3 ms·

Julia is too different from Fortran for the old wizards to make the switch.

catach 4y ago
Seems that most of the time, to be significantly better a language also has to be significantly different.
leephillips 4y ago
I’m not a wizard but I’m pretty old. After years of scientific programming in Fortran Julia was very easy to switch to. This is not like switching to a (conventional) Lisp or APL, The syntax is quite familiar; you need not use any part of the type system, until you learn more and want to take advantage of it. You’ll need to sprinkle in a few dots to reproduce Fortran‘s array expressions, in places.
- Moissanite 4y ago
  How do you get on with package management? That is probably the area where Julia had the most opportunity to trounce Fortran, and yet that is the part of the language I've never managed to get comfortable with.
  - leephillips 4y ago
    The package manager is one of my favorite things about Julia. From its integration with Git to the ability to effortlessly reproduce the precise environment in which any program was run, it’s programming language package management, finally, done right: https://lwn.net/Articles/871490/ https://lwn.net/Articles/871490/
    - Moissanite 4y ago
      Thanks - you inspired me to find the right search term for my issue ("CI/CD in Julia"). Basically, the idea of using the REPL (at all, and particularly for package management) never sat well with me; I was looking for the equivalent of "requirements.txt".
      - ChrisRackauckas 4y ago
        Project.toml or Manifest.toml? Every package has Project.toml which specifies bounds (https://github.com/SciML/OrdinaryDiffEq.jl/blob/master/Project.toml https://github.com/SciML/OrdinaryDiffEq.jl/blob/master/Proje...). Every fully reproducible project has a Manifest that decrease the complete package state (https://github.com/SciML/SciMLBenchmarks.jl/blob/master/benchmarks/Bio/Manifest.toml https://github.com/SciML/SciMLBenchmarks.jl/blob/master/benc...).

Seems that most of the time, to be significantly better a language also has to be significantly different.

I’m not a wizard but I’m pretty old. After years of scientific programming in Fortran Julia was very easy to switch to. This is not like switching to a (conventional) Lisp or APL, The syntax is quite familiar; you need not use any part of the type system, until you learn more and want to take advantage of it. You’ll need to sprinkle in a few dots to reproduce Fortran‘s array expressions, in places.

Moissanite 4y ago
How do you get on with package management? That is probably the area where Julia had the most opportunity to trounce Fortran, and yet that is the part of the language I've never managed to get comfortable with.
- leephillips 4y ago
  The package manager is one of my favorite things about Julia. From its integration with Git to the ability to effortlessly reproduce the precise environment in which any program was run, it’s programming language package management, finally, done right: https://lwn.net/Articles/871490/ https://lwn.net/Articles/871490/
  - Moissanite 4y ago
    Thanks - you inspired me to find the right search term for my issue ("CI/CD in Julia"). Basically, the idea of using the REPL (at all, and particularly for package management) never sat well with me; I was looking for the equivalent of "requirements.txt".
    - ChrisRackauckas 4y ago
      Project.toml or Manifest.toml? Every package has Project.toml which specifies bounds (https://github.com/SciML/OrdinaryDiffEq.jl/blob/master/Project.toml https://github.com/SciML/OrdinaryDiffEq.jl/blob/master/Proje...). Every fully reproducible project has a Manifest that decrease the complete package state (https://github.com/SciML/SciMLBenchmarks.jl/blob/master/benchmarks/Bio/Manifest.toml https://github.com/SciML/SciMLBenchmarks.jl/blob/master/benc...).