4 ms·

Interesting. You might be the first computational chemist I know who actually uses Rust. I know a lot of computational chemists! Python is the big one, all of

by akuro 4y ago

Interesting. You might be the first computational chemist I know who actually uses Rust. I know a lot of computational chemists! Python is the big one, all of the aforementioned chemists are either intermediate or advanced in that. The runner-up seems to be Julia, which I personally have no experience with. The big guys are Fortran and C++. I prefer C for tasks of this nature, but I also shill Scheme so don't listen to my opinions on programming languages. Best of luck on your computational chemistry endeavours!

doubleg72 4y ago
[flagged]
- akuro 4y ago
  Don't listen to my opinions, sure. But pay heed to the large number of computational chemists who do use these languages. There is a reason that professionals employ the tools that they do.
  - freehorse 4y ago
    Historical reasons, most often than not.
  - tialaramex 4y ago
    One reason we have lots of people in numerate disciplines using Python is that we taught them Python. Python is very teachable. Given a class of average 19 year old students from a numerate discipline, my colleagues will teach most of them Python to a reasonable standard in a single module (so e.g 2 hours of teaching per week over 16 weeks). The same would not be true if we were supposed to teach them C++ for example. It's a huge, sprawling language, full of difficult concepts that are unfriendly to teach, and equally filled with foot guns that if you don't teach them will definitely maim your students, but if you do teach them take up yet more precious time on minutiae. Safe Rust wouldn't be as hard to teach as C++ but it's no picnic. So even if the Chemists decided that ideally they'd like their undergraduates to learn Rust instead of Python, I think the argument would be that it can't be done on the existing timeline.
    - zozbot234 4y ago
      > Python is very teachable. Given a class of average 19 year old students from a numerate discipline, my colleagues will teach most of them Python to a reasonable standard in a single module Compared to Perl, sure, Python is teachable. But I'm willing to speculate that Safe Rust as a semi-pure functional language (basically: when in doubt, .clone() all the things and don't even think about the minor hit to performance) can also be taught as a first-time programming language. Rust may be at a disadvantage to Python wrt. bindings to the existing ecosystem of C/C++/F0RTRAN scientific codes, but that's a temporary state of things.
      - aragilar 4y ago
        To science majors? No. It lacks so many of the things you'd need to get them started (e.g. a REPL/notebook server, interactive plotting, a "here's everything you need" install (a.k.a conda)), and most (but not all) students have no real interest in learning to program. You could get the interested-in-programming ones to learn Rust, but then they're going to hit the large legacy of existing code and wonder why you taught them something that not related to what they're going to use.
        zozbot234 4y ago
        There's an interesting experimental project for a Rust REPL and notebook interface at https://github.com/google/evcxr https://github.com/google/evcxr . Other things, e.g. semi-officially endorsed collections of community crates with no strong backward compatibility guarantees, have seen some development already. > but then they're going to hit the large legacy of existing code and wonder why you taught them something that not related to what they're going to use. OTOH other languages should be easier to learn by reference to a comparatively elegant language like Rust. Python, Fortran, C++ etc. have almost nothing in common but Rust could arguably be a good introductory baseline that shares features with all of these.
    - aragilar 4y ago
      Now we teach them Python, because the ecosystem and mindshare is there. But why didn't Perl or Ruby (or Tcl) win? It seems like Ruby didn't get the right ecosystem around it, but Perl has PDL (which does seem to predate much of the Python scientific ecosystem), and had SWIG support before Python. Maybe there's a lesson there for the Rust scientific ecosystem?
- osigurdson 4y ago
  You take the comment far too literally.
abirch 4y ago
C++ and Fortran are what come to mind for computational chemistry due to the legacy libraries but it's been a long long time since I've used Gaussian.
- rsfern 4y ago
  i think there are only a couple exceptions to the Fortran/C++ rule. DFTK.jl is the only one I know written in Julia, and there’s GPAW and pySCF in python
boppo1 4y ago
What's discrediting about scheme?
- akuro 4y ago
  I don't know, but whenever I plead with people in computational science to read Structure and Interpretation of Classical Mechanics I'm usually met with an eye-roll.
  - boppo1 4y ago
    Interesting, I'd never heard of SICM, only SICP.

by akuro 4y ago

doubleg72 4y ago

[flagged]

akuro 4y ago
Don't listen to my opinions, sure. But pay heed to the large number of computational chemists who do use these languages. There is a reason that professionals employ the tools that they do.
- freehorse 4y ago
  Historical reasons, most often than not.
- tialaramex 4y ago
  One reason we have lots of people in numerate disciplines using Python is that we taught them Python. Python is very teachable. Given a class of average 19 year old students from a numerate discipline, my colleagues will teach most of them Python to a reasonable standard in a single module (so e.g 2 hours of teaching per week over 16 weeks). The same would not be true if we were supposed to teach them C++ for example. It's a huge, sprawling language, full of difficult concepts that are unfriendly to teach, and equally filled with foot guns that if you don't teach them will definitely maim your students, but if you do teach them take up yet more precious time on minutiae. Safe Rust wouldn't be as hard to teach as C++ but it's no picnic. So even if the Chemists decided that ideally they'd like their undergraduates to learn Rust instead of Python, I think the argument would be that it can't be done on the existing timeline.
  - zozbot234 4y ago
    > Python is very teachable. Given a class of average 19 year old students from a numerate discipline, my colleagues will teach most of them Python to a reasonable standard in a single module Compared to Perl, sure, Python is teachable. But I'm willing to speculate that Safe Rust as a semi-pure functional language (basically: when in doubt, .clone() all the things and don't even think about the minor hit to performance) can also be taught as a first-time programming language. Rust may be at a disadvantage to Python wrt. bindings to the existing ecosystem of C/C++/F0RTRAN scientific codes, but that's a temporary state of things.
    - aragilar 4y ago
      To science majors? No. It lacks so many of the things you'd need to get them started (e.g. a REPL/notebook server, interactive plotting, a "here's everything you need" install (a.k.a conda)), and most (but not all) students have no real interest in learning to program. You could get the interested-in-programming ones to learn Rust, but then they're going to hit the large legacy of existing code and wonder why you taught them something that not related to what they're going to use.
      - zozbot234 4y ago
        There's an interesting experimental project for a Rust REPL and notebook interface at https://github.com/google/evcxr https://github.com/google/evcxr . Other things, e.g. semi-officially endorsed collections of community crates with no strong backward compatibility guarantees, have seen some development already. > but then they're going to hit the large legacy of existing code and wonder why you taught them something that not related to what they're going to use. OTOH other languages should be easier to learn by reference to a comparatively elegant language like Rust. Python, Fortran, C++ etc. have almost nothing in common but Rust could arguably be a good introductory baseline that shares features with all of these.
  - aragilar 4y ago
    Now we teach them Python, because the ecosystem and mindshare is there. But why didn't Perl or Ruby (or Tcl) win? It seems like Ruby didn't get the right ecosystem around it, but Perl has PDL (which does seem to predate much of the Python scientific ecosystem), and had SWIG support before Python. Maybe there's a lesson there for the Rust scientific ecosystem?
osigurdson 4y ago
You take the comment far too literally.

abirch 4y ago

C++ and Fortran are what come to mind for computational chemistry due to the legacy libraries but it's been a long long time since I've used Gaussian.

rsfern 4y ago
i think there are only a couple exceptions to the Fortran/C++ rule. DFTK.jl is the only one I know written in Julia, and there’s GPAW and pySCF in python

boppo1 4y ago

What's discrediting about scheme?

akuro 4y ago
I don't know, but whenever I plead with people in computational science to read Structure and Interpretation of Classical Mechanics I'm usually met with an eye-roll.
- boppo1 4y ago
  Interesting, I'd never heard of SICM, only SICP.