How do we handle the carriage return in the cobol program

dick scherrer · Posted: Sat Feb 03, 2007 1:52 am

Possibly. . . .

From the lack of data, i was never sure i knew where the "break" was or if one eventual record could be "broken" in multiple places in the input.

Be interesting to learn what actually "fixed" this.

superk · Posted: Sat Feb 03, 2007 2:06 am

The first couple of times I read the post, I thought it was talking about handling "carriage control", and I couldn't for the life of me figure out why that would pose a problem.

dick scherrer · Posted: Sat Feb 03, 2007 2:18 am

Yup,

My first thought reading the initial post was that the cr/lf was coming in as data and the new output needed to be positioned accordingly by the program.

rahul_krg · New User Joined: 08 Aug 2007 Posts: 7 Location: India

Hi All,

After searching for the handling of CR/LF characters in EBCDIC characters, I came through this thread. I'm facing the same problem of getting the output record splitting in number of records depending upon the occurances of CR code (X'0D') in the file. This CR character will not be there in the final output in the file, but it was part of the record written to file. I've observed that if there are 3 CRs in the record, it gets splitted in 3 individual records.

Please help in decoding this problem of handling CR in Mainframe. I'm using ZFile to generate MVS datasets through JAVA program. My record contains some COMP field.

RG

dick scherrer · Posted: Fri Mar 14, 2008 10:28 pm

Hello,

rahul_krg · New User Joined: 08 Aug 2007 Posts: 7 Location: India

Hi Dick,

Thanks for the reply. Answering your question, as we are creationg the MVS dataset using ZFile API (IBM proprietary package)through Java, attributes are defined on run-time depending upon the record structure. Datasets are of type Variable block & attributes like lrecl, storage space are calculated on the fly. But the problem is quite general and will be there in any kind of dataset.

To describe the issue elaborately, as the record contains COMP fields we can get CR characters as a part of data. For example, decimal value 269 will be stored as X'010D' for a COMP field. Now X'0D' is representation of CR character in EBCDIC characters. So on encounter of this byte X'0D' record gets splitted in 2 lines, whereas this is a data and not the CR.

So I want to avoid this situation where CR is part of data. I can't avoid the occurances of CR in the data since they come from the data source.

I hope the situation is more clear.

Thanks.

enrico-sorichetti · Posted: Mon Mar 17, 2008 1:00 pm

rahul_krg · New User Joined: 08 Aug 2007 Posts: 7 Location: India

Hi Enrico,

Thanks for the reply. It seems that the situation can't be avoided. But the output of my work now depends on avoding CR character coming in data. Since the whole output dataset get garbled because of splitted records.

When I try to use binary mode for writing records, it puts everything in one records till maximum records length. This is not acceptable since I'm writing variable length records.

Is ther any solution or work-around for this. I hope there is a solution for every problem.

Thanks for the help.

enrico-sorichetti · Posted: Mon Mar 17, 2008 1:20 pm

As I see it

first of all too many confusing posts....

are You working :
under USS with unix like / stream oriented data management
under MVS record oriented data management...

You cannot mix the two architectures for the same dataset

if the data is text only then it might be ok to use the pc approach for record delimiting,
but remember ...
in a mainframe environment the standard record terminator is x'15'

anything else is ... garbage or pure binary data

if You have in Your record layout binary/comp data
You might need to review the application specification

enrico-sorichetti · Posted: Mon Mar 17, 2008 1:23 pm

/rant on

in this thread issues are posed by three different people, adding confusion
and making things harder to understand

I wish people would start theyr own threads for problems

so that the responders could face just one issue at the time ( expecially the terminology)

/rant off

enrico-sorichetti · Posted: Mon Mar 17, 2008 2:11 pm

dick scherrer · Posted: Mon Mar 17, 2008 4:35 pm

Hello,

The way we prevent this problem is to make sure that COMP data is not created on the UNIX or Win-based systems.

rahul_krg · New User Joined: 08 Aug 2007 Posts: 7 Location: India

Hi,

COMP data will be there since its a part of Record structure. I'm reading a the data file (which is a MVS dataset) corresponding to the record structure and after processing it, writing back to a new MVS dataset. My Java application is running under USS i.e. OMVS.

and yes, Enrico, docs are useful resources and I've already gone through all the docs related to ZFile. Also I've tried both the record mode and stream mode to write ZFile. The method "void write(byte[] buf, int offset, int len)" doesn't work really as it seems. It writes one record and terminates the writing operation as it reaches the end of buffer. I couldn't find the length of individual record as it was variable. With this method calculation of the value of the argument 'len' is not known, hence it also didn't help. Till now I couldn't write a variable length record in binary mode using ZFile with all possible combinations. Hence I was propelled to use text mode which then pose with CR character problem.

Well thanks for your replies.

dick scherrer · Posted: Mon Mar 17, 2008 5:45 pm

You're welcome,