Re: QMSHNDPM API

Simon,

I use a singleton procedure that contains the default program queue. If
you send a message without specifying the queue the default is used. In
an interactive job the default is set to the initial program. With
message subfiles I specify the long format and retrieve the default
queue from the singleton procedure. That way you have a queue that
stores messages and any procedure can display them, even procedures that
are not yet on the stack.

If the default has not been set it is *PRV. In a batch job, this is
usually fine. Like you, message API wrappers were about the first APIs I
supported. I started using this about 5 years ago and have found it is
easier. Overall, I have found this technique is more powerful (because
any procedure can send a message to any procedure) and easy to use.

One problem facing programmers that want to move to ILE is that it takes
a long time to build up a set of standard procedures. The APIs in the
iSeries-toolkit (at www.iseries-toolkit.org, along with some of the
other open-source sites) should give a lot of those programmers a head
start. Unlike a book or article, open source provides a way for
enhancments to occur as new techniques are identified or bugs are found.
With ILE and RPGIV, I think open source is one of the best ways to speed
adoption.

David Morris


>>> shc@flybynight.com.au 02/23/02 16:28 PM >>>

Hello David,

You wrote:
>To a certain degree you can use a fixed name. By default, I send all
>user messages for interactive jobs to a single program queue. In a
batch
>job I send them to *PRV. I also do this with a standard set of
>procedures I wrote a few years back. Those procedures hide many of the
>ugly details of PEPs etc. They do allow you to override this behavior.

I have a similar set of procedures to encapsulate the gory details of
message  handling -- I think they were among the first set of RPG
procedures I wrote.  Carsten reminded me of some additional options I
had
forgotten which nullify the effect of the generated RPG IV entry point
name.

How do you find the single program queue works in practice?  How do you
set
that up?  I mean does your application "know" the call stack and send
all
messages to the program queue of the main program?  Do you walk the call
stack looking for a specific program queue?  Do you send all messages to
the users message queue?  I'm curious because I would think that sending
to
*PRV regardless of where the job is running would be a simpler approach.

Regards,
Simon Coulter.