Python issues with math.floor: int too large to convert to float

Question

I'd like to calculate (⌊2^(1918)*π⌋+124476) in python but I get this error when I do it using the following code:

b = math.floor((2**1918) * math.pi) + 124476
print(b)

OverflowError: int too large to convert to float

How can you get this to work? In the end I just like to have it all as hexadecimal (if that helps with answering my question) but I was actually only trying to get it as an integer first :)

Answer 1

The right solution really depends on how precise the results are required. Since 2^1918 already is too large for both standard integer and floating point containers, it is not possible to get away with direct calculations without loosing all the precision below ~ 10^300.

In order to compute the desired result, you should use arbitrary-precision calculation techniques. You can implement the algorithms yourself or use one of the available libraries.

Assuming you are looking for an integer part of your expression, it will take about 600 decimal places to store the results precisely. Here is how you can get it using mpmath:

from mpmath import mp
mp.dps = 600
print(mp.floor(mp.power(2, 1918)*mp.pi + 124476))

74590163000744215664571428206261183464882552592869067139382222056552715349763159120841569799756029042920968184704590129494078052978962320087944021101746026347535981717869532122259590055984951049094749636380324830154777203301864744802934173941573749720376124683717094961945258961821638084501989870923589746845121992752663157772293235786930128078740743810989039879507242078364008020576647135087519356182872146031915081433053440716531771499444683048837650335204793844725968402892045220358076481772902929784589843471786500160230209071224266538164123696273477863853813807997663357545.0

Next, all you have to do is to convert it to hex representation (or extract hex from its internal binary form), which is a matter for another subject :)

Answer 2

The basic problem is what the message says. Python integers can be arbitrarily large, larger even than the range of a float. 2**1918 in decimal contains 578 significant digits and is way bigger than the biggest float your IEEE754 hardware can represent. So the call just fails.

You could try looking at the mpmath module. It is designed for floating point arithmetic outside the bounds of what ordinary hardware can handle.