Move mesa.experimental_cell to mesa.discrete_space

[quaquel]

As part of the stabilization of cell spaces, this PR moves everything to mesa.discrete_space. I have kept the mesa.experimental.cell_space, which should continue to function but will issue a deprecation warning.

I updated all tests and all examples that use cell spaces.

[tpike3]

I think the deprecation warning needs to be updated, from the build failure in test examples it is saying

"DeprecationWarning: you are importing from mesa.discrete_space, all cell spaces have been moved to mesa.discrete_space"

shouldn't this be "DeprecationWarning: you are importing from mesa.experimental.cell_space, all cell spaces have been moved to mesa.discrete_space"

[tpike3]

@quaquel Thanks for doing this!

Before you merge could you update the tests or the pyproject so the build isn't failing with the deprecation warning.

[EwoutH]

I would like to take a good look at namespaces, also to make it futureproof for other network spaces etc. Hopefully I can do that Tuesday.

Also we should split off a 3.1 maintenance branch before merging.

[quaquel]

Before you merge could you update the tests or the pyproject so the build isn't failing with the deprecation warning.

Yes, this is very much on my list of things to do before merging.

I would like to take a good look at namespaces, also to make it futureproof for other network spaces etc

No idea what you mean with this. There is a network in the current mesa.spaces, and one in mesa.experimental.cell_spaces. These both assume that the network structure is fixed. So there is room for a changing network space (although I am not sure about a real use case for it).

[quaquel]

build/examples is now failing on the deprecation warning as intended. This is running mesa-examples. All other tests and builds work fine.

[tpike3]

build/examples is now failing on the deprecation warning as intended. This is running mesa-examples. All other tests and builds work fine.

If possible, can you ensure the build passes, just because it may hide other issues, pytest should be able to handle it either

in the test_examples with something like

def test_with_deprecation():
    with pytest.warns(DeprecationWarning, match="you are importing from mesa.experimental.cell_space, all cell spaces have been moved to mesa.discrete_space"):
        warnings.warn("you are importing from mesa.experimental.cell_space, all cell spaces have been moved to mesa.discrete_space", DeprecationWarning)

or globally in the pyproject.toml

[pytest]
filterwarnings =
    ignore:.*some specific warning message.*:DeprecationWarning

[EwoutH]

Currently the from mesa.discrete_space.discrete_space import DiscreteSpace feels a bit weird. Maybe it makes more sense to keep the top level space named cell_space, it would make it very distinct and the concept of cell is intuitive.

Otherwise, if you, Tom and other @projectmesa/maintainers agree, don't let me block you.

[quaquel]

you can actually use a shorter import: from mesa.discrete_space import DiscreteSpace.

[quaquel]

@tpike3: the failing action is the one that runs mesa-examples. So, this build runs all examples in the mesa-examples repo. All tests on the mesa repo itself, including the examples that ship with mesa itself, pass without problems. Our most complete integration test is erroring and only because of a deprecation warning.

About the only thing I see that can be done to get the built to pass is to explicitly ignore the deprecation warnings by changing this line in built_lint.yml. Let me know what you think about that.

[tpike3]

@tpike3: the failing action is the one that runs mesa-examples. So, this build runs all examples in the mesa-examples repo. All tests on the mesa repo itself, including the examples that ship with mesa itself, pass without problems. Our most complete integration test is erroring and only because of a deprecation warning.

About the only thing I see that can be done to get the built to pass is to explicitly ignore the deprecation warnings by changing this line in built_lint.yml. Let me know what you think about that.

Works for me, thanks!

[quaquel]

Ok, all tests now pass correctly so this is ready to be merged. However, I want to hold of for a short while still. (1). we need to first create a 3.1 maintenance branch, and (2) I ran into a small bug in hexgrid. I know how to resolve the bug, which is due to mixing up (row, column) indexing in hexgrid and (x, y) coordinates when visualizing.