RE: QWCCVTDT Weirdness, solved!

>>I'd really like to understand the rationale behind writing the code
that 
>>way to begin with.  Was it just random trial and error until something

>>worked?  Why would you ever do that?  What was the thought process
behind 
>>it?

You're of course correct in your statement that some of this code has
been stumbled upon - experimenting with options until it worked, often
not understanding the underlying details. I'm the first person at this
location to even think in terms of modules, user spaces, prototypes,
etc. - so unfortunately I've had little precedent to follow (and as we
all know, programming is partially a plagiaristic art). But that's why
when I do reach a point of things not adding up, I ask for feedback -
and I'm always looking for a better way. It's not important that my way
is better necessarily...

>>You could also use QUSCRTUS and QUSPTRUS to create and get a pointer
to a 
>>user space instead of allocating RAM with the %alloc() BIF.  It'd
require 
>>more code, but you'd end up with the same results (more or less).

Since I already have the modules written to create a user space and
return the pointer to it, it's not a lot of work.

I understand how you use alloc to get more space assigned to the
variable than the d-spec definition would normally allow. Since I prefer
the user space method, how would I set that up?

Let's say my user space pointer is USP_Ptr.
I'll need to define a variable based on that.
So,

D MyVar         ??????A  Based(USP_Ptr). 

How big do I make the variable. I kept thinking that if I make it the
largest I could define on the D-spec (32767) that's all the API would
return for me. Or - is the 'length of receiver variable' (staying with
the QMHRDQM API example) the key? Will the API use however much space I
tell it to in that variable, regardless of the definition of MyVar.
Example.

D MyVar             1A   Based(USP_Ptr).
D MyVarLen         10I 0 Inz(1048544)                      *-or whatever
size I want-*

Then when the API is called, it will start putting data into MyVar but
continue as far as the MyVarLen lets it go? If this is the case, I can
run with that. I think that's a little obscure myself, but I'll agree
that it's a little more clear than my weird pointer structure I'd forced
to work before. (The reason I went with odd pointer definitions is that
I couldn't resolve in my mind how long to make MyVar).

Thanks for the help. Always looking for a better way.

JPW









 

 

The bitterness of poor quality remains long after low pricing is
forgotten! 

 

Cautillo, Leon M.



-----Original Message-----
From: rpg400-l-bounces@xxxxxxxxxxxx
[mailto:rpg400-l-bounces@xxxxxxxxxxxx] On Behalf Of Scott Klement
Sent: Tuesday, 28 February 2006 20:08
To: RPG programming on the AS400 / iSeries
Subject: Re: QWCCVTDT Weirdness, solved!



> I am still interested on any thoughts about how the pointer based on a
> pointer thing works (because it does), and if there is a better way.

Oh, now that I've looked it over again and read Tom's reply, I can tell 
you exactly how it works.

Basically what you were doing is equivalent to doing this:

      D                 ds
      D   InVar                 1     16*
      D   DTSField              1     17A


I don't know what the name of the field containing the DTS timestamp was

(since you didn't post it) so I made one up.  But hopefully you
understand 
what I'm saying.  They occupied the same spot in memory.

Granted, you used pointer logic rather than a data structure, but it had

the exact same effect.

It didn't matter which one you passed to the API, since they contained
the 
same value.  As long as you made sure DTSField was set to a valid DTS 
timestamp (and therefore made InVar an illegal pointer value) the code 
would work.

It makes no sense to do that -- but that's what you did.

> For example, if you wanted to dump all the messages in a data queue
into 
> a user space with the QMHRDQM API, and the API expects a character 
> argument for the output variable, how would you do it?

Well, the method you're using wouldn't help with that.  But you could do

it this way if you wanted to:

      D mySpace         s               *
      D InVarPtr        s               *
      D InVar           s           1024A   based(InVarPtr)
      D x               s             10I 0

       ** Ask the OS for 1mb of memory:
      C                   eval      mySpace = %alloc(1024 *
%Size(InVar))

       ** set the entire space to blanks:
      c                   for       x = 0 to 1023
      c                   eval      InVarPtr = mySpace + (x *
%size(InVar))
      C                   eval      InVar = *blanks
      c                   endfor
       .
       .
       ** when you're done, let the OS have it's memory back
      C                   dealloc                 mySpace


Basically, InVar views 1k of memory at a time.  So in the above example,

I've reserved 1 mb of memory to the MySpace pointer.  And then I've 
changed InVarPtr to tell it to view different sections of that 1mb of 
memory.  (In this example, I used that functionality to blank it out.)

To pass the contents of the 1mb space to the API you'd do this:

       ** Ask the OS for 1mb of memory:
      C                   eval      InVarPtr = mySpace
      C                   CALL      'SOMEAPI'
      C                   PARM                    inVar

Since the current address of InVar is the same as the address stored in 
mySpace, it'd be the memory that's allocated to mySpace that'd be used
by 
the API.

> I needed a lot more than 32767 characters for a large data queue.

In this case, I've used the %alloc() BIF, which is capable of allocating

up to (approx) 16mb. That's a lot more than 32k!

You could also use QUSCRTUS and QUSPTRUS to create and get a pointer to
a 
user space instead of allocating RAM with the %alloc() BIF.  It'd
require 
more code, but you'd end up with the same results (more or less).

I don't know if the data queue APIs support it, but if 16mb wasn't
enough, 
you COULD allocate the memory in teraspace.  If you did that, you'd be 
able to allocate as much as 1 terabyte to MySpace.  So...  that should
be 
enough, eh?


> By basing a pointer variable on a pointer that is the address to the 
> user space, and then passing that based pointer to the QMHRDQM API it 
> works nicely, but I've sometimes wondered if there is a better way.

It makes no sense to base a pointer on a pointer for what you're trying
to 
do.  Base a character field or data struture, or whatever you need to
work 
with on the pointer.  Not another pointer!

Anyway, I'm glad you figured it out.