In 3.x the AffineScalarFunc class has two attributes that tell users about the contribution of other Variables to the uncertainty: the derivatives property and the error_component method.

x = ufloat(1, 0.1)
y = x**2

y.derivatives
y.error_components()

One is a property, so it is accessed without the () parentheses, the other is a method so it needs the parentheses (). In 4.x the derivatives property is being removed from the UFloat class. error_components remains. This issue is asking if error_components should be changed to a property.

• The pro is that error_components() is just returning something it reads off the UFloat._uncertainty (which is a UCombo) attribute. I would say semantically it is more like a property than a method. The simpler notation would reflect that.
• The downside is that it would not be backwards compatible. Also I generally find it tricky to remember when an attribute is a property and whether or not I should use (). This change in behavior will make that even worse.

I would take a weak preference towards changing error_components to be a property and will do so if no one pushes back. But I'm very open to pushback. If we want to change it we need to put in a deprecation/change coming warning in 3.2.x.