Joinkeys to match both conditions

mfchnuser · New User Joined: 21 Sep 2012 Posts: 12 Location: India

I have a file A that either has a 10 digit inc number or 7 digit priority no.

FILA

enrico-sorichetti · Posted: Mon Dec 17, 2012 7:30 pm

how many records in file A ?
if only two records a two step approach using symnames might be the most economical way out of it

from the description noo way with joinkeys
to understand why just look at the syntax of it ...

one key for the first file one key for the second one ==> ONE CONDITION
and You ask for TWO

mfchnuser · New User Joined: 21 Sep 2012 Posts: 12 Location: India

Thanks for your reply

The data given was just a sample

The number of records in FILEA are not fixed. they can vary from 0 to 100.

but the record can be either 10 digit inc number or 7 digit priority no in FILEA.

this needs to match with the same in FILE B , having the 10 digit inc no. at 11 pos and priority no. at 21 pos. B

file B has more than 20k records, i need the unpaired records from FILE B

and i cannot use rexx

dick scherrer · Posted: Mon Dec 17, 2012 7:53 pm

Hello,

Then use 2 steps.

Besides rexx is Not a good tool for matching "many records". It would use exponentially more system resources than the 2 sort steps.

mfchnuser · New User Joined: 21 Sep 2012 Posts: 12 Location: India

thanks for your respnse!!

so you mean to say that JOIN KEYS does not work for multiple conditions?

i wanted 1 sort step since already the current data is a result of 2 sort steps, wanted to reduce the steps

Pandora-Box · Posted: Mon Dec 17, 2012 8:04 pm

What is the key in file B??

daveporcelan · Posted: Mon Dec 17, 2012 8:04 pm

dick scherrer · Posted: Mon Dec 17, 2012 8:05 pm

Hello,

If you explain why there are 2 sort steps before this process, someone may be able to suggest an alternative.

enrico-sorichetti · Posted: Mon Dec 17, 2012 8:09 pm

it would be wiser to review the whole requirement and/or the file layout and content

from the data posted You have NOO way to understand how to join,
You would need anyway to carry on some additional processing ( checking for blanks )

in order to understand how to use the file a <keys> for the join

and even if feasible it adds complexity because of a poor design

daveporcelan · Posted: Mon Dec 17, 2012 11:05 pm

This seems like a slightly different flavor of this thread:

Bill Woodger · Posted: Tue Dec 18, 2012 4:09 am

We know you have SyncSort, but we don't know if the SyncSort you have supports JNFnCNTL files. If it does, you can probably do it in one step.

Doing another Sort step on the 0-100 record file should be a simple solution. JOINKEYS supports multiple keys, but the keys have to be on the same record. So you have to take the pairs of records and make them into one record.

saiprasadh · Active User Joined: 20 Sep 2006 Posts: 154 Location: US

I Used DFSORT for your requirement.

Bill Woodger · Posted: Tue Dec 18, 2012 5:57 am

I've slightly amended Sai's. The F1 is now just a key, and the "other one" of the paired records contains blanks. The GROUP is only established for the records with an "odd number" from the SEQNUM.

enrico-sorichetti · Posted: Tue Dec 18, 2012 12:14 pm

My replies were based on an arbitrary sequence of the FILE A records

I agree that a sequence LONG/SHORT LONG/SHORT ... makes things workable !