3 ms·
While this looks like great stuff and there is no way I will discourage anyone from developing tools for scientific computing in any programming language, the v
by pm90 10y ago
While this looks like great stuff and there is no way I will discourage anyone from developing tools for scientific computing in any programming language, the vast majority of users of these tools IMO are scientists and engineers who are not programmers. Or to put it another way, most of them don't have a degree in CS, they are usually Mathematicians, Physicists etc. When I hung out in that community about 3 years ago, there was already a wholesale movement going on in moving from MATLAB and/or Fortran to Python... and I don't think the vast majority of that community (which is generally conservative when it comes to tooling) will adopt yet another tool unless it offers some significant benefit. (MATLAB was easy to write and use but proprietary. Python was easier and free. Both make use of high performance libraries written in C/Fortran).
- gautamdivgi 10y agoYup. I'm one of those people. Well, partially at least. I do a lot of distributed computing. But my PhD was in a statistical area. I used Octave for a while (matlab clone). But then as my data sets got larger and I mean really large (non-linear equation solving on 11m points on my puny desktop), I went full force on Python (NumPy, SciPy, Matplotlib). The reason - its is super easily scriptable, the libraries are great and easy to use and python allows me to use multi-processor capabilities to distribute my work to all my cpu cores. Keep in mind I could have used GSL. But Numpy/Scipy & matplotlib were just faster and more productive for me. That being said, I'd rather have one source of truth - i.e. GSL and have this be a wrapper around it as opposed to re-creating all the logic. Either way - its a fantastic effort. If at anytime I create systems that need numerical computation (as opposed to the hacks that I did for my thesis) I will definitely look into this.