When choose Julia

As we priviously mentioned, Julia is not magic and there is no silver bullet, we have to choose our language for certain problems, here are some suggestions:

When you need Julia

• if you usually need to optimize Python with C/C++/Fortran/etc.

• if your program reqiures a lot loops and matrix multiplications and other computation heavy tasks

• if your problem requires a lot abstractions but also sensitive to performance.

• if your interface will looks better with a DSL, or some addons on its original language. (like einsum for tensor contraction)

• Your collaborator or supervisor use Julia.

When you don't need Julia

• Your program is heavily based on pure Python and its packages (like Pytorch, MXNet, TensorFlow, etc.), and there is no better or equivalent Julia packages.

• You care about performance, but most of your functions and commands will only be executed few times. (because we have a JIT overhead in Julia)

• You don't care performance very much at the moment, about 5x slower is acceptable. (tasks like GUI, plotting, etc.)

• Your collaborator or supervisor do not accept new language.