conversion from float to Decimal in python-2.6: how to do it and why they didn't do it

Question

Direct conversion from float to Decimal was implemented in python-2.7, both in Decimal's constructor and with the Decimal.from_float() classmethod.

Python-2.6 instead throws a TypeError suggesting to convert to string first:

TypeError: Cannot convert float to Decimal.  First convert the float to a string

so my usual workaround is this:

if sys.version_info < (2, 7):
    Decimal.from_float = classmethod(lambda cls, x: cls(str(x)))

That's just a literary translation from the error message - and I just don't bother implementing it in the constructor too.

If it's that simple, why didn't they implement it in the first place instead of telling the user to do it in a TypeError? Is that the best method available (and by extension is that the one used in python-2.7 and newer?)

Answer 1

Your workaround is not the RightWayToDoIt(tm) because it loses information. The lossless way to convert is shown in the recipe for float_to_decimal() shown in the Decimal FAQ.

The reason we didn't include Decimal.from_float in Python 2.6 is because we were being conservative about introducing unintentional interactions between binary floats and decimal floats. By Python 2.7, this was all worked out and you can just write Decimal(f) where f is a binary float.

Other than the small nuisance in 2.6, I hope you're enjoying the Decimal module

Answer 2

Probably because the behavior of a direct conversion can be counterintuitive if you don't know a few implementation details about floats. As stated in the docs:

Note Decimal.from_float(0.1) is not the same as Decimal('0.1'). Since 0.1 is not exactly representable in binary floating point, the value is stored as the nearest representable value which is 0x1.999999999999ap-4. That equivalent value in decimal is 0.1000000000000000055511151231257827021181583404541015625.

If you convert to a string, you can control the precision you want to use, so you can get an accurate conversion to Decimal.

The new method was introduced in Python 2.7 - that's why it isn't in 2.6. New features are not backported to older versions.