Reading COBOL Variable length file with multiple layouts

venuhunev · New User Joined: 26 May 2007 Posts: 70 Location: chennai

Hi,

I have a problem in reading a variable file in a COBOL program which uses copybook having multiple layouts.

My actual file record length is 151 (155 in JCL) and Block size is 23467. since the file is existing and cannot change attributes, I am not sure how to solve it.
No Abends or any other errors. I am splitting detail 1 and detail 2 record along with header record. So I just read it and direct the input to specific output.

Following is an example of the problem what I face.

Bill Woodger · Posted: Mon May 21, 2012 1:41 pm

What are you DISPLAYing?

With multiple types of record, you should have something on the record to tell you which layout to use. If you use the wrong layout to DISPLAY a particular record, then you well get more/less data shown than expected (sometimes, when lucky, a S0C4).

Nothing to do with anything except defining more/less data than exists (ie using the wrong record-layout, ie coding error - at least from what you have explained so far).

Robert Sample · Posted: Mon May 21, 2012 2:40 pm

As Bill says, depending upon what -- exactly -- you are doing, it does not appear there's an error or even anything to change. The system fills the buffer when it does a read; if your COBOL program attempts to reference data past the length of the currently defined record then you'll see data from the next record(s). This behavior is the way COBOL works and has worked for many, many, many years -- so there's not much chance you are going to be able to change it.

Bill Woodger · Posted: Mon May 21, 2012 2:55 pm

Just to add, you blocksize is not very good for a 3390 track. Wasting 10k per track, or close to 1/6 of the space available.

venuhunev · New User Joined: 26 May 2007 Posts: 70 Location: chennai

I am displaying all 01 variables when reading each record.

READ INFILE.
DISPLAY RECORD.
DISPLAY HEADER.
DISPLAY TRAILER.
DISPLAY DETAIL1.
DISPLAY DETAIL2.

Bill Woodger · Posted: Mon May 21, 2012 3:11 pm

Well, there's your problem.

I don't know what RECORD represents, presumablye you've carefully hidden the real name from us, as RECORD is a reserved word, but the other record areas would seem to relate to the first byte of data.

So, 88 for the first byte

venuhunev · New User Joined: 26 May 2007 Posts: 70 Location: chennai

Thanks for your replies.. sorry, I have edited my reply.

I actually dont need display in my code. I just did it to know what was wrong.

My file is Variable (151). Atleast one of the record layout should be 151. So I have a 01 varaible which is RECORD.

File Structure:
header record layout is 40
detail1 record layout is 50+10 (where last 10 bytes are not mandatory kinda comments)
detail2 record layout is 40
Trailer record is 40.

When i execute READ INFILE for the first time, it takes 40 bytes of HEADER record as the layout is for 40 bytes.
But my detail1 record layout varies as the last 10 bytes are not mandatory. so my detail1 record could be 50 bytes or 55 bytes.

Consider my first detail1 record which is 50 bytes, but my read statement populates 50 bytes from first detail1 record and 10 bytes of next record.

dbzTHEdinosauer · Posted: Mon May 21, 2012 3:30 pm

all you have provided us is the info that you have no idea what is going on.

you need to provide, FD, i/o statements (at a minumum)
before anyone can provide you with the info that you need.

Bill Woodger · Posted: Mon May 21, 2012 3:37 pm

We didn't really expect that you were only interested in DISPLAY. The comments apply to the records, not to any Cobol verbs.

You started off by telling us you had multiple layouts in a copybook. What you showed were different sizes, so you'd have a variable-length-record file.

Now, possibly, you are telling us that the records themselves can also be different lengths but the same type.

If that is what you mean, then there must be something on the record which distinguishes between the "fixed" part of the record for a particular type and the "variable" part, and within the fixed part there should be something which, usually indirectly, indicates the length (as in presence) of the variable part.

If you have 10 bytes which are there sometimes and not there at others, there must be something to indicate that. Only access the extra 10 bytes when the "something" is true.

venuhunev · New User Joined: 26 May 2007 Posts: 70 Location: chennai

FD INFILE
LABEL RECORDS ARE STANDARD
RECORDING MODE IS V
BLOCK CONTAINS 0 RECORDS
RECORD 151 CHARACTERS.
COPY COPY1.

I gave 151 characters like FB. I think I should give RECORD VARYING 40 to 151.

Bill Woodger · Posted: Mon May 21, 2012 3:44 pm

The shortest record you showed previously was 20 bytes.

If that is the FD you ran with, and managed to display all your records to the end, you'll not get much different if you do 20/40 to 151.

You have to use, in your program, every time, only the data which belongs to the current record.

To do that, you have to know the record type.

And if the record, by type, can be variable you have to know what tells you it is variable, and use that. Normally this would be an indicator or a count.

Only access the variable part if you know there is data there belonging to the current record.

Otherwise you will be copying data from the subsequent record(s) and even, on occasion, accessing storage you are not allowed to (possible when final record in final current block/buffer) and getting an S0C4.

Before looking at anything else, know which record it is, and use the correct layout.

Before looking at anything variable, know if/how many are present.

No problem, just normal stuff.

dick scherrer · Posted: Mon May 21, 2012 7:23 pm

Hello,

READing INTO a working-storage variable that has redefines for each record type may help you reduce confusion(personally, i believe this is a better way to code also).

Bill O'Boyle · Posted: Mon May 21, 2012 8:43 pm

Click here - www.ibmmainframes.com/viewtopic.php?p=191731&highlight=#191731

In the FD, you only need a single 01 area defined to the largest-length record.

As Dick has suggested, issue your READ into the largest WORKING-STORAGE area, defined to the largest maximum-length.

Then, check the value in WS-INPUT-LGTH and you'll know which type of record you have, because it will be set to the length of the record just read.

Once you've determined the record-length, you can them move it to the correct area.

venuhunev · New User Joined: 26 May 2007 Posts: 70 Location: chennai

thanks everyone for the suggestion... infact.. solution.... it worked

dick scherrer · Posted: Tue May 22, 2012 9:05 am

Good to hear it is working - thank you for letting us know

d

Bill Woodger · Posted: Tue May 22, 2012 11:23 am

Mmmm.... I was afraid of that.

You have a max record-length (data) of 147 bytes. For more than 100 bytes, with the definitions you have shown, you are effectively going to "guess" whether theres is something there or not.

For that extension, you have more than one field. If more than one field of those more than one fields is not there, then you are lost.

Good luck.