SED or COBOL to insert a line in many programs?

I'm running MicroFocus COBOL 4.0 on UnixWare 2.1.3.

I'm thinking about a change to every program which would involve
searching for some text that each program has (a section name) and
then inserting a MOVE command 2 lines down.

As I'm running on UNIX is SED a good option (N.B. I've not used it
before) or would clc groupies all prefer to use a COBOL program to
convert other COBOL programs?!

David L.

P.S. There will be other mass changes in the future so that whichever
technique is used will be employed again.

In article <a9026f4e.0406080249.5d26b437@posting.google.com>, david@quantumcat.demon.co.uk 
(David Latimer) writes:
>
> I'm thinking about a change to every program which would involve
> searching for some text that each program has (a section name) and
> then inserting a MOVE command 2 lines down.
>
> As I'm running on UNIX is SED a good option (N.B. I've not used it
> before) or would clc groupies all prefer to use a COBOL program to
> convert other COBOL programs?!

That's "sed".  Native Unix filesystems are case-sensitive.

Personally, I'd probably use awk, partly because IMO it's better-
suited for editing files where the changes are some distance from the
search text (two lines down, in this case), but mostly because I've
used it more often than sed, so I know more of its features and
quirks.

Which suggests that the answer to your question might be to use the
tool you know, unless you have good reason to suspect that the effort
spent in learning a new tool is worthwhile.  There are, of course,
many reasons to learn a new programming language (which is in effect
what you're contemplating), but it may be that none of them are
compelling in this case.

Many Unix programmers, by the way, would use Perl for this.  It's what
Perl was designed for, more or less - text file processing.  (I
wouldn't, because I think Perl is rarely worth the effort or aesthetic
trauma, but some are very fond of it.)

--
Michael Wojcik                  michael.wojcik@microfocus.com

How can I sing with love in my bosom?
Unclean, immature and unseasonable salmon.  -- Basil Bunting

david@quantumcat.demon.co.uk (David Latimer) wrote:

>I'm running MicroFocus COBOL 4.0 on UnixWare 2.1.3.
>
>I'm thinking about a change to every program which would involve
>searching for some text that each program has (a section name) and
>then inserting a MOVE command 2 lines down.
>
>As I'm running on UNIX is SED a good option (N.B. I've not used it
>before) or would clc groupies all prefer to use a COBOL program to
>convert other COBOL programs?!

Clc groupies don't agree on anything. This is a religious issue. I have and
would do it in Cobol .. because I could do it in a hour and it would look
beautiful.

Others might use sed or awk or Perl, which would look yucky.

I would (and have) used sed to do this - it's the simplest way among the
options available for what you want to do.

We use sed and perl scripts every day on our Cobol source, originally on
AIX, and now on Windows using Cygwin.  We have a sed script that we run on
screen section source generated from Acucobol's Acubench to add and subtract
a few clauses.  We have a perl script that I wrote that executes before the
compiler that does some source checking to enforce some standards and catch
some common mistakes.

"David Latimer" <david@quantumcat.demon.co.uk> wrote in message
news:a9026f4e.0406080249.5d26b437@posting.google.com...
> I'm running MicroFocus COBOL 4.0 on UnixWare 2.1.3.
>
> I'm thinking about a change to every program which would involve
> searching for some text that each program has (a section name) and
> then inserting a MOVE command 2 lines down.
>
> As I'm running on UNIX is SED a good option (N.B. I've not used it
> before) or would clc groupies all prefer to use a COBOL program to
> convert other COBOL programs?!
>
> David L.
>
> P.S. There will be other mass changes in the future so that whichever
> technique is used will be employed again.

robert.deletethis@wagner.net (Robert Wagner) wrote

> Clc groupies don't agree on anything. This is a religious issue. I have an
d
> would do it in Cobol .. because I could do it in a hour and it would look
> beautiful.
>
> Others might use sed or awk or Perl, which would look yucky.

Well, _you_ certainly treat it as a religious issue.

I'm an atheist and would use whatever is appropriate whether that be
Cobol, sed, Python or Java.

Functionality is far more important that what RW thinks is
'beautiful'.

If sed does it then do that.

Well at the risk of adding fuel to the fire here, I would strongly recommend
 looking at awk for
this. Not because it is pretty or because of some religious affiliation. :)

Avoid "sed" because it is designed to edit a stream of data very well, but d
oes not lend itself
well to repetitive tasks and/or to more complex programming.  Basically, it 
is great for editing
reports on the way to the printer, or chomping on HTML on the way back to th
e browser, but
for doing things where the rules are a little more complex than a straight u
p substation,
well, awk is a better tool. It was designed to do exactly what you want to d
o in fact.


I any case, here is a little sample awk program and input that I think comes
 close to doing what
you want done. There may be errors in the example code, I typed it in so any
 mistakes are
almost certainly the result of my typing and not the awk program.

Also, there are slightly more elegant ways to do this, but this is the easie
st way to quickly understand. :)  I did add more than
the program actually needs to give you a feel for the structure of an awk pr
ogram. They can get pretty complex and do some amazing
text transforms with very little effort.

Here is the awk program. Put it in a file called something like "test.awk". 
 It will simply add a line to the Working Storage
section of the input file.

****************************** CUT HERE ********************************

BEGIN {
ignorecase
}

/WORKING-STORAGE SECTION\./ {
print $0
getline
print $0
getline
print $0
print "       *ADD USER NAME HERE"
print "        01 WS-USER-NAME                      PIC X(4) VALUE 'PAUL'."
print ""
}

## Default rule - print the line!
/^./ {
print $0
}

/* Do this after all input is processed */
END {
print "         * File processed by awk\n"

****************************  CUT HERE ************************

And here is a test input file. (Mistakes, if any, are typos! :) Put this int
o a file (test.cbl?)
and run the awk program with the following command.

awk -f test.awk test.cbl

The output will show on your screen. You can use standard UNIX stuff to redi
rect it to a file,
or you can do the file manipulations inside the awk program if you want to.

**************************** CUT HERE *****************************

INDENTIFICATION DIVISION.
PROGRAM-ID. AWKTEST.

ENVIRONMENT DIVISION.

DATA DIVISION.

WORKING-STORAGE SECTION.
01  WS-USER-CHOICE                   PIC X(1).

PROCEDURE DIVISION.
START-PROGRAM.
DISPLAY 'HELLO ' WS-USER-NAME
ACCEPT WS-USER-CHOICE
GOBACK.

***********************  CUT HERE ************************************


"Richard" <riplin@Azonic.co.nz> wrote in message news:217e491a.0406082224.6c4dcd30@posting.
google.com...
> robert.deletethis@wagner.net (Robert Wagner) wrote
> 
>
> Well, _you_ certainly treat it as a religious issue.
>
> I'm an atheist and would use whatever is appropriate whether that be
> Cobol, sed, Python or Java.
>
> Functionality is far more important that what RW thinks is
> 'beautiful'.
>
> If sed does it then do that.

Thanks Paul,

It's great to be be presented with code that I can practice on.

I look forward to meeting this particular UNIX utility!

David


"Paul Raulersonv" <pkrauleson@verizon.net> wrote in message news:<KjGxc.120$tA6.66@nwrddc03
.gnilink.net>...
> Well at the risk of adding fuel to the fire here, I would strongly recommend looki
ng at awk for this.

david@quantumcat.demon.co.uk (David Latimer) wrote:

Hi David,

>I'm thinking about a change to every program which would involve
>searching for some text that each program has (a section name) and
>then inserting a MOVE command 2 lines down.

I simply use my editor (VIM) for these kind of tasks. Maybe your
editor has some means of recording keystrokes and writing them to a
file for further use? Then *show* it what to do once and let it repeat
the work on other programs. This way you don't have to learn any new
language.

Since MF Cobol wasn't (isn't?) able to debug object code, I used a
small script to find problems that only occured in object code:

/PROCEDURE DIVISION.*\.
:.,$ s/^.\{6}\ \([A-Z\-0-9]*\)\.$/\0^M           DISPLAY "\1".

This inserts a DISPLAY "Paragraph-Name" as the first statement in each
paragraph. I still use a similar script to trace event driven GUI
programs.

A solution to your problem could look like this:

/^.\{6} FOOBAR SECTION\.
:normal jjO            MOVE A TO B^[
:wq

Search for FOOBAR SECTION.
go two lines down and insert a new line above the cursor with MOVE A
TO B
Write & Quit

In 1996 I added a new file exception handling to about 150 Source
files written by different programmers with a single script containing
editor commands. Today, VIM has a full featured script language with
control structures like if/then/else or while...

> As I'm running on UNIX is SED a good option

Maybe you already use the right tool every day...

Kind regards

Sven

David Latimer wrote:
.....
> I'm thinking about a change to every program which would involve
> searching for some text that each program has (a section name) and
> then inserting a MOVE command 2 lines down.
......
> P.S. There will be other mass changes in the future so that whichever
> technique is used will be employed again.

This sort of job is a natural for SNOBOL4.  Others have put
their $0.02 in for COBOL, sed, Awk, Perl, .... but this is
the sort of thing that SNOBOL4 was designed for and has
proved to be excellent for through the years.

I've used SNOBOL4 for many things, many of them small once-
off utilities to make a bad situation workable, but some big
ones too.  I once wrote a program in SNOBOL4 (Asymbolize --
it's in the SHARE Library) to make it possible to extend a
very big I.B.M. 360 assembler program.  The problem was that
all the references to registers in the assembler source were
coded in numerics, not with symbols, so that the references
to registers did not show up in the assembler cross-reference
listing, and one could noy discover all the uses of the several
machine registers.  My program replaced all the uses of
numerics used to indicate registers (not offsets and constants,
e.g.) by assembly-time variables; during assembly, these
could be substituted for by ordinary symbols which would show
up in the cross-reference.

SNOBOL4 is rich in the ability to set up patterns and to match
character strings against them, taking actions as the match
proceeds or fails, or after a whole pattern match has succeeded.
It has the built-in ability to create and look up in tables such
as dictionaries and symbol tables, and it's easy to write whole
compilers in SNOBOL4 (as has been done in many compiler-writing
courses).

Do visit  http://www.snobol4.com ; there, you can get on the
SNOBOL4 mailing list, buy one of the excellent SPITBOL compilers
for SNOBOL4, or download a free copy of the sample Vanilla
SNOBOL4 system (it's adequate for many small jobs!).  The
documentation that's included in the Vanilla SNOBOL4 system is
a really great introduction to SNOBOL4, and the documentation
that comes with SPITBOL is superb.  There are also links at
the site to other implementations of SNOBOL4, and a library of
contributed programs in and connected with SNOBOL4.

All the best!

John.

--
John H. Lindsay  jlin_DELETE_THIS_SPAM_ZOT_dsay@kingston.net
48 Fairway Hill Crescent, Kingston, Ontario, Canada, K7M 2B4.