Dropping support for Python 2

[rjleveque]

@ketch points out that Numpy and Matplotlib are dropping Python 2 support on Jan 1, 2020, effectively forcing us to do the same.

I think everything is currently working fine in Python 3 and so perhaps for new code to go in the next release we should not worry about Python 2 support, e.g. clawpack/clawutil#138.

In the meantime, Clawpack users should prepare for this by making sure any Python code they have related to Clawpack works well in Python 3 and with the latest versions of Numpy/Matplotlib.

[mandli]

I am fine dropping Python 2 support. Perhaps we should try and do the same as the NumPy/SciPy teams are doing by pointing users to the last fully Python 2 compatible version of clawpack.

[ketch]

pointing users to the last fully Python 2 compatible version of clawpack

Yes, we should definitely do that, and include the versions of dependencies that it works with.

Perhaps we should post a message to claw-users some time in advance of the first Python-2-incompatible release.

[mandli]

Sounds good to me. We should probably pick a version to coincide with this for ease of pointing as well. Perhaps we should start a discussion on claw-dev or claw-users in terms of dating this change?

[rjleveque]

We could make 5.7.0 the first Python3 only version since it may be a few months before that appears. There are some major GeoClaw improvements I hope to get in over this time frame. It might be nice to have them in the Python2 compatible version, but I suspect they may get further improved or debugged in future releases so best if we don't have something half-baked in the "final Python2 release".

In the meantime if there are smaller changes we can release a 5.6.1 etc. and try to clean up anything that we want to make sure is available forever as Python2, and even I suppose we could add more to the 5.6.x line after 5.7.0 is released if necessary. So we could say 5.6.x is still Python2 compatible but >= 5.7.0 is not necessarily.

Does this make sense?

We could send a note to claw-dev for starters pointing to this issue and encourage more input. Then something to claw-users before long encouraging them to start switching if they haven't already.