MVS 390, Char columns with complex data types

[roy]

Hi All,
This is a MVS / 390 Related Topic
I'm facing a case where there are db2 table fields of type char(x) that holds an array of records (by occurrs) or a records with redefines.
The redifines are less of an issue since only one type of record is relevant.
The problem is that in those char columns are actually holding a sub-record somposed of both character and numeric data (some holds local NLS chars) and I need to decompose them using DS.

Any help would be apreciated

P.S.
In my desposal are both 390 and server jobs.

Thanks in advance for your time

[ArndW]

Have you looked into using the Complex Flat File stage in server jobs, which were created expressely for this type of data.

[roy]

My problem is that my source is a DB2 table
and correct me if I'm wrong DS only reads CFF files, it can't create them adding the fact I was told they won't build me a "normal" file I can read representing a sain version of the data I'm stuck with 3 people to help me named: MMI (Me Myself & I; sorry Craig , forgive the explained first time used acronim will you )

[DSguru2B]

Have you tried staging the table in a flat file and then using the CFF to read the file?
We have a similar situations in my current shop where the arrays are just jammed into a single column. WHen the time comes to process them, I drop it into a staging file, create a cfd, and let the cff stage do the rest.

[ray.wurlod]

Why not use a Relational Database stage in a mainframe job?

[ArndW]

And since you are in Server, you can read the DB2 table as suggested above, then use a named-pipe sequential file for the cff source and thus don't even need to land your data on disk.

[roy]

The string contains comp-3 elements that do not read well via db2 connect as a string.
There is no logic for determining what kind of record were dealing with, perhaps since it seems only one kind of record is currently used.

I eventually used a substr function in the sql to get the fields I need
Then used a SV to check Is Numeric to validate my numeric data, else I put Zeros and in the derrivation an implicit cast to decimal data type.

As for the Occurrs, a REXX program was built to decompose the data to seperate records (we could have done it in DS just as well, but it wasn't my call).

[roy]

So I think I failed to make cllear what was the solution we used
So to complete the picture from my previous post ...

We used Mainframe jobs to bring the data to Teradata work tables.
that includes handling of all occurs/redefines and such interesting variants.
Once the work tables were populated in a normal relational format we continued the business logic in server jobs.

IHTH,