Bug in Angle Dimensionality

[rbreslavsky]

I noticed an issue with regards to angle dimensionality. For one, when working with degrees to some exponent, n, pint does not distinguish degrees ** n from degrees. Furthermore, I can convert to degrees from degrees ** n, but this changes the numbers in a way that is obfuscated.

import pint

ureg = pint.UnitRegistry()

test = 5.0 * ureg('deg')
test_squared = test ** 2

Then I get:

In: test_squared
Out: <Quantity(25.0, 'degree ** 2')>

In: test_squared.is_compatible_with('deg')
Out: True

In: test_squared.to('deg')
Out: <Quantity(0.436332313, 'degree')>

The last output has converted <Quantity(25.0, 'degree')> to a radian equivalent, which was not the intent.

[jules-ch]

Do you have a physical example or use case in mind ?

This comes from the opinionated choice to treat angle as dimensionless:
https://pint.readthedocs.io/en/stable/user/angular_frequency.html

Difficult to treat this, I'd say. what would you have expected there ?

[jules-ch]

It is just a matter of convention, but conversion for steradian are correct when dealing with degrees squared:

>>> test = 180**2/math.pi**2 * ureg('deg')**2
>>> test.to("sr")
<Quantity(1.0, 'steradian')>

[jules-ch]

Related #1288

[rbreslavsky]

Thanks for taking the time to respond. This occurrence actually happened due to a bug in some code where a pint.Quantity of unit degrees was multiplied by ureg('deg') a second time.

The problem is that I tried to implement a check to make sure dimensionality was correct (using .is_compatible_with()). This check failed, and then I was allowed to strip units using .m_as() in another function, but that action effectively converted the answer to radians.

I think an error of some sort would be useful here (for example, when attempting to convert a non-radian angular dimension of an nth power to a root of itself). Could that be accommodated for angles? I know it's a niche issue, but pint has been fantastic at avoiding this kind of problem elsewhere.

[andrewgsavage]

possibly a warning yes, I had a similar idea here:
#1554