×
The internal search function is temporarily non-functional. The current search engine is no longer viable and we are researching alternatives.
As a stop gap measure, we are using Google's custom search engine service.
If you know of an easy to use, open source, search engine ... please contact support@midrange.com.
Le 25/07/2020 à 20:01, Patrik Schindler a écrit :
Hello Marc,
Am 25.07.2020 um 13:09 schrieb Marc Rauzier <marc.rauzier@xxxxxxxxx>:
Maybe to be sure that records remain sorted in arrival sequence in the member, for programs reading the file using some sort of RPG cycle?
Hm, I can't see an use case for that, nowadays. Most often, databases have an ID counter or something like that, automatically incremented. I forgot the special IBM term for it. But this is of course my perception.
Hello Patrick
Other people already gave you their opinion about this topic. In mine,
the important word in your questions is "nowadays". I do agree that, if
one has to start creating a brand new application, they will probably
use SQL based database design and all the modern ways to access the
data. And this is a highly valuable idea.
But we are talking here about IBM i which is supposed to be used for
business related activities. This is not a system for developers (all
Unix/Linux based systems do exist for that). This is not a system for
players (all Windows systems do exist for that). This is a system, just
like zOS, for production activities. The main purpose is to be efficient
and reliable. As Vernon says, there are a lot of existing applications
softwares which are still relying on techniques which were provided by
the operating systems a long time ago. And, I do agree with Bruce that
changing the default value, for example from RESUSEDLT(*NO) to
REUSEDLT(*YES) for CRTPF, of a parameter of a command (I mean with an OS
upgrade or PTF install) is not a good choice, even if it is described in
the Memo to users or cover letter. There is a risk for the applications
to be changed if they still need to use the old default but did not
forced it.
Other contibutors talked about performance reasons in regard to
REUSEDLT(*NO). I am quite sure that if you want to read an entire table
on the arrival sequence, using an old RPG cycle like program reading the
physical file will be much more efficient than reading on an index based
on SQL ROWID or similar column. If you do not mind, I can share an
experience about such a situation. I had to work on a job which was
reading an entire large file, on the sequence of keys set up in a
logical file (the logical file was used in the program). The job was too
long. To improve it, before reading the file, we added a step to sort it
according to the logical file keys (check out FMTDTA command) into
another temporary file, and we replaced the program which was reading
the logical file by keys with another one wich was reading the sorted
physical file by arrival sequence. Even with the extra time for sorting
the data, the overall duration was divided by 3 or 4. The OS was able to
block as much as possible the IO when reading the physical file in
arrival sequence, while blocking was not possible using the keys sequence.
Let's go back from the 90s to nowadays. It is possible that SQE is smart
enough to provide the same kind of performance improvements when reading
a table on a ROWID column. I do not really know.
Just my opinion and sorry to be long.
:wq! PoC
PGP-Key: DDD3 4ABF 6413 38DE - https://www.pocnet.net/poc-key.asc
midrange.com
