merge files with duplicate in both the files...

Escapa · Posted: Thu Feb 05, 2009 3:12 pm

Input1:

Escapa · Posted: Thu Feb 05, 2009 7:14 pm

I searched in dfsort forum for it and it looks like this is the case of cartesian join.
And replies on such topics tells
cant be done using splice (dfsort)

what will be best suited to do so?
any thoughts/suggestions?

Skolusu · Posted: Fri Feb 06, 2009 12:21 am

Sambhaji,

There is a DFSORT solution for Cartesian join. I need the following details.

1. What is the LRECL and RECFM of input files
2. What is the position and format of the key to be matched
3. Which file has the max dups on the key?

Escapa · Posted: Fri Feb 06, 2009 11:52 am

Thanks Kolusu for update.

Details you required are as follows

file1:
LRECL=45
RECFM=FB
max no of dups=55

file2:
LRECL=32
RECFM=FB
max no of dups=10

Skolusu · Posted: Fri Feb 06, 2009 10:33 pm

Sambhaji,

Since you did not mention the key position and format of the key, I assumed that it is in the first 4 bytes in character form in both files.

The format of output file is

Escapa · Posted: Tue Feb 10, 2009 3:11 pm

Hi Kolusu...
Thanks a lot...
I tried this solution for 10% of records then works fine but when i tried with actual file it is failing with EB37 on tmp file
I tried increasing space but still result was same..

dick scherrer · Posted: Wed Feb 11, 2009 2:17 am

Hello,

Suggest you talk with your storage management people about your "special" requirement. . .

You mention the abend being related to the "tmp" file, but there is no tmp dd. . .

Terry Heinze · Posted: Wed Feb 11, 2009 4:13 am

EB37

Arun Raj · Posted: Wed Feb 11, 2009 7:36 am

Sambhaji,

The logic used here involves repeating 'n' times the records in file-2, where 'n' is the maximum number of duplicates in file-1 and storing it in T1. It would be better if you post the actual number of input records and the value of 'n' from your job run(Search for REPEAT in your DFSMSG output).

Escapa · Posted: Wed Feb 11, 2009 11:59 am

Terry Heinze · Posted: Wed Feb 11, 2009 12:08 pm

Did you look up what an SB37 is? Did you look up the accompanying IEC030I message and the reason code?

Skolusu · Posted: Wed Feb 11, 2009 10:14 pm

Sambhaji,

Your input files are huge. I do not think the REPEAT solution is viable for this job as we are repeating every record 69 times. Your input count shows more than 11 million and the repeat solution would create roughly about 826 million records which is the reason for your SB37 errors.

We have an COBOL program for such requests. Please contact me via e-mail offline (skolusu@us.ibm.com). Make sure that you send me the following details

1. LRECL and RECFM of both files
2. Position of the key and format in both files
3. OUTPUT file layout ( what fileds do you need to pick from file 1 and file 2)
4. Do you also need the unmatched records from both files?

Escapa · Posted: Mon Feb 16, 2009 11:00 am

Thanks Kolusu. I got it.
Working much faster than expected timings..