Use type hints for returned objects

[alixdamman]

Now we officially do no longer support Python versions under 3.6, it is (maybe) the right time to introduce type hints the code of LArray.

[gdementen]

It is a good idea to at least experiment with this but I wonder how to avoid duplication with what we have in the docstrings. Do you have an example of project using numpydoc style docstrings also using type hints?

Assuming we go down that path (this needs a bit of research before we decide), I wonder what is the best strategy: all at once (a very large task) or as we go.

[alixdamman]

The idea behind is to (as for the ConstrainedSession, see #832) enable auto-completion on PyCharm. When users write something like:

arr = ndtest((3, 3)).transpose()
arr.

PyCharm doesn't know that the tranpose() method returns an array and then writing arr. below the call to transpose() doesn't make PyCharm to propose any auto-completion on Array's methods.

In other words, I am mostly interested to specify the type of the returned object by a method than to specify the type of each argument.

It is a good idea to at least experiment with this but I wonder how to avoid duplication with what we have in the docstrings. Do you have an example of project using numpydoc style docstrings also using type hints?

xarray partly does (for the returned type ;) ): https://github.com/pydata/xarray/blob/c788ee44008cdd65c8b6de40c737f1b28e173496/xarray/core/variable.py#L65

I don't know about pandas but they have a typing.py module.

Assuming we go down that path (this needs a bit of research before we decide), I wonder what is the best strategy: all at once (a very large task) or as we go.

We can start with the type of the returned object if there is one. In that case, I can try to do it for all functions and methods.

[gdementen]

Ok, I missed that relatively recent development of things. It seems like the community is moving towards type hints in addition to numpydoc style docstrings. As you say, the return type is the most important and less verbose change, so we can start with that.