> any evidence of their still strong position will be greatly appreciated
Fortran still dominates certain areas of scientific computing / HPC, primarily computational chemistry and CFD. - https://fortran-lang.org/packages/scientific - you don't hear about most of them because they're generally run on HPC centers by scientists in niche fields. But you do get their benefit if you buy things that have the chemical sector in their supply chain.
The common thread is generally historical codes with a lot of matrix math. Fortran has some pretty great syntax for arrays and their operations. And the for-loop parallelization syntax in parallel compilers (like openmp) is also easy to use. The language can even enforce function purity for you, which removes some of the foot guns from parallel code that you get in other languages.
The kinds of problems those packages solve tend to bottleneck at matrix math, so it's not surprising a language that is very ergonomic for vector math found use there.
Same for Matlab, it's mostly used by niche fields and engineers who work on physical objects (chemical, mechanical, etc). Their marketing strategy is to give discounts to universities to encourage classes that use them. Like Fortran, it has good syntax for matrix operations. Plus it has a legitimately strong standard library. Great for students who aren't programmers and who don't want to be programmers. They then only know this language and ask their future employer for a license. If you don't interact with a lot of engineers at many companies, you aren't going to see Matlab.
I work for the National Solar Observatory, creating Level 2 data for the DKIST Telescope’s observations of the sun. (For example, an image of the Temperature that lines up with the observation)
Just the other day, the solar physicist I work with said “yeah, that code that runs on the supercomputer needs to be rewritten in Fortran” (from C, I think.
He’s nearing retirement, but it’s not that he’s behind the times. He knows his stuff, and has a lot of software experience in addition to physics
While it's not as ubiquitous as it used to be, Matlab is still very heavily used within shops that do a lot of traditional engineering (ME, EE, Aero, etc.).
This surprises people who just write software, but consider:
- The documentation and support is superb. This alone can justify the cost for many orgs.
- Did I mention the support? MathWorks has teams of application support engineers to help you use the tool. The ability to pay to get someone on the phone can also justify the price.
- The toolkits tend to do what specific fields want, and they tend to have a decent api. In contrast, you might end up hacking together something out of scipy and 3-4 weirdly incompatible data science packages. That's fine for me, but your average ME/EE would rather just have a tool that does what they need.
- It also has SIMULINK, which can help engineers get from spec to simulation very quickly, and seems deeply embedded in certain areas (eg: it's pretty common for a job ad with control systems work to want SIMULINK experience).
Is Python gradually displacing it? Probably.
(Honestly, I wish it would happen faster. I've written one large program in Matlab, and I have absolutely no desire to write another.)
Fortran still dominates certain areas of scientific computing / HPC, primarily computational chemistry and CFD. - https://fortran-lang.org/packages/scientific - you don't hear about most of them because they're generally run on HPC centers by scientists in niche fields. But you do get their benefit if you buy things that have the chemical sector in their supply chain.
The common thread is generally historical codes with a lot of matrix math. Fortran has some pretty great syntax for arrays and their operations. And the for-loop parallelization syntax in parallel compilers (like openmp) is also easy to use. The language can even enforce function purity for you, which removes some of the foot guns from parallel code that you get in other languages.
The kinds of problems those packages solve tend to bottleneck at matrix math, so it's not surprising a language that is very ergonomic for vector math found use there.
Same for Matlab, it's mostly used by niche fields and engineers who work on physical objects (chemical, mechanical, etc). Their marketing strategy is to give discounts to universities to encourage classes that use them. Like Fortran, it has good syntax for matrix operations. Plus it has a legitimately strong standard library. Great for students who aren't programmers and who don't want to be programmers. They then only know this language and ask their future employer for a license. If you don't interact with a lot of engineers at many companies, you aren't going to see Matlab.