxxxxxxxx
2003/09/18 11:55



To
Midrange Systems Technical Discussion <midrange-l@xxxxxxxxxxxx>
cc

Subject
RE: Sequence error in MR






Actually, I already raised the performance issue.  Based on the code that
was posted, unless the vast majority of the records in the secondary file
were to be deleted, this program wastes time unnecesarily reading records
it isn't going to delete.

As to your experience, the devil is in the details, I'm sure, Neil.

GA

--- Neil Palmer <neilp@xxxxxxxxxxx> wrote:
> There's one word missing from this discussion.  Performance. 
> Now I'm citing numbers here from 10 years ago at least, on ancient 
> hardware, but I rewrote a program someone had written to update a large 
> master file from a transaction file.  It was written using either CHAIN
> or 
> SETLL/READ (I can't recall) and I changed it to use MR logic.  The run 
> time dropped from approx 4 hours to 40 mins.  And before anyone asks, no
> I 
> don't care to recreate a similar test today, but if anyone want to go 
> ahead I'd be happy to hear the results.  :-)
> 
> ...Neil