Improve `unicodedata.numeric` return value

[unicodeFun]

Bug report

Bug description:

As 1/20 != 3602879701896397/72057594037927936. Maybe add a Fraction function to the unicodedata object?

import unicodedata
from fractions import Fraction

c = '൛'
str(Fraction(unicodedata.numeric(c)))

CPython versions tested on:

3.11

Operating systems tested on:

Linux

[picnixz]

I don't understand the issue. Isn't it a feature request? what is it the rationale of transforming it to a fraction?

[unicodeFun]

I think there should be a way to get unicodedata.numeric to match to a specific fraction. Or just get the 'Numeric value:' as a text string. Getting it as a float might get different output from system to system.

[zware]

I'm afraid this is just a hazard of mixing floating point numbers (which are generally the standard way to represent non-whole numbers in computers) with fractions; see the Fraction.limit_denominator method if you really need to do so.

[unicodeFun]

The raw unicode data has numerator & denominator. But python's unicodedata seems to have no access to this.

[unicodeFun]

Example: ൛

[StanFromIreland]

The raw unicode data has numerator & denominator.

unicodedata.numeric returns a float of that fraction, and not the fraction itself, but there is no difference, e.g. 1/2 is equivalent to 0.5?

The "odd" output you are seeing is because you are mixing floats and Fractions.

>>> import unicodedata
...
... number = unicodedata.numeric('൛')
... number
0.05
>>> 0.05 == 1/20
True

The Fraction output is expected:

>>> from fractions import Fraction
... Fraction(1/20)
... 
Fraction(3602879701896397, 72057594037927936)

Therefore,

>>> 1/20 == 3602879701896397/72057594037927936
True

I think this should just be closed. I do not see anything reasonable to be done here, unicodedata is not the place for a(nother) function that converts floats to fractions.

[picnixz]

I see your point but are the numerator and denominator meaningful separately? if not, the floating-point representation is sufficient which I believe it is.

[skirpichev]

I see your point but are the numerator and denominator meaningful separately?

Sure!

E.g. 0x2154 glyph is "Vulgar Fraction Two Thirds". While numeric() just returns a float result for 2/3, which can't be represented exactly as a binary fraction. And you will have to guess how to limit denominator to get a correct fraction back.

Function, that returns a pair instead (or just a Fraction) - will be much more useful than the floating-point result.