Charatcer Handling

nirmal.poikaikumaran · New User Joined: 29 Nov 2005 Posts: 66 Location: Bangalore

Hi Team

I have a fixed length file size X(80)

01 DEPT.
05 DEPT-NUM PIC 9(10).
05 DEPT-NAME PIC X(70).

data looks like
-----------------------------------------X(80)----------------------------------
0000000001HR
0000000021INFORMATION TECHNOLOGY
0000003333FINANCE
----------------------------------------------------------------------------------

Requirement is to create a variable length file as per below (~ is delimiter)

1~HR
21~INFORMATION TECHNOLOGY
3333~FINANCE

I understand byte by byte comparisons should be done. But what is the simplest way to do this?

Thanks
Nirmal

Robert Sample · Posted: Fri Oct 23, 2009 5:18 pm

I would say the simplest way would be to use the FUNCTION LOG10 of DEPT-NUM to get an integer power of ten, add 1 to it, subtract that total from 11. Use reference modification from that position to 10 to move the DEPT-NUM, move in the delimiter, then find the length of the DEPT-NAME field (using FUNCTION REVERSE and looking for leading spaces) and move those bytes over.

Why not just keep the leading zeroes and trailing spaces and move in a delimiter between the fields -- the easiest way?

nirmal.poikaikumaran · New User Joined: 29 Nov 2005 Posts: 66 Location: Bangalore

Hi

I have not used the FUNCTION LOG10 & REVERSE. I did not understand that

The thing is I have an input file of 5 Billion rows. Keeping it that way would require a lot of memory. to eliminate it , is what the requirement.

Can you give me a sample of LOG10 & REVERSE

Thanks
Nirmal

Robert Sample · Posted: Fri Oct 23, 2009 5:41 pm

Files do not have rows -- they have records. Data base tables have rows but not records. Are you sure which you have? 5 billion 80-byte records would be 400 billion bytes, so I am assuming your data is on tape? Any way you do the conversion, it will take a lot of time to run.

Click on the manuals link at the top of the page, find the COBOL Language Reference manual, and read up on the functions there. You will learn much more by doing the reading yourself than you would by having me tell you the information.

dbzTHEdinosauer · Posted: Fri Oct 23, 2009 6:05 pm

Robert,

Looking thru Nirmal's previous posts show a pattern of behavior identical to that of this topic.

He does not want to learn how to do something -
he wants a solution that he can cut&paste for his own.

personally, I thought your trick of log-and-field-length to determine first significant character of a numeric field was slick
- I will add that to my bag of tricks, thx.

As far as Nirmal is concerned,
I imagine either another post/topic here (or another forum)
for a cut&paste solution for his problem.

Robert Sample · Posted: Fri Oct 23, 2009 6:20 pm

dbzTHEdinosauer · Posted: Fri Oct 23, 2009 6:26 pm

and the other possible solution:
modify the process that creates the file - but I imagine that is too obvious.

Nirmal,

Craq Giegerich · Posted: Fri Oct 23, 2009 7:51 pm

With "3 billion records" the cost of compacting the data would not be trivial, in addition there is the cost of uncompacting it when you want to use it. If the first field is truly numeric then changing it to packed decimal would cut the size for that field almost by 50%. What are you going to do with this file?

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

Maybe I'm missing something, but what company on this planet would have 5 billion Deparments? Unless it's a Government project.

jctgf · Active User Joined: 04 Nov 2006 Posts: 109

0002001abc

Considering the record above I'd do the following:

dick scherrer · Posted: Sun Oct 25, 2009 4:26 am

Hello,

With that example:

How is w-new-record defined?
How is the variable output file (FD) defined?
How do you tell the WRITE to write only w-length output data?

jctgf · Active User Joined: 04 Nov 2006 Posts: 109

jctgf · Active User Joined: 04 Nov 2006 Posts: 109

STRING W-NUMBER(W-COUNTER:W-LENGTH) '$' W-NAME DELIMITED BY " " (2 BLANKS) INTO MAX-NEW-RECORD.

dbzTHEdinosauer · Posted: Sun Oct 25, 2009 2:56 pm

jctgf,

you have managed to go from 80 char fixed length records to
80 char var length records - which means you are extending the record length
with a vli.

jctgf · Active User Joined: 04 Nov 2006 Posts: 109

yes, you're right.
my memory leaks.
unfortunately the site allows the poster to correct the message only during the 10 minutes following the posting.
I wonder why.
code looks a little long.
little correction may be needed.
my doubt is about the variable-record format in FD (I'm not at work now).

dick scherrer · Posted: Mon Oct 26, 2009 2:29 am

Hello,