u.cycle is insane and causes astropy 4 to fail

[paulray]

See my email to the list, but I think we should make Phases unitless instead of using u.cycle which has many unexpected behaviors. Currently it causes one of the tests to fail when using astropy 4.0 since the unit propagates into photonphase in a way it didn't use to

[rossjjennings]

@paulray -- I've tried to reproduce the weird behavior you wrote about in your email, and found that it isn't the default behavior in astropy 4.0. In an astropy 4.0 environment, if I just open up ipython and run

import astropy.units as u
int(20*u.cycle)

I get a UnitConversionError. But if I import pint.models before the second line, I get 125 as output, with no error.

[rossjjennings]

Aha! I think the culprit is u.set_enabled_equivalencies(u.dimensionless_angles()) here. This is the only place where it appears outside a with block.

[paulray]

Glad this is fixed! But, I think we should have a discussion at some upcoming PINT telecon about whether it might be better to remove the u.cycle units on Phases.

[rossjjennings]

Should we close this now that #616 is merged?

[paulray]

I still think we should hold it open pending a discussion on whether to stop using u.cycle at all for pulsar phases.

[paulray]

Fix is #685

[Zac-HD]

FYI you can put a comment in your PR description which auto-closes linked issues 🙂