This seems reasonable to me. Python 2.7 is no longer supported by any major packages, and switching from Python 3.5 to 3.6+ should be seamless for users.
What binaries are available?
This means that for Pandana v0.5.x, we distribute binaries on Conda Forge for the following environments:
osx-py36, osx-py37, osx-py38
linux-py36, linux-py37, linux-py38
win64-py36, win64-py37, win64-py38
The last version of Pandana with binaries for Python 2.7 is v0.4.4, on Conda Forge.
The last version of Pandana with binaries for Python 3.5 is v0.3, on Pip.
Additional compatibility if you compile your own binaries
The actual code in Pandana v0.5 still works fine with Python 2.7 and 3.5, although I don't think we'll go out of our way to support them going forward. You just need to compile the C++ locally when you install. There are instructions for this in the documentation.
And note that all the binaries mentioned above are for x86-64. We'll cross the ARM bridge when we come to it..
Writing this up based on a Slack discussion with @sablanchard, who noticed that Conda Forge stopped building Pandana for older versions of Python.
Looking into it, Conda Forge policy was to stop building for Python 3.5 in Oct 2018, and to stop building for Python 2.7 in Mar 2020. (https://conda-forge.org/docs/user/announcements.html)
This seems reasonable to me. Python 2.7 is no longer supported by any major packages, and switching from Python 3.5 to 3.6+ should be seamless for users.
What binaries are available?
This means that for Pandana v0.5.x, we distribute binaries on Conda Forge for the following environments:
The last version of Pandana with binaries for Python 2.7 is v0.4.4, on Conda Forge.
The last version of Pandana with binaries for Python 3.5 is v0.3, on Pip.
Additional compatibility if you compile your own binaries
The actual code in Pandana v0.5 still works fine with Python 2.7 and 3.5, although I don't think we'll go out of our way to support them going forward. You just need to compile the C++ locally when you install. There are instructions for this in the documentation.
And note that all the binaries mentioned above are for x86-64. We'll cross the ARM bridge when we come to it..
To do