Thank you very much for the PR.

Do you think you can create a simple test for it?

You can add it to test_debugger_json.py

Some references that do something like you should be testing here (which you can use as a template for the new test) would be : test_pandas, test_stack_and_variables_dict.

One thing I noticed is that this will change things as other containers such as list/tuple/etc.

The current display is:

<class 'list'>: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9,

and with the change it'd be:

len: 50, value: [0, 1, 2, 3, 4, 5, 6, 7, 8, 9,

I think the type info shouldn't be changed on those cases... Maybe instead of customizing something as generic as "anything with a __len__", it could be a bit more scoped to the actual types you are interested in?

-- I'm also a bit worried that we don't have support for priorities in the current mechanism, so, something so generic could override other representations that are more scoped.

Hi fabioz, you are right, perhaps using the Sized type is too generic, leading to an over design. But I think displaying len/shape for np.array and torch.tensor would be super useful for us AL/ML people. I will make a new PR with more restricted code, equipped with tests.

Hi @fabioz , @qige96 ... I made a followup PR doing the same, as I think this would hugely benefit the ML/AI community.