28.4 % admired Fortran in StackOverflow survey

Meta
1

StackOverflow
StackOverflow1802Ă—973 128 KB

Technology | 2024 Stack Overflow Developer Survey

4 Likes
2

Nice, Fortran made the page. The absolute numbers are pretty good I would say, but even more important are the relative numbers. I was able to find the 2023 survey here: Stack Overflow Developer Survey 2023, Fortran has the following admired-desired percentages and rank (I manually counted the position, hopefully correctly):

  • 2023: 0.59% - 24.37%, 45th place
  • 2024: 0.9% - 28.4%, 44th place

So the percentages show clear improvement. The admired (“the proportion of users that have used the same technology in the past year and want to continue using it”) increased by over 50% year to year. So I think this is good, if the trend continues, we are in good shape.

6 Likes
3

This is a key point that I think is sometimes undervalued when we discuss language improvements. It’s not enough that we focus on adding features while rejecting aesthetic improvements. We need to make the language easy, or even an pleasure, to use; otherwise people will not use it. So when we see complaints about aspects of Fortran that are overly verbose or require multiple statements to do something that can be done with a single statement in another language, they are worth serious consideration.

3 Likes
4

Yes, and I’d say the same for the community and the quality of discussions. A language with unfriendly or obscure/closed communities attract fewer new users. I think this has significantly improved in the past 5 years.

5 Likes
5

Slightly off topic, but related to the popularity and future of Fortran in general…

Did you know Python has largely replaced Fortran for HPC applications because Python performs really well due to its ability to interface with Fortran. :grin:

image
image676Ă—522 36.4 KB

6

I often see python + some fast libraries have been used rather often for computationally “heavy” calculations. For example, a recent one (for climate modeling, by Google Research, published recently in Nature) is:

https://research.google/blog/fast-accurate-climate-modeling-with-neuralgcm/

I remember the Boltzmann generator (a kind of canonical sampling method for molecular/protein configurations) also used Python (together with ML libraries):

Another example may be FermiNet,

7

Apart from Python, RE the SO survey, I always wonder to what extent the percentage of that survey is actually meaningful, because it does not give any information about the correlation with actual users. As for me, I’ve never participated in that survey, and I wonder how many Fortran users actually voted in the survey (even for active SO users).

PS. Not trying to “critisize” the OP’s post at all, but just my impression of the SO survey overall…

8

I’ll add one of the more impressive Python based codes from the CFD/turbulence modeling world, PyFR.

https://www.pyfr.org/

https://www.sciencedirect.com/science/article/pii/S0021999116307136

https://www.computer.org/csdl/magazine/cs/5555/01/09430703/1tzui7afJ6g

1 Like
9

A programming language with snowflake-free communities that welcome and withstand fierce criticisms of the language and work hard and positively to do more and faster become worthy of its users.

2 Likes
10

I couldn’t agree more!

11

There are also some FDM and spectral-based solvers in Python:

Also several powerful Python FEM frameworks, such as FEniCs and FireDrake:

The UFM language these are based on isn’t too far from Fortran.

One more recent one, again using a Python code generator that outputs C++ (or even SIMD-heavy assembly):

The authors went all the way to 10^12 degrees of freedom.

1 Like
12

Filtering the Python Package Index by programming language gives

C 1763
C++ 1941
Cython 1214
Fortran 220
Java 160
Javascript 2331
Python 10000+
Rust 1822
SQL 334
Unix Shell 388

Other languages are used in fewer packages.

Searching for Python in my list of Fortran codes on GitHub gives 129 hits.

1 Like