Re: Populating fixed-length fields from IBM DB2 Services OBJECT_STATISTICS

Joe

If you were pulling in an entire table with SQL, you should be able to declare a DS based on the table that SQL can use, not just for native RPG I/O - you hinted at the issue in your original post.

But in this case you are pulling only a couple of the columns, so you have to define the VARCHAR stuff yourself, right?

I feel kind of redundant in my reply!!

Cheers
Vern

On 3/8/2021 4:59 PM, Joe Pluta wrote:

Yes, that's probably what I'll do.  If I use a DS defined like(record) for the file I/O and a DS with VARCHAR for the SQL, I can hopefully do EVAL-CORR between the two.


On 3/8/2021 4:52 PM, Joep Beckeringh via RPG400-L wrote:

Joe,

My 'workaround' is to fetch into a (qualified) data structure with fields that match the definition, and then move the data to database fields (mostly in a qualified data structure as well).

Joep Beckeringh


Op 8-3-2021 om 22:43 schreef Joe Pluta:

Hit my head against a wall for three hours today, so thought I'd bring it to the list to try to save someone else time.

I have a physical file with (among others) fields OZLIB and OZOBJ.  These are the library name and object name, and both are defined as 10A.  Declaring a cursor like this: SELECT OBJLIB, OBJNAME from TABLE(OBJECT_STATISTICS('MYLIB', 'ALL')) works wonderfully as long as I do FETCH NEXT FROM C INTO :OZLIB, :OZOBJ.  But as soon as I stick those two fields (OZLIB amd OZOBJ) into a data structure, I get the dreaded "unusable" message telling me that a subfield doesn't match.

Well, after a whole lot of hair-pulling, I figured it out. The fields in OBJECT_STATISTICS are defined as VARCHAR, and while embedded SQL will convert VARCHAR to fixed-length if you specify individual fields (albeit with a warning at runtime about potentially poor performance), it won't do so on a data structure.  The matching rules are much stricter, probably because it assembles the data from SQL and then moves it as a single block of memory into the data structure (although I'm guessing there).

Anyway, since I don't plan to standardize on VARCHAR fields in my databse, my workaround is to do all the fields individually. When I have time I'll try defining the fields with the same name but with VARCHAR in a qualified DS and then try an EVAL-CORR.