Is support for __round__() missing at the moment?

[shovnyk]

Hi. I was writing my own class for an RGB color library and decided to define the __round__() special function to convert fractional RGB values into the nearest integer. Having round() implicitly do that would be convenient. But it does not appear to work and I get the following error:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: can't convert ColorRGB to float

Admittedly this is quite trivial and I could very well just write another method for doing this. But I was just wondering if there are plans to add this.

[GitHubsSilverBullet]

An RGB value contain three distinct values for the three named colors.
What esoteric function would combine those three into a single meaningless float value?

[shovnyk]

I wasn't intending to return a float. I wanted to return another RGB object with the values of R, G and B rounded to the nearest integer in the range [0, 255] (because some methods in my class perform computations that make them fractional from time to time). Something like this (in CPython):

class ColorRGB:
   def __init__(self, r, g, b):
      self.r, self.g, self.b = r, g, b
      ...
   def __round__(self) -> ColorRGB:
      return ColorRGB(*[max(0, min(255, round(value))) for value in self])

But as @jimmo explained this is not possible as round calls __float__ instead. That's fine. I'll just use another method instead for this.

[andrewleech]

Did you try adding float function to your class? That's the normal way to allow conversion of a class to float.

Once that's there, the round function should work. I'm actually not sure if round functionality is supported or not.

[jimmo]

@shovnyk __round__ is not supported. This appears to be a deliberate decision, see e.g. #3610

And @andrewleech is correct, round(x) implicitly does float(x) which will call x.__float__.