git.net

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Python-ideas] Re: Magnitude and ProtoMagnitude ABCs — primarily for argument validation


On Tue, 10 Mar 2020 at 13:41, Chris Angelico <rosuav at gmail.com> wrote:
> It makes good sense for
> division by 0 and division by 0.0 to both result in the same
> exception.

But Python 3 returns a float, for example, in division between
integers. 4 / 2 == 2.0. So some_integer / +0 should return +Infinity.
This is what IEEE 754 says.
Luckily Python does not adhere completely to IEEE 754.

Furthermore, there's a precedence in Python: decimal. decimal.Decimal
does not returns NaN in any of the operations that should return it:
https://en.wikipedia.org/wiki/NaN#Operations_generating_NaN
On the contrary, it returns decimal.InvalidOperation.

@Alan Bawden: thank you for letting me know about William Kahan. I
think it's analysis of IEEE 754 is very detailed, and I didn't know
that traps exist.
And yes, I agree that traps and exceptions slows down the computation.
But in a language like Python, when simplicity and usability is a
priority, and when "no exception should pass silently", IMHO
operations that returns NaN are only an heritage from C, since Python
is written in C.

PS: I'm quoting import this with Tim Peters in the discussion, to Alan
Bawden... It's like teaching physics to Fermi.