Rec count for multiple flat files - PS

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

Hi!

I have a PS file which contains the list of all the flat files for which I need to get record counts.

INPUT: ---> This the PS file containing names of other GDG/flat files

Filename
======
ABCD.DATA.FILE1.G0001V00
ABCD.DATA.FILE2.G0005V00

OUTPUT:

Filename Count
====== ========
ABCD.DATA.FILE1.G0001V00 0000000050
ABCD.DATA.FILE2.G0005V00 0000001000

Since there are huge no of GDG/flat files mentioned in the PS(in 1000s!), its difficult to write individual steps to get record count of each and every file mentioned using the below code:

//SYSIN DD *
OPTION COPY
OUTFIL REMOVECC,NODETAIL,
TRAILER1=(20:'NUMBER OF RECORDS',COUNT=(M10,LENGTH=8))
//*

Also the content of PS file (which contains the list of files for which record count is required) keeps on changing every day. Is there a way to get the record counts for a huge set of flat files mentioned in a PS.

Nic Clouston · Posted: Sun Apr 17, 2011 1:12 pm

Yes - you could use Rexx or sort or a program to read your input file nd generate the necessary job steps/control cards

prino · Posted: Sun Apr 17, 2011 2:05 pm

superk · Posted: Sun Apr 17, 2011 4:29 pm

Prino, by my count, I think there have been about a dozen topics I've read in the past month all on this same theme. Never an explanation, and most seem to end up unresolved.

I don't understand this statement:

expat · Posted: Mon Apr 18, 2011 1:56 pm

But surely if the programs creating these datasets were correctly written, they would display a count of records created, and any programs reading these files would display a count of records read.

Why read the file once again to get the counts when if in the first place things were done correctly there would not be the need.

Bill Woodger · Posted: Mon Apr 18, 2011 7:41 pm

If you really want to go ahead anyway, at least pick on a sub-set of the files first, so the user gets a feel for what you think their requirements are. Say, a sub-set of 50 files or so, you can do that all by hand/qad automation using what you know already.

BUT

I am also at a loss as to why anyone would want to do this. In what way would it be useful to know how many records are on each of a vast number of files?

It seems from your comment about the input file-of-files changing daily, that the intention is to run it daily. I'd like to see the scheduler dependencies for the job(s)! Or will it not matter if some files are from yesterday and some are from today, or haven't been updated for a week, quarter, year, or from a job which abended and has not yet completed, whatever.

You would also need to collate all the outputs. Then stick them in a spreadsheet so that someone can deal with all the flaky figures in a convenient way.

We can continue to pick holes in your request (I'm sure we can find lots more), but it would be more fruitful if we knew what the business requirement was - forget "how to do it" for now, and let us know "what the user actually wants".

dbzTHEdinosauer · Posted: Mon Apr 18, 2011 7:55 pm

his user is probably like the manager who used to be here.
anytime a copybook or model is not longer used,
normal shops remove the element from the production library.
but to him,
that meant 1 less module or copybook
that he could tell others that he was responsible for.

enrico-sorichetti · Posted: Mon Apr 18, 2011 7:56 pm

Bill Woodger · Posted: Tue Apr 19, 2011 4:53 am

enrico, you make an interesting point. You're suggesting that he might be wanting the record counts from thousands of files so that he can do some sort of reconciliation to see where data is being lost or duplicated?

Yoiks. It actually makes sense. I had considered Dick's point, but then thought someone like that must prefer Cylinders to records. I thought of a press release (XYZ Mainframe Services processes billionth transaction this year!), but then why daily (although the daily was only my guess, I suppose)?

And you think all these duplicates requests and counting requests might be the same sort of thing? Mmmm....

OK, lots of self-reconciliation retro-fitted. Outsource it to me, "Old School" solution, 500 euros per program. Your own solution, TBD euros per program.

expat · Posted: Tue Apr 19, 2011 12:54 pm

I actually worked on a project where the count reconcilliations were automated rather than having the operators fill in paper sheets from counts displayed on the screen.

Back in 1979 that was.

Bill Woodger · Posted: Tue Apr 19, 2011 1:50 pm

You learn don't you. It's a computer. Get it to do all the (meaningful, I guess I'd better add, given the thread) work it can.

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

hI All,

By difficult I meant that it would not be advisable to write code to get count for each and every file being reffered in the job. Since these jobs were created long back , the programs which created the datasets donot have logic to display the record counts.

My client Requirement was to display the details of jobs run - JOb id, Job Name, start time, end time, return code etc and all the file names used in the job and their respective counts . The job details could be fetched from SAR unload. SImilarly the file names, but I dont see the file counts getting displayed unless a SORT step is used or the programs used in the job have the logic to display the code.

I believe under such a circumstance, there might be a better solution than what I am thinking (i.e displaying count of each and every file). Need your feedback and suggestions. Please advise how to display counts in such a case where we dont have programs taking care to display the count and we have multiple files used in the job.

Thanks,

Skolusu · Posted: Wed May 04, 2011 11:46 pm

kitchu84,

Please answer all of the questions below

1. What is the LRECL and RECFM of the file that contains the DSN names?
2. Do all the files in the list have same DCB properties (LRECL ,RECFM ..)?
3. Do you have authority to submit a job via INTRDR?
4. What is the max number of files can you have in the list file?

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

hi Skolusu,

1. The LRECL= 150 and RECFM=FB of the file that contains the DSN names
2. all the files in the list have different DCB properties ...

3. Unfortunately no , we donot have access to submit a job via INTRDR
4. The List file can contain at the max 1000 files.

Thanks,
Priyanka.

Skolusu · Posted: Thu May 05, 2011 9:49 pm

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

hi Skolusu,

Still if you could let us know the solution, it wil help us learn.

Thanks,
Priyanka.

Bill Woodger · Posted: Thu May 05, 2011 10:07 pm

Akatsukami · Posted: Thu May 05, 2011 10:11 pm

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

hello Skolusu,

It would be great if you could suggest a solution if we had access to submit jobs through INTRDR ...

will be useful from learning perspective ...

Thanks,

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

@Akatsukami

No we cannot use REXX as well.

Bill Woodger · Posted: Fri May 06, 2011 12:14 pm

Cobol? It's not a complicated thing. If not, what languages do you have?

PeterHolland · Posted: Fri May 06, 2011 12:27 pm

SAS could do the job.

prino · Posted: Fri May 06, 2011 12:55 pm

Let me guess the next reply:

kitchu84 · New User Joined: 02 Dec 2006 Posts: 33 Location: chennai

hi Prino,

Just that I let the group know about the available options that we have.

hi Skolusu - I figured out that its possible to submit jcls via INTRDR. Only my id didnot have access which can be rectified by raising special requests.

Please suggest on the way to determine the record count for a list of files in file when the programs are creating the datasets in the jcl are not taking care to display the file counts.

sqlcode1 · Active Member Joined: 08 Apr 2010 Posts: 577 Location: USA

See if this works...you may have to modify according to your need as below job would work for max. 99 files...