Timelines around support for legacy Python

[PeterJCLaw]

Just wondering what the plans were for keeping supporting running this package under legacy Python?

As demonstrated by #106, doing so isn't free (even if it is relatively low effort), so it might be a good idea to consider dropping Python < 3.7 at some point.

It's worth noting that IPython, possibly the biggest package which depends in some form on asttokens hasn't supported < 3.5 for over 5 years now, which hopefully means that that isn't the driving reason to keep supporting the older versions.

[dsagal]

The driving force for supporting Python2 is actually Grist (https://github.com/gristlabs/grist-core), which was the driving force for creating asttokens. Specifically, the reason we need to keep supporting Python2 is that Grist has users with documents created when Python2 was still the default (this was in turn driven by a choice sandboxing that only worked with Python2).

If the pain of supporting both version were greater, we could update the major version of asttokens, drop Python2 support, and say that any packages that need it have to use an older asttokens version. At this point, this seems a greater hassle than supporting both Python2 and Python3 :)

[PeterJCLaw]

Thanks!