sequential file processing

imain · New User Joined: 25 Jun 2006 Posts: 14 Location: US

hi all,

i have two files in sequential format and also sorted
1.transaction file.(taccno,tdate,tamt and status(credit or debit)

2.master file.(accountno, name, balance)

I am reading transaction file and tries to find the taccno in master
file,for the first time it works well and good but for the second time
the master file will not read records previously read how can i solve
the problem.can i use file status codes to handle this condition.

thanks in advance

DavidatK · Posted: Thu Oct 19, 2006 11:58 pm

If the files are already in sort order, why do you need to read previously read records?

Dave

vijayamadhuri · Active User Joined: 06 Apr 2005 Posts: 180

one way u can do is
load u r master file into an array.search (use search command) to search the array (keep incrementing the index( ix) till u find a match.once u find a match reset the index .

so for u r second record u would be reading the file from the begining..

i think the other method is binary search...

hope that helps...

DavidatK · Posted: Fri Oct 20, 2006 2:48 am

VM,

Your solution is possible for small files, however not practical. What happens when the master file cantains large numbers of records. i.e. 150,000,000 records? And the transaction file contains 10,000,000 records?

The solution for "imain" is a single pass through the master file and the transaction file.

Dave

azshukla · New User Joined: 31 May 2006 Posts: 9 Location: Pune

Hi DavidatK
The problem faced by imain is perhaps that he wants to search Accounts in master file corresponding to each taccno in transaction file. If it is one to one mapping then definitly he doesn't need to read a previously read record again , but if it is not one to one mapping there would be multiple records in any of the file corressponding to other file. in this case it is inevitable to do a search from starting each time.. i don't think there is any other possible way.. and if it is.. please share with us..
Mainframs are used for handling large files.. and thats thier advantage that they can handle these lengthy files.. if there is no other way, we will have to sacrifice the performence..

MFRASHEED · Active User Joined: 14 Jun 2005 Posts: 186 Location: USA

Let us look at this with functionality prespective.

We have a Master file (Assuming one record per Account)
We also have a transaction file (Assuming multiple records per Account)

If both files are sorted on Account# what would be need to look back on master file? I guess probably I don't understand what is need here can you clarify imain.

I would agree with DavidatK. Table processing for large files is bad design. In our shop we used to have program where a large 2 dimensional table was defined which increased the program run time 5 folds. This table was so large there were region issues, missed timelines etc. This situation was resolved by spliting program processing in two runs. If there is need for look back, these records can be written to a seperate file and next program will process only these records. And believe me this is much better than trying to do everything in one program. I guess if we understand the requirements a better solution can be thought of.

Even though mainframes can handle large volumes etc, but applications should be designed to make best use of available resources. MIPS add up and can prove costly and hurt bottom line.

azshukla · New User Joined: 31 May 2006 Posts: 9 Location: Pune

Well i think these all talks are useless... untill you have the requirement , you can't design.. and what i believe assumptions made by a vender are the most serious mistakes done by their side.. untill you have a clear requirement .. you can not go ahead with assumptions.. for the lrage chunk of data pple usually keep them in segmented files which are again interlinked.. it is a complex design.. and i'm not sure whether IMAIN asked for that... his problem is
how to read the previous read record and do you have any other way to do that?? or else there are always better sollutions..coz nothing is perfect!!!

dbzTHEdinosauer · Posted: Sat Oct 21, 2006 3:08 am

I think these forums are not useless: I can determine from answers who knows what they are talking about and those who want to just talk.

Actually, the problem is reading the previous record for sorted input files means faulty design/programming.

and your are right, nothing is perfect ........ but giving someone a means to do something slick, when the program sucks in the first place, is useless and does not aspire to the philosophy of these types of forums. So, I will continue to answer questions with what they need to know, not what they what to know.

yeah, I know. means someone has to learn something or do something different. some point or another you have to start over if what you have is wrong.

dbzTHEdinosauer · Posted: Sat Oct 21, 2006 3:51 am

Basic Update 101----- Transaction file is the trigger file > T = t in case one forgets.

Masterfile is only read when the transaction file presents an account number greater than the masterfile account number. You never read backwards in sorted sequential file processing.

Sometimes I think that random access was developed because so few could write a logical program.

If anyone is offended by my remarks, I don't apologize. This level of silliness does not require diplomacy.

mmwife · Super Moderator Joined: 30 May 2003 Posts: 1592

Hi Imain,

This is one of two batch pgm algorithms that comprise more than 90% of batch pgms in existence today... so you better learn the concept.

The other one is the control break algorithm.

You begin by sorting both files on the same key.
I've included code as an example (sorry, I couldn't find how to use the attach function):