I don’t agree. A key aspect to progress on any front is Sustainability . Outside of this Discourse component, there is no underlying cohesive foundation providing any sustainability to anything else at fortran-lang.org. Based on a convergence of voluntary interests at a particular point in time, it can thrive for a while but eventually voluntary interests have a way of diverging and things tend to wither away.
If Fortran language is to make a meaningful difference to the domains in scientific and technical computing, an ORG somewhat similar to the one lifting up python.org with some formal and more sustainable structure and backing will be essential. Otherwise, Fortran risks missing out completely on so many monumental computational advances that are on the near horizon which so many other languages and paradigms - and there are a dozen or more of them as listed on the IEEE annual survey and promoted in Nature News articles - are all well-positioned to contribute meaningfully.
This is not a knock on anyone who have done or are doing all the hard work at fortran-lang.org, it’s just a matter-of-fact. And there is no implication building a formal ORG and perhaps a foundation behind it is easy; rather it’s hard, extremely, extremely hard. The fact Fortran suffers severely from the “currency of eyeballs” (not enough people have their focus on this language) makes it even harder.
Vipul, you seem to have good ideas–why don’t you join our monthly calls, with suggestions on how to make things better? We need all the help we can get.
According to Google Trends, from 2010 on, of the Fortran 77, 90, 95, 2003, and 2008 standards (you can only view up to 5), Fortran 90 was the most often searched-for version of Fortran. Well into the 2010s, Fortran 77 was second, outpacing Fortran 95. I think Fortran programmers, who are often primarily scientists, engineers, numerical analysts, or statisticians rather than programmers, (and who don’t visit sites like this) are relatively slow to adopt new standards.
Due to no standard file name suffix for Fortran source files, and default behavior varying between compilers, many programmers, including myself, use .f90 for free-form source files. I wonder if and how much has this contributed to the mainstream perception of Fortran. If you randomly browse free-form Fortran source code on the internet, you’re likely to see predominantly .f90 files, even for new projects, Fortran-lang projects included.
I think @Beliavsky got the point here. The programmers, specially the younger generations, go Java/Python/C++ or some of the newest “toys” - Julia, Scala, whatever. The former group (scientists etc.), even if they are not using Python/Numpy/Scipy or C++, will hardly need the most sophisticated features because the effort of learning them would not be worth their time. They are happy with array syntax, derived types (simple, just to be able to construct linked lists or trees).
There is a potential in parallel programming (coarrays) but frankly speaking, the support has been poor until very recent Intel decision to make oneAPI free. GFortran requires external libraries (opencoarrays) to make the executable truly parallel and installing those libs is not an easy task in many OSes, believe me. RHEL/CentOS has abandoned them in the newest versions. Even oneAPI does not have coarrays for all platforms (MacOS for an important example). And the ability of modern compilers to auto-parallelize may be an easier alternative.
In the 90’s, .f90 was sounding modern. Thirty years later, it sounds like old stuff… Modernity is always now…
Something like .ff (like Free Form) would have been better for the long term (and Fortran people already knew what was long term)… But perhaps the .ff extension was already used at that time?
You are of course right, let’s concentrate on showing what Modern Fortran can do (and also let’s first learn ourself what Modern Fortran can do… In my case I need to learn OOP and Coarrays for example…) And let’s improve its ecosystem (awesome but scattered).
I don’t remember .ff. There was (still is?) .F extension used for files which were to be C-preprocessed before compilation. On the C-side, however, .cc and .C have always been (in case-sensitive environments, of course) equivalent, so maybe that was the reason for not adopting .ff
Per the standard’s own terminology, there are 2 source forms: fixed form and free form. So .ff can imply either and still remain confusing.
So I had light-heartedly suggested .mf in a similar discussion thread a few years ago at comp.lang.fortran to imply Modern Fortran, an abbreviation that can remain ever-green yet be somewhat vague and which can be intentional of course: https://groups.google.com/g/comp.lang.fortran/c/7KJnPThsD4A/m/elOXYURqAgAJ
.mf came in 2nd place in a limited, informal polling then.
Nonetheless, as the strong opinions on that CLF thread showed, it’s way too much of an effort to replace.f90 now, too much of the existing tooling and the build chain remain wedded to that convention.
However a quick thought suggests to me different conventions can coexist just as they do with everything in life, it’s the will of the practitioners that can prevail. Resistance to change and conformity can be a bane to progress.
I have seen some projects use this convention. I see no downsides for large projects where you adopt common conventions and have an automatic build system.
For small projects, where I am just compiling a few source files from a terminal window, I like things to just work directly, without the need to memorize additional flags. Of course I could create a wrapper compiler command, but I’m too lazy for that. Hence, I stick to .f90 for free-form.
For the time being, fortran.com will continue, more or less, as it is. Ron Green, who has worked with Walt for many years, will continue to run it, and he’s recruited me to help. It needs some updating, to be sure… If at some point Ron doesn’t want to do it anymore, I and another “friend of Fortran” have an open offer to take it on ourselves.
Maybe fpm can help us out here, setting free-form flags as default and allowing us to go back to .f. It might be an additional, good selling point for its extensive adoption.
The main page of dlang.org lists some prominent companies that use D, with links to more, including testimonials. It would be nice if the fortran-lang.org had something similar. Lahey Fortran has a list of corporate customers, and Absoft has a list and testimonials. NAG has case studies, although I don’t know which ones are for its Fortran compiler and library. I think Intel Fortran has the largest share among commercial compilers but don’t know of similar pages for it. A less-popular language has to deal with the criticism that nobody uses it anymore.
Given that .f90 is the most common suffix for free format Fortran source files, one can google “foo filetype.f90” to find such code containing foo. There is some code online with the .f95 suffix. For fixed source form, .f is by far the most common, with a smattering of .for.
