[pjmlp]

Python will be beaten by those caring about performance for writting applications instead of plain server scripts.

Those will migrate to Julia, Go, or something else.

I used Python a lot in the last decade (2000 - 2004) for automatation and little applications.

Nowadays my choice would be OCaml for the same type of tasks.

EDIT: typo, automatic => automatation

[andreasvc]

Those caring about performance are already using Python, in combination with extension modules written in C/Cython. You would have to complete your argument by explaining why they would rather switch to a completely new language without a huge (scientific) ecosystem of libraries, instead of writing more extension modules in the places where it really matters.

The appeal of a scripting language is that "automatation and little applications" do not take a lot of friction to write, I'd say OCaml is less suited in this regard.

[pjmlp]

As someone that cares about performance I don't use Python any longer, except on cases where customers require me to do so.

I don't need to convince others to join any kind of cause.

[andreasvc]

I don't follow. I am not talking about any kind of cause. However, you made a specific prediction, that "Python will be beaten", and I'm saying you haven't made a proper argument why that is so.

As someone who cares about performance I still use Python for the non-performance critical parts of code, which is the vast majority.

I'm pretty sure that in the end this is just a matter of personal preference, but you make it sound as if there are objective arguments without actually presenting them.

[pjmlp]

My line of thought while replying to the parent is that if a rewrite is needed, those that are more performance minded, are likely to use a language that doesn't suffer from GIL or a pure interpreter as canonical implementation.

Somehow you can already see it on the interwebs. The majority of Go and Julia users come from Python and Ruby, not from languages that have AOT/JIT compilers on their toolchains.[0]

This is just my gut feeling from almost 30 years watching similar technology transitions.

I have no data to back it up and may be completly wrong.

[0] There is PyPy, but I never saw it being deployed in production systems.

[jghf]

Numba + Blaze boosts numerical python to speeds on par with or greater than julia, with additional out of core, delayed expression and multiple backend capabilities that Julia currently does not have.

This might stem the tide for a while... Then we have pyston which maintains C api compatibility.

[pjmlp]

Are they written in Python?

[jghf]

Blaze is, Numba is a JIT compiler to LLVM. You can write numpy functions and JIT/Multithread them with just a decorator and zero to few code changes. Blaze makes this transparent to the user, chunks backends, provide out of core array object, abstract expressions etc.