fix issue with repr and Decimal

[ghost]

Given the issue #431 fixing a precision issue while using python 2.x, as demonstrated by the example below:

$ python
Python 2.7.13 (default, Jan 19 2017, 14:48:08)
>>> float(repr(0.8289445733333332)) == 0.8289445733333332
True
>>> float(str(0.8289445733333332)) == 0.8289445733333332
False
>>>

We have now an issue as described in #430 and #482 with both Python and Python3 and the use of both Decimal Python type and repr, because :

>>> from decimal import Decimal
>>> str(Decimal(0.8289445733333332))
'0.828944573333333156739399782964028418064117431640625'
>>> repr(Decimal(0.8289445733333332))
"Decimal('0.828944573333333156739399782964028418064117431640625')"

In Python 3, str() and repr() now provides the same result with a float:

$ python3
Python 3.5.3+ (default, Jun  7 2017, 23:23:48) 
>>> float(repr(0.8289445733333332)) == 0.8289445733333332
True
>>> float(str(0.8289445733333332)) == 0.8289445733333332
True
>>>

But the problems remain while using both Decimal and repr(). The pull request offers a solution which should work both for python2 and python3.

Regards,
Carl

[sebito91]

Why make this so convoluted, since we're passing this to _is_float already (and succeeding), why not just return repr(float(value))?

[sebito91]

Why not just make this return repr(float(value)) since you've already passed the _is_float test?

[ghost]

I guess I went that way because I didn't want to lose precision of the Decimal, but I don't know if it's worth it, your solution is quite shorter and more elegant.

[sebito91]

If you're referring to your example earlier, even if Decimal resolves to a more granular value than what you sent it may not be what you actually want. If you wanted that value down to the full granularity you should provide that to begin with (IMHO of course).

Do you want to take a stab at changing your PR? If so, can you please add a quick test to check both a non-Decimal and Decimal value?

[sebito91]

To clarify, since your Decimal value is set to 16 digits, you should get the repr value of the 16-digit float instead of a (potentially) longer, more granular definition.

In [1]: from decimal import Decimal

In [2]: str(Decimal(0.8289445733333332))
Out[2]: '0.828944573333333156739399782964028418064117431640625'

In [3]: repr(Decimal(0.8289445733333332))
Out[3]: "Decimal('0.828944573333333156739399782964028418064117431640625')"

In [4]: repr(float(Decimal(0.8289445733333332)))
Out[4]: '0.8289445733333332'