Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!weyrich!orville
From: orville@weyrich.UUCP (Orville R. Weyrich)
Newsgroups: comp.lang.fortran
Subject: Re: Where can I find a Fortran style guide?
Message-ID: <1991May24.031430.1891@weyrich.UUCP>
Date: 24 May 91 03:14:30 GMT
References: <2167@tharr.UUCP>
Reply-To: orville@weyrich.UUCP (Orville R. Weyrich)
Organization: Weyrich Computer Consulting
Lines: 48

In article <2167@tharr.UUCP> gombo@tharr.UUCP (Alun Jones) writes:

>I've reached the stage in my conversion of a large suite of Fortran that I
>feel it would be useful to inform my bosses of where I feel their program is
>'badly' written.  Although I don't have any syntax problems with their code,

I did that once, and my boss took it as a personal attack. [He had written
much of the code himself, in FORTRAN-II]. I lost my job. Hope you fare
better.

>there are one or two things that I feel are bad style - using too many
>implicit variables, assuming variables are initialised to zero, mixing and
>matching variable types in common blocks, that kind of thing.

Avoiding implicit variables doesn't help style much, unless your compiler has
an "IMPLICIT NONE" statement that will trap unexpected implicit variables
arising due to typos.

Assuming variables to be initialized to zero is definitely non-portable. 
What hardware platform are you using? It is non-standard to expect unitialized
variables to be zero.

Using COMMON to purposely overlay variables of different types is non-standard
and will cause trouble porting to machines with different word sizes.
The standard does permit defining different sets of variables in different
(unnamed) COMMON statements, but only in the case that the intent is to
overlay working storage usage by different routines [an obsolete usage that is
not necessary in virtual-storage systems]. 
It is definitely against the rules to do this to accomplish type-conversions, 
etc.

>Now, most compilers will quite happily accept these as valid, although I know
>at least one is against ANSI standards.  But I'm fairly sure they're 'bad'
>style.  Now, I can't really say that to my bosses without some convincing
>arguments, so I'm looking for a Fortran programming style guide.

I would suggest "FORTRAN Poisons and Antidotes" by Brian T. Smith, in the
book "Portability of Numerical Software".  It is Springer-Verlag Lecture Notes 
in Computer Science #57 (1977). The whole book is good reading.



--------------------------------------           ******************************
Orville R. Weyrich, Jr., Ph.D.                   Certified Systems Professional
Internet: orville%weyrich@uunet.uu.net             Weyrich Computer Consulting
Voice:    (602) 391-0821                         POB 5782, Scottsdale, AZ 85261
Fax:      (602) 391-0023                              (Yes! I'm available)
--------------------------------------           ******************************