Sum based on criteria for each id, each name

Soumya Bhat · New User Joined: 01 May 2020 Posts: 1 Location: India

Hi All,

I have a query. I want to know if this can be achieved by using JCL or a COBOL program is required. If JCL or COBOL, then what is the way to achieve it? I m trying to do via COBOL program. Its seems to be getting more complex than required.

Please find the more details below:
INput file ---->
1st column is id, 2nd column is type, 3rd column is record type, 4th column is value, 5th column is name

123 0000 000 00000 -> Header rec
123 0001 002 00555 XYZ
123 0001 002 00600 XYZ
123 0001 002 00200 ABC
123 0001 002 00250 ABC
123 0008 002 00050 XYZ
123 0009 002 00010 LMN
234 0000 000 00000
234 0001 002 00200 XYZ
234 0001 002 00250 XYZ
234 0008 002 00050 OOO

Output criteria:
1. no change in header record, should be copied as it is
2. Total for value should be done only when type is '1' and record type is '2'(detail record)
3. Different total should be done for each 'id'
(a) within each 'id', different total for each "name"
3. rest all records other than type '1' should be written as it is

Output record:

123 0000 000 00000
123 0001 002 01155 XYZ
123 0001 002 00450 ABC
123 0008 002 00050 XYZ
123 0009 002 00010 LMN
234 0000 000 00000
234 0001 002 00450 XYZ
234 0008 002 00050 OOO

daveporcelan · Posted: Sat May 02, 2020 12:20 am

I am sure you can do it with COBOL, it does not appear too complex.

I am sure you CAN NOT do it with JCL, as all IT professionals know that JCL can ONLY call a program to perform a function.

With that being understood, I was able to do it easily with Syncsort.

I ignored most of your 'criteria' and assume that your input data is a correct representation of reality (probably foolish of me).

By this I mean that 0000 000, 0008 002, 0009 002 are never duplicated for a particular ID and name.

If this is indeed true, then here is a very simple Syncsort solution: