Object proxy

[Korijn]

Closes #103

I think I got it down pretty nicely this time.

It required a very different approach, and a way to distinguish stateful attributes from callables.

[berendkleinhaneveld]

Nice surprise to see that you've managed to tackle this!

Would love to play around with this. I guess the test suite might need some more love, but that can also wait until we've tried it out for a bit.

Would be nice to see an update to the benchmark to get a sense of how it affects performance.

Just came to mind: does the to_raw method still work? Or did I miss that in the review?

[Korijn]

Just came to mind: does the to_raw method still work? Or did I miss that in the review?

Not for objects. I can't think of a way to implement it.

[Korijn]

Nice surprise to see that you've managed to tackle this!

I am very relieved that I can finally put this to rest. It's been in the back of my mind for years. 🤣

I have to say I am surprised this works as well as it does. It's still technically possible to circumvent the mechanisms here, but it requires very specific, outlandish setups.

For example, ndarrays are objects but I do not think they will be reactive. We have to test I guess.

[Korijn]

For example, ndarrays are objects but I do not think they will be reactive. We have to test I guess.

ndarrays (and memoryviews) are definitely not reactive. We would have to write a separate ndarray_proxy.py module for that and I would not recommend that anybody use it. :) So I extended the type_test to exclude ndarray's from proxying.