> This pattern, for example, can’t be lazy: try: import numpy; except ModuleNotFoundError: ...

Wait, that seems bad. These kinds of modules (especially numba) are often exactly the ones that you want to delay importing. Why not provide a way to check the existence at import time? How are you supposed to handle nonexistence in that case?

> The error here will move to the first usage of something from numpy. There is a semi-lazy alternative:

  import importlib.util

  if importlib.util.find_spec("numpy") is None:
  ... # whatever you wanted to do if numpy is missing

  lazy import numpy

...crash?

Seriously: crash if your dependencies aren't available. There are better ways to do optional dependencies. ImportError ain't it.

What is the right way to handle an optional dependency?

I don't understand, sorry. How does declaring an optional dependency in a toml file solve the problem of accessing the symbol in your code?

Also, you need something that works for simple one-file scripts, not just a whole package or project.

I think this import flow tends to be the canonical one.

I might recommend doing something like importing from something like `numpy.version` (or some other "random" very small utils package) so that the work done on file load is still fairly small.

Common? Yes. Canonical/ideal? No. Optional dependency specifiers in your package metadata is a much better way to go. You can use those even if you don't publish your package as an artifact; `pip install -e .[optional-thing]` should work, or can be made to, easily.

That's only part of the puzzle. Your program needs to branch on whether the dependency is installed or not!

"Oh you don't have optional dependency? Then do X" needs to happen at runtime right?

It's not clear such a thing even exists in general though. It's certainly not guaranteed by packages.

What would that be for numba?

Yep it's not clear, and a bit of a case-by-case thing. Fortunately most codebases only need to do this in a handful of spots so you can really "just" look at a package and find a file that probably is safe.

For example, numba._version only imports standard library stuff so is probably good enough here[0]

It would be nice to have general querying capabilities here, of course. I just think that in practice there's at least a halfway-decent workaround in almost any real situation

[0] https://github.com/numba/numba/blob/main/numba/_version.py

I've got my reservations about lazy imports, but scipy in particular is such a memory hog, I've increasingly been vendoring smaller utilities out of it. Good lazy imports would be a big help.

doesn't `from scipy import ...` help with memory?

This feature seems really valuable for commandline tools! However this paragraph gave me pause:

> Currently, disabling lazy imports disabled the syntax keyword, which means that you can’t use it for circular imports, type checking, etc.

Imo this is a good thing, laziness shouldn't be semantically important. The ability to force disable laziness while maintaining semantics is important for linting and testing, and more often than not, circular imports are a sign of bad code structure.

This seems like an ugly hack. We really need a way to snapshot a python program after its imports have been loaded, similar to temacs/emacs that's been around for decades, or a similar thing that exists in gforth. Emacs's old unexec scheme became unmaintainable but maybe there are still good alternate ways to do it.

Is there a language with a hybrid laziness approach? similar to "do work only if needed" but if you have ressources, do work most likely to be needed, like lazy anticipating?

Good feature but god I hate there’s a whole new keyword for a minor modification on an existing concept. Very unnecessary.

It's a soft keyword: https://docs.python.org/3.15/reference/lexical_analysis.html...

> As soft keywords, their use in the grammar is possible while still preserving compatibility with existing code that uses these names as identifier names.