> But so it Python.

For a very different definition of "good enough," i.e. unless you're working on a large codebase maintained by a large team for years, need high performance, or need deep low-overhead observability. So Java is "good enough" for a much wider selection of "serious" (large/mission-critical/long-lived/ high-performance) software.

> And for specialized things, there are better choices available to me.

That's OK, but in reality Java does very successfully compete with almost all the languages you've mentioned; it is they that, in practice, find it very hard to compete with Java.

Dropbox, instagram, google and youtube all were written in Python at some point. Languages don't scale, architectures do.

But even when the language became the bottleneck, they went the road of rust and go, not the road of java.

Again, not to say Java is not used by those companies with great success.

Python is a distant fifth place at Google today behind Java, C++, JavaScript and Go. It may be behind Typescript too at this point.

Performance critical, lower-level servers are typically written in C++ at Google - think things like Spanner or GCS. Google only allows a very restrictive C++ subset.

Application servers are typically written in Java or Go. Which is preferred depends on the organization.

Google of course makes extensive use of Guava and Guice, and Bazel, which itself is written in Java.

There is an absolute mountain of Java at Google. Pretty sure it's the most used language by LOC.

As if LoC was a usable metric when it comes to a language as verbose as java.

Plus all the boilerplate coming not from the language but the OOP design patterns.

No one said it was a usable metric for anything other than the fact that a lot of Java is written at Google.

I bet it has more to do with their HQ location and feeling trendy than anything else.

Well, it's true popularity in tech is not necessarily related to quality.

Trending does have an impact.

But by experience, having a killer feature, and a good adoption story usually does much more.

E.G:

- PHP: easy to make web page in 2001

- JS: monopoly on the client side web

- Go: simple concurrency and binary production

- Rust: borrow checker

- Python: "yes, you can do that too. Also in 2 lines"

Etc.

re Python: "...All actual work is done by C libraries though, you could just as easily use BASIC instead"

Well, believe it or not, VBA is doing to have to do with Python competition in Excel in a not so distant future:

https://excel.uservoice.com/forums/304921-excel-for-windows-...

> While we don’t have specific plans to announce at this time, we have been researching the topic, conducting customer interviews, and are working with the Python team at Microsoft so we can build a plan that we think can address the scenarios you told us about and ensure it can run wherever Excel runs.

Any long time developer on MS stack knows that actually means "thank you very much but no".

They just hired Guido, the father of Python, at Microsoft.

So, I would not be so sure.

Yeah, because he did so much for Python tooling at Google.

Google is hugely invested into Python, especially the data engineering branch. TensorFlow, Jax, etc. Though they will probably get replaced by Dex lang, they even dropped Swift 4 Tensorflow to focus on Dex [1]. The language looks very Pythonic compared to Julia which is a 180-degree turn in terms of language design.

[1] https://github.com/google-research/dex-lang

All of the Tensorflow underlying libraries are implemented in C++ and nowadays usable from almost any language with native interop.

As for Dex,

> Research language for array processing in the Haskell/ML family

> This is an early-stage research project, not an official Google product.

Where is Python?

Take a look at the code examples, looks like Python & Numpy + Haskell had a baby. And yes, it is developed by the creators of PyTorch (the irony) and some of the guys from Tensorflow, Jax and S4TF at Google Research & DeepMind. Read their paper here, it's pretty interesting. [1] It may never become an actual production language but it is many times more interesting than both S4TF and Julia.

[1] https://openreview.net/pdf?id=rJxd7vsWPS

Oh this looks nice.

But it’s also a google side project, so previous experience dictates that it’ll be dead somewhere between arrival and 5 years time.

- Ruby: "yes, you can do that 5 different ways in 1 line"

Ruby: Buggy async io for 20 years

Ruby, readable Perl since 1995.