Diffing two floating point numbers

[seperman]

Diffing two floating point numbers

[seperman]

@bwest-at-xometry I'm finally back and catching up with the tickets.
Interesting point you bring up regarding the definition of the significant digit. I think it might be a good idea to give the user the flexibility to pass their definition of "significant digits". What do you think about that?
The math.isclose is tricky since we don't know how "close" the numbers are during the hashing process. The hashing process is used when ignore_order=True and during that process, the significant digits are used to transform the numbers and then hash them.

[bwest-at-xometry]

Flexibility would be great.

[seperman]

I will try to address this in v4.2.0 whenever I can...

Or if you have time, please feel free to make a PR. What I'm imagining is maybe a parameter called: significant_digits_format and people can pass their own functions. For example the default one is:

lambda x, significant_digits: ("{:.%sf}" % significant_digits).format(x)

And then you could pass:

def scientific_significant_digits_format(x, significant_digits):
      return lambda x, significant_digits: ("{:.%se}" % (significant_digits - 1)).format(x)


DeepDiff(t1, t2, significant_digits_format=scientific_significant_digits_format)

That gives the full control to the user.
If you make the PR, please have that function passed into the DeepHash too.

Thanks!

[seperman]

forgot to tag you! @bwest-at-xometry

[seperman]

@bwest-at-xometry This feature is now published. There is the number_format_notation now that you can set to e for scientific notation. Also for power users the number_to_string_fun is introduced.
https://deepdiff.readthedocs.io/en/stable/diff.html