Re: [open-cobol-list] possibly bug: accept into numeric field

[Thomas Biehler]

On Friday 07 November 2003 23:55, Keisuke Nishida wrote:
> At Tue, 4 Nov 2003 13:54:39 +0100,
>
> Thomas Biehler wrote:
> > i am not sure, but there is possibly a bug in open-cobol
> > in the accept statement.
> >
> > The problematic case is an accept into a numeric field
> > when inputing less characters as in PICTURE-SIZE given.
> > The actual behaviour is the alignment to left (ok!) and
> > the filling up to the SIZE of the field with SPACES! ( IMO not ok)
> >
> > I read in both ANSI-Standards (1985 / 2002) about the ACCEPT-Statement.
> > As far as i unterstand it, only the left alignment is ok.
> > These standards are saying nothing about "alphanumeric fillup".
> > (The MICRO-FOCUS Compiler does not makes the "alphanumeric fillup"
> > but IMHO the ANSI-Cobol Standards are not very precise in the
> > Specfication of the ACCEPT - "Format 1"; perhaps it is question of
> > interpretation ?! )
>
> I checked the COBOL 2002 standard right now.  It says
>
> | 14.8.1.3 4) a) If the size of the receiving data item [...] exceeds
> | the size of the transferred data, the transferred data is stored
> | aligned to the left in the receiving data item [...], and additio
> | data is requested.
>
> so the correct behavior might be to read additional characters from
> the terminal.  Is this what you expect?  What other compilers do about
> this?
>
> Keisuke
>

Hi Keisuke 

I agree with you.
The "correct" (= standard conform) behavior in COBOL 2002 
and COBOL 1985  Standard (Nucleus Module Level 2 = (only in "High"-Subset))
is to read additional characters.

In COBOL 1985  Standard (Nucleus Module Level 1 = ("Minimum" and 
"Intermediate"- Subset):
6.5.4  
 Format-:1
  (4) ... In  level 1, only one transfer of data is provided.

But this "old" behavior was at least "problematic". (IMO errorprone)
As i can remember me, you have to handle this case with 
other statements after the ACCEPT in older "COBOL-Standard's"! 
I have talked with other experienced cobol-programmers 
and they have commited this. 

I have some suggestions: 
===================
Perhaps you should start a thread about the "standard-conform" 
behavior from ACCEPT (FORMAT-1) in the comp.lang.cobol newsgroup. 
(I can do it also and act as a "filter" :-) ; but a shorter way is a 
more faster way to the real solution)

Or you can ask the developper from the tiny-cobol project 
why they implemented it as i now works?! 

Other compilers: (which i have acutally access to them)
========================================
I have tested the behaviour in MF-COBOL and with tinycobol 0.61.
Also, a friendly colleague on a BULL DPS-7000 (GCOS 7)
with that "Cobol-85" Compiler. Always the same intuitive behavior!  :-))
(A old COBOLA / COBOLM  = BULL DPS-6000 (GCOS 6) which 
i have used long time ago, is no longer in use. I could have a look
in the handbook in the archive, but i think this does not make sense!)

The intuitive behaviour:  (conform to ... ?!)
==============================
From the user sight it looks like a 
numeric MOVE from a shorter Field to a longer Field !!!!
(Right alignment , Fillup with Zero's)   
But only one request to the User is made
(ACCEPT from  Terminal / SYSIN / Console ...)
  = only one Transfer of Data !?

(But this seem's to be compiler differences from the ANSI standard 
(= Compiler Dialects ?!  / non-conforming behavior ?!))

I hope, it is now clear why i can not say 
what I expect, only in a simple sentence.
I don't now what i should request from you.
(I have some ideas but i would like to see your answer first)
In general, i like compiler that follows standards !
(And i like good standards too :-) )

Further discussion with other people is necessary.

Sorry for the confusion and thanks for your patience!

Thomas

P.S.  For me it does not hurry !   
        (Only ACCEPT FROM ENVIRONMENT-... are used in our application)