Python 3.10 is out, and because GHA > travis, we can use it in the build straight away rather than waiting 6+ months... We should also start testing on ubuntu 20.04 (vagrant already uses it so shouldn't be a problem).
We should also drop 3.6. The dependency molesq depends on recent numpy (maybe molesq doesn't need it, but it's a pain to find the MSnumpyV). Numpy (and most of the numerical stack) stopped supporting 3.6 in June last year; 3.6's proper EOL is at the end of the year anyway. This may necessitate a change in docs/ other scripts.
Drop 3.6
[x] CI
[x] docs
[x] check docker, vagrant etc.
Add 3.10
[x] CI
[x] Bump core numpy dep to match molesq's minimum
[ ] Test on ubuntu 20.04
[ ] Run lints etc. on py3.9
Once 3.6 is dropped, we will need to continue supporting 3.7 at least until pypy supports 3.8 (in the works but no predicted release date).
Python 3.10 is out, and because GHA > travis, we can use it in the build straight away rather than waiting 6+ months... We should also start testing on ubuntu 20.04 (vagrant already uses it so shouldn't be a problem).
We should also drop 3.6. The dependency molesq depends on recent numpy (maybe molesq doesn't need it, but it's a pain to find the MSnumpyV). Numpy (and most of the numerical stack) stopped supporting 3.6 in June last year; 3.6's proper EOL is at the end of the year anyway. This may necessitate a change in docs/ other scripts.
Once 3.6 is dropped, we will need to continue supporting 3.7 at least until pypy supports 3.8 (in the works but no predicted release date).