Re: overlaying arrays in v5r1

Hi Ken

I've used this same technique in the past, probably without problems. Bit I think I've been told (maybe Jon Paris) that this technique is not guaranteed to be safe. There is no guarantee that contiguous fields will be brought into the program buffers in the same order - each field is brought into internal buffers individually - unless using LIKEREC or EXTNAME.

Now maybe we could use a pointer to the subfield in the LIKEREC DS as the based pointer of the stand-alone array - then it wouldn't matter, so long as the fields in question remain contiguous in the format.

Whadda y'all think?
Vern

On 1/20/2014 8:41 PM, Ken Sims wrote:

Hi Tom -

On Mon, 20 Jan 2014 15:17:33 -0600, Thomas Garvey <tgarvey@xxxxxxxxxx>
wrote:

Since I needed to use the externally described data structure, which
brings in the original fields as subfields, I could not create separate
data structures to name the subfields again and have them overlaid
(which I would LIKE to do
to avoid the overlay(field:startpos) format of the overlay keyword).
So, I did use the StartPos format. Fortunately, the file this is based
on is mature and won't be changing.

What I've done when the field are contiguous in the record format, and
that aspect will NOT change, however there could be fields added or
changed such that the absolute position in the record format could
change ...

... is to code the array stand-alone, based on a pointer, and have the
pointer defined to initialize to the address of the first of the
fields in the record format.

Ken
Opinions expressed are my own and do not necessarily represent the views
of my employer or anyone in their right mind.