IBM Mainframe Forum

Read the DFSORT tricks but...

Problem:
Inputfile up to 200k rows (fb10) can also be 100 rows

Want that file to be split into x outfiles with maximum 20k rows in each.

(The "tricks" shows how to split into x files (x=hardcoded amount))
Can You help me with this?

//Lasse

Lasse,

20k each file with an lrecl of 10 would only have (20 X 1024 = 20480 /10) = 2048 Records. You would be creating 100 datasets to accommodate all the rows. You can define all the 100 datasets and use SPLIT1R=2048 to split the records. If you are concerned about the number of empty datasets , you can run an IDCAMS step to check for an empty file and delete all the empty files in a single step.

Is not the TS looking to create output data sets with a maximum of 20K records (which he miscalls "rows") rather than 20K bytes?

Sorry for the miscall, yes, I ment 20k records.
(200.000 records will end up in 10 outputfiles, with 20.000 records each, and 21.000 ends up in 2 files)

LasseH wrote:Sorry for the miscall, yes, I ment 20k records.
(200.000 records will end up in 10 outputfiles, with 20.000 records each, and 21.000 ends up in 2 files)

It is still not clear as to how you want to split. Do you dynamically allocate the output files each with 20k records?

ex:


If that is indeed the case as I mentioned earlier define the max outfil files (100) and then in the next step use IDCAMS to delete all the empty datasets which is the simplest task.

Another solution is to read the file, get the count of records and then dynamically generate the ICETOOL JCL and submit it via INTRDR.

Thanks, I go with the "create a lot, delete after"
//Lasse