Re: Overlaying a DS Array In a LIKEREC DS

[Barbara Morris <bmorris@xxxxxxxxxx>]

Dennis Lovelady wrote:
> ... *BUT* I don't know what
> happens when the READ is to a data structure.  Seems to me at that point
> that the fields should be populated.  If they're blank in that scenario,
> then it seems to me that either there's mystery happening and the
> READ-INTO-DS doesn't work like it appears to be designed (which I doubt), or
> the compiler is wiping the unreferenced fields (which I also doubt).
>
> But I think (if this is the issue) that you can hand-define those fields all
> you want, and it won't eliminate the 7031 "error," and therefore neither the
> compiler's propensity to remove them for efficiency's sake.
>
> Barbara, am I off base here?

When the READ is to a data structure, the input buffer is copied to the 
data structure, so the entire data structure is changed, whether or not 
the individual subfields are referenced or not.

The 7031 behaviour only applies to whether fields are loaded during 
I-spec processing, which isn't used when there's a result DS.

Fields are sometimes loaded even when they have a 7031.  For example, 
subfields, when the data structure itself is used, or any overlaying 
subfields is used.  There are other reasons that would cause a field to 
be loaded, such as its being exported, or a parameter, or based, etc. 
The compiler can only skip it during I spec processing if it's sure the 
field's value can only be accessed through the field itself.