PyOxidizer VS dh-virtualenv

Bundling Python isn't too bad if you find the right tools for it.

I really like https://github.com/indygreg/python-build-standalone and https://github.com/indygreg/PyOxidizer

A bundled, built standalone Python can be 16 to 32MB (including the full standard library, which you can strip down to just the bits you use to save size). Not tiny, but probably not worth switching programming languages over.

But really, I would suggest thinking about what you want to build before "how" or "with which tool" - one of the signs of a person becoming a good engineer is having an array of tools at their disposal and being able to choose a correct tool for the correct task. Rust also excels in integrating with other languages - with JS via WebAssembly (a bit of self-promotion, for example), with Elixir via Rustler, with Python via PyO3 and PyOxidizer, etc. So you absolutely can start writing a frontend app with JS, or a distributed system with Elixir, or a data processing/ML app with Python and use Rust to speed up critical parts of those. Or, in reverse, you can start with Rust & add new capabilities to whatever you're building, that being a frontend, a resilient chat interface, or an ML model.

Thank you, although this is not exactly on topic. I'd not heard of PyOxidizer, but it appears to have the same goal as PyInstaller, py2exe, and cx_Freeze -- as the PyOxidizer readme says, it produces

Here is some example Github Action from PyOxidizer as a Kickstarter: https://github.com/indygreg/PyOxidizer/blob/main/.github/workflows/build-exe.yml

A starting point to try out binary modules by the way would be https://github.com/indygreg/PyOxidizer - could already have benefits by rolling in all dependencies of modules (so no more pip/apt/dnf/... installs on target hosts). Setting this up should be relatively straightforward and could probably be automated enough to even manage to build binary modules for all modules in the community ansible distribution eventually.

PyOxidizer might be another option.

XAR signing is effectively just an RFC 5652 CMS signature plus some minimal data structure manipulation. Code at https://github.com/indygreg/PyOxidizer/blob/faa7dfcea5d66bf5....

Mach-O and bundles, by contrast, require a myriad of additional data structures requiring thousands of lines of code to support. To my knowledge, nobody else has implemented signing of these far-more-complicated primitives. (Existing Mach-O signing solutions just do ad-hoc signing and/or don't handle Mach-O in the context of a bundle.)

This is how Azure package functions, iirc. I think I used this https://github.com/spotify/dh-virtualenv for similar purpose some time ago.

I'm in a devops role where we actually reroll the Tensorflow whl in-house (to get a few tweaks like specific AVX flags turned on), but because the rest of our deployment is apt/debs, we then turn around and wrap that whl in a deb using Spotify's excellent dh-virtualenv:

https://github.com/spotify/dh-virtualenv

There's no expertise for Bazel in-house; when we run the build, it seems to fail all its cache hits and then spend 12-13h in total compiling, much of which appears to be recompiling a specific version of LLVM.

Every dependency is either vendored or pinned, including some critical things that have no ABI guarantees like Eigen, which is literally pinned to some a random commit, so that causes chaos when other binaries try to link up with the underlying Tensorflow shared objects:

https://github.com/tensorflow/tensorflow/blob/master/third_p...

And when you go down a layer into CUDA, there are even more support matrixes listing exact known sets of versions of things that work together:

https://docs.nvidia.com/deeplearning/tensorrt/support-matrix...

Anyway, I'm mostly just venting here. But the whole thing is an absurd nightmare. I have no idea how a normal distro would even begin to approach the task of unvendoring this stuff and shipping a set of normal packages for it all.

Rarely do I develop software at my host level, I'm almost always doing it inside of a container or a virtual machine (local or remote), and part of that is simple muscle memory to solve several problems that you've mentioned in your experience with Python, as well as other languages (bonus points for me that I was never plagued by a horrid npm bug). Those are problems that are easily solvable with Docker and a properly formatted requirements.txt (which should only be used in development — use pip/setup.py for proper deployment) and for the rare times when I do need to use Python at the host level of a Linux system, I use Spotify's dh-virtualenv.

Maybe I misunderstood the question, but this would be the point of CI/CD and containerization as far as industry standards are concerned. If you had to deploy at host level, then this would be the point of dh-virtualenv that’s tied in by your CI/CD system.