Get 1/10 of a file

vicky11121982 · New User Joined: 21 Jun 2007 Posts: 9 Location: gurgaon

I was asked this question of getting 1/10 of file of millions of records in a flat file, record number varying in each run, through COBOL code. First 10 bytes of the code has sequence numbers. Using read and count was discounted as it will have to read the whole file. Can anyone suggest anyother way?

Antonio Barata · Posted: Thu Jul 10, 2008 2:49 pm

Hi
Have you considered using SORT/ICETOOL instead of application program?
I think that it could solve it with much less effort.
Try to put the question in the DFSORT/ICETOOL forum.
Regards

gcicchet · Posted: Thu Jul 10, 2008 3:06 pm

Hi,

why do you need to read the whole file unless you don't know the count before hand, then you don't have any choice whether it's cobol or any other tool.

I'm not sure having the sequence number matters, you read and count and stop when you have reached your target.

Gerry

vicky11121982 · New User Joined: 21 Jun 2007 Posts: 9 Location: gurgaon

While digging found READ LAST.Can we make the access mode as dynamic and use READ LAST to read the last record and get the record sequence number.Dividing that by 10 will give us our limit till the count can reach.Any suggestions??

superk · Posted: Thu Jul 10, 2008 4:39 pm

vicky11121982 · New User Joined: 21 Jun 2007 Posts: 9 Location: gurgaon

My bad..

.. you are right Kevin it was for Integrated Language Environment COBOL (ILE COBOL) ....link:http://publib.boulder.ibm.com/infocenter/iadthelp/v7r0/index.jsp?topic=/com.ibm.etools.iseries.langref.doc/c0925395485.htm

gcicchet · Posted: Thu Jul 10, 2008 6:11 pm

Hi,

I take back what I said earlier, if you have FILEAID you can get the last record by using the following

Craq Giegerich · Posted: Thu Jul 10, 2008 7:37 pm

Which 1/10 of the file do you want, the first, the last, a random select or just 1 record of each 10 as the file is read?

enrico-sorichetti · Posted: Thu Jul 10, 2008 7:40 pm

that' s just an interview question with no business logic behind

Robert Sample · Posted: Thu Jul 10, 2008 7:50 pm

Enrico: actually, I've had to do this for business reasons -- statistical sampling of the input file; we had to pull out the sample, run some statistical analyses on the sample, and then compare the sample results to the full file results to verify the sampling tests were providing valid results. However, statistical sampling usually only needs a few thousand records; pulling 10% of a multi-million-record file would be overkill.

Vicky: you're not going to get away from having to read the file to pull the records. In COBOL you can use FUNCTION RANDOM to return a random number and use that for selecting the records.

enrico-sorichetti · Posted: Thu Jul 10, 2008 7:53 pm

I was not implying the lack of business reasons as a general point,
it was just the way the question was posed

vicky11121982 · New User Joined: 21 Jun 2007 Posts: 9 Location: gurgaon

Thanks Robert and Gerry

Enrico:Even i/w question have answers and I was just curious to find one.

enrico-sorichetti · Posted: Fri Jul 11, 2008 12:26 am

the point is that answering interview question has a different approach...
it' s more like puzzle solving,

when I ask what is the business need I want to know as much info as I can in order to provide the best answer

and whenever, like in the original question, also the how is specified
I have the impression that something is not going the right way

vicky11121982 · New User Joined: 21 Jun 2007 Posts: 9 Location: gurgaon

hanucob · New User Joined: 15 Jul 2008 Posts: 3 Location: bangalore

Hai!!
My answere is like write a query using count(*) ,then u will get
no of records in it,then divide it by 10.......then can read required records
...

dbzTHEdinosauer · Posted: Wed Jul 16, 2008 11:08 am

the topic is: Get 1/10 of a file

you can not execute sql against a non-db2 organization.