FORTRAN 2003 Accepted as Standard

GraWil writes "Despite the nay sayers citing its death in 1965, the FORTRAN standards committee has now released the final FORTRAN 2003 specification. In an announcement to the comp.lang.fortran group, Michael Metcalf annouced that 'Fortran 2003 has passed its ballot with flying colours: 20 yeses, 0 noes, 8 abstains.' Strictly speaking, the 2003 and past standards are not freely available but drafts can be found online. FORTRAN 2003 is an upwardly-compatible extension of the current standard, FORTRAN 95, adding and extending support for exception handling, object-oriented programming, and improved interoperability with the C language. In other FORTRAN news, the GNU FORTRAN 95 compiler has made amazing progress over the past year. Gfortran will be part of gcc-4.0 when released (probably in 2005)."

Comments

Have they addressed any of the weirdnesses?

[gowen]

Three things bug me about F95, which I use every day

1. No standardised/portable method for handling command line parameters.
   
2. Undefined behaviour for short-cutting logical ors. e.g. the behaviour of
   if(flag.and.function(var)) then...
   is undefined if flag is false and function() has side effects.
   
3. The things you *can't* use parameters for (like fixing lengths in format statements) without running fpp/cpp on the code first
   
4. No standardised meanings for

   real(kind=8) x

   Does that mean an 8 byte real? Or a 8 bit real? It depends on the compiler... (and yes, I know the portable solution is

   real (kind=kind(0.0d0)) x

   and the such like, but *thats* really ugly, compared to

   double precision x.

Re:Have they addressed any of the weirdnesses?

[dubious9]

I find Fortran (especially old fortran like 77) is most easily worked when you do the setup and application logic in C/C++. Once you've got the type conversions down (ie, when passing Fortran strings, their lenth is appended as an argument, or that array subscripts are reverse and start from 1), it's quite easy to call C from Fortran and Fortran from C.

Since C is much more standard and cross platfrom, you can get that stuff out of the way and do the heavy lifting in Fortran. Data types, yes, those can be inconvenient cross-platfrom wise, (though I've never had to worry about them much on HPUX or Sun) but otherwise use the right tool for the job.

Deafening silence

[bonniot]

Halh an hour after Fortran 2003 was announced on slashdot, the silence is deafening. Have most people migrated to other languages? I often heard that the amount of legacy code will make fortran survive for a long time. Or is it just that the sets of fortran users and of slashdoters do not intersect?

Re:Deafening silence

[kawika]

Sorry it took me a while to respond. I was down in the computer room re-punching my card decks to comply with the new Fortran 2003 standard.

Re:Deafening silence

[traveyes]

You joke.... Seriously, where I work there's 2 people that still write in Fortran/66. No kidding.

They don't use card-punchers, but they might as well.

.

What future programming languages will be

[tod_miller]

Taking aside the idea of syntax for moment, this is a RT on 'what is a language' nothing to do with turin complete, OO or AOP or whatever.

What we really care about: libraries. Being able to do things quickly, without fsking about.

That is why php is successful, people can just run phpnuke/postnuke etc.

Perl is also successful because of its roots and flexibility, and easy to get into, and you could just run slashdot on your site if needs be.

What really helps these, is foundations. php/mysql, perl/whatdoesslashdotuse? People will write in anything if they see an easy way to get something done. Tutorials and support material.

The point - except for people studying 'computer languages' (as someone woudl study the history of world languages) who will pick up fortran as an option for a new language?

I am first to admit I do not know the dissadvantages or advantages of it. Are there any? or is it just syntax?

Java is a language, but much more, it abstracts the whole idea of a language ( no it isn't correct to cite /nets multiple compilers here, no no really this is a different point) and makes it a process.

So it isn't about the clean syntax OO language, but the process of programming. Through design and development and testing, it has all be rbought up with testing, this is true of almost all languages, but when I think of Java I have a view of all the testing frameworks, libraries and standards.

To be honest, Fortran now is just a syntax specification, that says, take this line, and make that byte code. That doesn't do it for me.

One mans syntax is another mans syntax error.

Error on line 1: Insert ; to complete statement

Re:What future programming languages will be

[TeknoHog]

I am first to admit I do not know the dissadvantages or advantages of it. Are there any? or is it just syntax?

Disclaimer: I have an M.Sci in physics with a slant towards computational physics. I'm most familiar with F90 but its advantages should apply to the later revisions as well.

Fortran is a slightly higher-level language than C. It has lots of math capabilities built-in; particularly matrix types and operations meaning good potential for parallel processing. Decent compilers can produce multiprocessor and MMX/SSE code from Fortran, and probably the equivalent on non-x86 platforms. I'm not sure if you can do this in C portably.

Fortran has a fairly simple overall syntax where newlines matter. I found it very easy to learn, having some experience with Python and C. In some ways the syntax is rather arcane but at least it's quite clear (compared to C or Perl at least IMHO).

Fortran does have pointers, but they are unnecessary because of other memory handling methods (namespaces or 'modules'). Dynamic memory allocation is a bit quirky IMHO, but way simpler than in C.

The language feels well suited for physicists. It's not necessarily good for general programming, but I'd happily use it for many speed-critical sections instead of C.

One important reason for the persistence of Fortran is the number of libraries for computational sciences, for example LAPACK. They can be included in modern Fortran programs despite being written in F77. There have been a number of attempts to make a scripting front-end for these libraries; Matlab is one of them, even though its F77 roots are not very visible.

The common complaints about Fortran are mostly true for F77 and older revisions. F77 is a truly horrible language, and it has none of the nice features that make F90+ good for scientific computing.

Re:What future programming languages will be

[gowen]

Dynamic memory allocation is a bit quirky IMHO, but way simpler than in C

Quirky? I really like it.

i) Declare a variable allocatable.
ii) Allocate it, and check IFAIL
iii) Deallocate it when you're done.
iv) Profit!

The real paradigm shift you need for a C programmer moving to fortran, is that nothing is passed by value and side effects are used so much. But the intent(in) and intent(out) directives are your friends for that sort of thing.

Function overloading is pretty straightforward too, once you start using modules with well defined interfaces.

Re:In other news

[dubious9]

Then I challenge you do some complex multi-dimentional array calculation and manipulation. Most applications, yes, but why would scientists still use it if *everything* was better in c++? Not everything is maintaining applications from the 70's; there *is* new development.

There's a language for every problem domain and when that problem domain includes heavy lifting with math, it's often fortran. Besides you can quite easily compile C and Fortan together, Fortran for the math, C for the application data and GUI.

GFORTRAN != G95

[Anonymous Coward]

The article talks about GFORTRAN being released with the next GCC, but then links to G95 on sourceforge. From reading the gcc.gnu.org list trsffic, G95 was orginally on sourceforge, but then forked when its main developer wouldn't cooperate with anybody else, especially the GCC people. The fork was GFORTRAN, and it will be GFORTRAN, not the g95.sf.net project, which gets released.

Re:In other news

[noselasd]

It seems you think of Fortran as a general purpose language, such as e.g. C and C++ it really is not. Fortran is strong at numerical computing, and in many areas there it is way better than C++.

Re:In other news

[bhima]

C & Fortran represents far more code where I work than C++ (or any other single language solution) and there is not single line of java in production code (we use a lot of Perl in house come to think of it).

I think the main reason is that Fortran can describe the problem space in ways that scientists can use well (and hence is maintainable) and C handles the rest fine.

I don't understand why some many younger people are so quick to put older languages down, my sister has made a small fortune maintaining / extending COBOL apps for various financial institutions using more or less the same method (C & COBOL together). I guess it's because they've never understood the various problem spaces to begin with and see the world in a web-centric sort of way.

What sectors are people using this in?

[SLiK812]

When I was a young lad in college (all of 10 years ago), I had to learn Fortran for one of my Chem Eng classes. We were learning Fortran77, mostly cause my profressor didn't think we would need it in the future, and didn't want us to be concerned with some of the new structures in more current versions.

The Aero's also had to learn it (I know cause I taught it to them, since their prof sucked). So what sectors of industry are people working in with Fortran? Is it still just the Chemical and Aeronautical fields, are other places (where a different language might be more beneficial, say) still using it, cause no one wanted to convert systems?

Re:What sectors are people using this in?

[j_cavera]

The nuclear engineering communitiy is (still) standardized on F77. The reasons: 1) many great neutronics codes were written in the 60's (often for weapons research) using F77 and are still valuable today and 2) the NRC has some strange requirements concerning foward and backward compatability - almost anything from a DEC PDP-3 to a new G5 iMac can compile F77 code. Maybe someday they'll let us use F95...

Re:What sectors are people using this in?

[Anonymous Coward]

Writing F77 code in the 60's????

Libraries and support tools

[Per Abrahamsen]

You gave the answer. For scientific computing, Fortran is where the libraries and support tools are.

You can think of it as being for scientific computing what php is for web development. Except, of course, that Fortran is far more dominating in the scientific computing field than PHP is for web development.

I know, I write scientific computing in C++, and I must occationally consider if it wouldn't have been more cost efficient to use Fortran, where most of the relevant libraries, support tools and expertise is.

Stop...the...nightmare

[b-baggins]

Twisted visions of late nights hunched over the green glow of my IBM XT sweating through trying to write a Fortran 77 program to data crunch my analytical Chem class lab reports are running through my brain like nails on a blackboard.

Holding out.

[base3]

I'm waiting for FORTRAN XP.

MODERATION MADNESS STRIKES AGAIN

[hummassa]

Moderators: the grandparent was not flamebait, I was serious.

Ok, I will start over:

dubious9: I challenge you do some complex multi-dimentional array calculation and manipulation

You mean like:

Matrix a, b, c, d;
a = b * b / c + d.dot(a);

unh? easy answer: use the right lib.

dubious9: why would scientists still use it if *everything* was better in c++? Not everything is maintaining applications from the 70's; there *is* new development

I don't know!? Because they don't know C++? Because they would have to put a C++ course in Engineering School? Because they don't know the whole awful lot of ultra efficient, STL-based, highly-parallelized numeric/scientific computation libraries available for C++? (an initial search of freshmeat returns at least seven interesting, relevant results)

bhima: see the world in a web-centric sort of way

what does a web-centric way has to do with c++?

noselasd: Be very much aware that there are lots which measure "efficient" in development time, NOT execution time.

See below :-)

tigersh: That would hardly make Fortran efficient, but then in comparison to C++

I am really more comfortable -- and efficient -- with c++'s features and quirks than with fortran's. But as I answered to dubious9, this is not what I said in the grandparent post. It was: Everything that can be done in FORTRAN can be done more efficiently and just as expressively in c++. So, it kind of lost its raison d'etre. And I stand for it. I can do things like the matrix manipulation of the first question in c++ and have the compiler parallelize what it can, use SSE or whatchmacallit, using templates (and more on this in the last answer) and operator overloading.

noselasd: It seems you think of Fortran as a general purpose language, such as e.g. C and C++ it really is not. Fortran is strong at numerical computing, and in many areas there it is way better than C++

No, it is not. That was my point to begin with.

gowen: Really? How do you get around the fact that the flexibility of C/C++ pointers make it almost impossible to completely optimise/parallelise code is a safe manner?

Answer: Partial Template Specialization. It works.

Fortran lives in scientific research.

[CharAznable]

Back in college, a couple of years ago, I worked as a research assitant for an astronomer. He made me learn Fortran77, and then Fortran 90. F77 is horribly stiff, no different from punching stuff into a card. But F90 was really great, easy, straight forward syntax, and very, very fast. My job was basically working with 15 year old f77 code and writing new f90 code used to calculate IR emissions from protoplanetary disks around stars. I leared a lot and made me a better programmer, I think.