Even if the TIOBE index were a meaningful indicator, Fortran being in a tie with Delphi Pascal for tenth place shouldn’t really be a cause for celebration for its few remaining enthusiasts.
Fortran should be compared with its competition, which, in my opinion, is only MATLAB in the top 20 TIOBE list. All others are either non-scientific or primarily used for non-scientific tasks. Fortran won’t likely go much higher than it has achieved in ranking because that is almost the saturation level of its community. Less than 2% of the world’s population has a (STEM) graduate degree, the primary Fortran userbase, which matches TIOBE’s 1.24% Rating of Fortran. This is why MATLAB’s TIOBE ratings have stagnated for over a decade and never surpassed ~2%.
This comment was written by the lead developer of Flang. Peter, it sounds like you have given up on Fortran. Is Fortran saved? No. Is it still falling into obscurity? I am not sure, but the fall might have stopped. We still have a long way to go, either way. However, this is definitely a small win, it’s the highest ranking Fortran has ever got at TIOBE. So certainly a cause for a celebration!
The index is currently limited to 29 languages. You can still analyze the popularity of your favorite language and compare it to others, using Google Trends. C++ has the same popularity as C on Google trends: to avoid duplication, it is not included in the PYPL index.
The position of Fortran related to other languages below 2% rating is not interesting in itself as there is much noise. But the TIOBE Index is measuring something (although we don’t know clearly what) and that something is clearly rising Fortran into the spotlight (of the Top 20). Whatever we think of that ranking, it is good marketing for the language. The important thing is that people now ask “what is going on?” and search information about Fortran to try to find an explanation.
C is Python’s evil twin, or, in chronological order, Python is C’s good-natured twin. C gives you performance, and doesn’t care about usability or safety; if any of the footguns go off, tell it to your healthcare provider, C isn’t interested. Python on the other hand gives you safety, and it’s based on a decade’s worth of research into usability for beginners. It doesn’t, however, care about performance. They’re both optimized aggressively for two opposite goals, at the cost of ignoring the other’s goals.
But on top of that, Python was built with C extensions in mind from the start. Today, from my point of view, Python functions as a packaging system for popular C/C++ libraries.
Ideally modern Fortran gives you performance, usability, and safety.
Searching “fortran” in the journal Physical Review gives 3657 hits and indicates that it has been used extensively in all domains of physics. Below are the numbers of hits by field. A paper containing the string “fortran” may not have underlying Fortran code but may be referring to previous works that do.
3-dimensional systems (62)
Atomic & molecular collisions (44)
Atomic & molecular processes in external fields (73)
I, Fortradamus , prophesies that in the year of our Lord 2026, Fortran will be available as a scripting language in MsOffice and Libre Office. In 2028, it will be included in Android to reduce the energy consumption. Yes, the Great Compiler will come back. Let it be.
At the recent Platform for Advanced Scientific Computing (PASC) Conference, @rouson was asked in an interview why agencies should keep funding Fortran development. Here is my manual transcription of his answer.
The number one thing I always say is that Fortran is the only widely-used language that has a standards body – a standards committee – that sees scientists as their target audience. If you’re working in most other languages, certainly any of the really popular languages – C, C++, Java, Python – certainly there is a lot of great stuff happening in science with those other languages, but the science is a very thin slice of what those languages are doing, and so it can sometimes take a very long time to get features into those languages that are really supporting science.
The program of the Fortran minisymposium at the conference is here.
The above quoted statement is inaccurate, arguably it’s a falsehood and worse. Whoever makes such a statement will struggle mightily to show any evidence where the “standards body” has taken any action based on a line of sight of “scientists as their target audience”.
I would welcome if this were really the case.
A fundamental issue with the functioning of the standards body, since it has taken on the mantle of language development as part of standardization is …
For whom Fortran, and for what?
The actions by the standards body are DECIDEDLY unclear on this, one can make a far better argument the standards body is merely attempting to …
“Keep the lights on”
“Manage the decline”
Seek some organic growth with some processor sales of vendors,
Adhere to the terms of big taxpayer body contracts such as with DOE or other US-based national labs and bodies …
P.S.>
When someone calls things exactly as they are, the unvarnished facts, all the moderators do is delete my post.
There is nothing offensive to anyone here, the only code of conduct violation here will be when my right to express my thoughts are suppression via the shameful and cowardly deletion of this post.
I would say it is at least arguably accurate. If you extend the audience a little to include engineers and mathematicians (i.e. numerical analysis), then it would be more correct. Certainly one would say that the fortran committee does not bend toward operating systems, computer graphics, or data mining, which are the domains of some of those other languages mentioned. I think fortran is not doing so poorly now in these areas, but certainly in the 1980s and 1990s I would have preferred for the fortran standard to be more responsive to those areas (e.g. the POSIX standard in 1988). For example, floating point arithmetic is mostly used in the domain of science and engineering, and there isn’t a language that supports IEEE 754 better than modern fortran.
Numerical computing is a small fraction of the needs of scientists in scientific programming.
The “standards body” has done little to nothing that would suggest a direct line of sight of “scientists as a target audience” in any of their actions.
That the few language feature additions every five years overlap in some ways with what scientists too can make use of does not amount to “see scientists as a target audience”.
Very impressive. We have achieved another milestone!
@FortranFan, @rouson, I think it is an aspirational quote about the goal of the committee. I think we can all agree that should be the goal. What we can also agree is that in the top 20 languages in TIOBE, it’s only Fortran and MATLAB that has science as the target audience. @RonShepard I think the committee’s actions treat certain compiler vendors as the target audience, and since the compilers are for a language used by scientists, there is a large overlap. But there are also differences, for example compiler vendors are often reluctant to prototype new features, and that’s something the committee and compiler vendors should improve upon. Having more users of Fortran at the committee would help a lot. If you are a user, please join the committee.
@FortranFan I’ve been saying something like that since long before I joined the Fortran committee, but I’m coming to understand the extent to which I’m in a tiny minority of people who have long been happy with the decisions of the committee. Even when I invested quite a lot in developing and presenting an idea that the committee shot down, I ultimately agreed the committee was right. LOL.
With all that said, there appears to be so much widespread disappointment with committee, including on the part of those on the committee, that I think I should revise my statement and remove the committee from it so that the statement is just about the language. As I’m sure you know, Fortran has math right in the origins of its name: Formula Translation. So hopefully what is inarguable is that Fortran aims to support mathematical work and that describes a lot of science. I suspect this is why the language has multidimensional arrays in addition to so many array features plus all the math intrinsics and math-oriented feature sets like user-defined operators. I also suspect the emphasis on high-performance computing, which arguably relates to an emphasis on scientific computing, also shows that the language aims to support science. That’s why there’s support for distributed-memory parallel programming and support for things like GPU programming via do concurrent, which three compilers can now automatically offload.
Is it more neutral and less controversial if I describe the language instead of the committee? I’ve been thinking of making that change for a long time anyway. It was fine to praise the committee’s work before I joined, but not a great idea to praise the committee’s work if I’m on it. I apologize for bad form.
There is an interview with Steve Lionel from 2012 on the RCE Podcast (RCE 74: Modern Fortran | Podcast | rce) in which scientists and engineers are mentioned as the primary target user group for Fortran.
I hope we all agree that Python plays a gigantic role in scientific computing today. If I think about language changes that have impacted me the most, they are
There is a lot more stuff going on with standard libraries in the Python Enhancement Proposals. The large volume of third-party libraries is also very attractive for scientists. The solutions for code documentation are much better in Python, Julia, and Go, and sometimes I have the feeling this plays a bigger role than raw performance.
I just notice Tiobe’s readings about July 2024 are in. Regarding Fortran, the report is a new all-time high, both in percentages (2.05%) and rang (#9); the first streak to be among the top-ten for four months
, prophesies that in the year of our Lord 2026, Fortran will be available as a scripting language in MsOffice and Libre Office. In 2028, it will be included in Android to reduce the energy consumption. Yes, the Great Compiler will come back. Let it be. 
