Re: COMPILE ERROR RNF7595

On 2011/1/28 8:48 AM, Vern Hamberg wrote:

I have no idea what is different. What is means is, don't use an *OUTPUT
DS for CHAIN. It is clearly documented that you use *INPUT in certain
cases and *OUTPUT in others.
...
There is no evident difference when I look at them in debug. And
EVAL-CORR works fine between them - probably they ARE the same for data
files. They could be different, I suppose, for display files. Maybe
there is an internal flag that knows the difference. That's the kind of
thing I love to know about, but in the end, it doesn't matter, eh?

There probably isn't any difference at all the the *INPUT, *OUTPUT and *ALL data structures for a database file. There could be differences for a logical file (there can be input-only fields in a logical file) but I'm pretty sure that for all database files, the shape of the three data structures is the same, and the only difference would be that some of the subfields in the *OUTPUT DS would not have names.

They can indeed be very different for display files.

The RPG rule is in place to ensure that the data structure used in the result field is an explicit match for the I/O buffer. It might happen that the *OUTPUT buffer is an exact match for the input I/O buffer, but it's not an explicit match.

That said, I personally would like to see this rule relaxed. Such as an extension of the *ALL support for display files to all the file and record types. (Support for *ALL with EXFMT support was added in 6.1, and support for *ALL with any subfile record I/O was added in a 6.1 PTF as part of the Open Access PTF). See my RPG Cafe blog post about the subfile support: http://preview.tinyurl.com/subfile-all-ds