Inconsistent arithmetic with large numbers (possible floating point imprecision)

[apcamargo]

When performing arithmetic with very large numbers in numbat, the order of operations appears to affect the result:

>>> let big_number = 10^100

  let big_number: Scalar = 10^100

>>> big_number + 1 - big_number

  (big_number + 1) - big_number

    = 0

>>> big_number - big_number + 1

  (big_number - big_number) + 1

    = 1

Mathematically, both expressions should yield 1. This discrepancy is could be due to floating point imprecision.

I'm using version 1.16.0.

[sharkdp]

This discrepancy is could be due to floating point imprecision

Yes. Numbat is a pure numerical calculator at the moment. Internally, it uses 64 bit floating point numbers, and 10^100 is outside the range of exactly representable integers (https://stackoverflow.com/a/1848762). So this is expected behavior, but see #4