View Full Version : [SOLVED] Python large numbers

kroq-gar78

September 25th, 2010, 07:15 PM

Hello Ubuntu Forums community,

I am making a python program that requires the math.pow function. I need it to go up to 2^31 (2 to the 31st power) and preferably even farther. When I run the program in the terminal, it says "OverflowError: math range error"

Any way to fix this?

Thanks in advance.

Bachstelze

September 25th, 2010, 07:20 PM

What are you doing, exactly? math.pow(2, 32) doesn't raise an exception for me.

Queue29

September 25th, 2010, 07:32 PM

Python supports arbitrary precision numbers out of the box. What version are you using? And can you post the code that seems to be troublesome?

ssam

September 25th, 2010, 07:57 PM

Python supports arbitrary precision numbers out of the box. What version are you using? And can you post the code that seems to be troublesome?

thats not right. its just long ints that have infinite range.

http://docs.python.org/library/stdtypes.html#numeric-types-int-float-long-complex

>>> math.sin(pow(2,1000))

-0.15920170308624243

>>> math.sin(pow(2,10000))

Traceback (most recent call last):

File "<stdin>", line 1, in <module>

OverflowError: long int too large to convert to float

worksofcraft

September 25th, 2010, 08:54 PM

thats not right. its just long ints that have infinite range.

http://docs.python.org/library/stdtypes.html#numeric-types-int-float-long-complex

>>> math.sin(pow(2,1000))

-0.15920170308624243

>>> math.sin(pow(2,10000))

Traceback (most recent call last):

File "<stdin>", line 1, in <module>

OverflowError: long int too large to convert to float

I thought it would calculate pow as floating point number anyway... what on Earth is Python doing... 10000 multiplications?

Anyway as far as I can tell most languages use floating point as provided by their processor which is usually IEEE-754. I think it give about 16 digits of precision and exponent up to something like 300 in decimal.

Bachstelze

September 25th, 2010, 09:07 PM

I thought it would calculate pow as floating point number anyway... what on Earth is Python doing... 10000 multiplications?

pow is integer, math.pow is float.

>>> pow(2, 1000)

10715086071862673209484250490600018105614048117055 33607443750388370351051124936122493198378815695858 12759467291755314682518714528569231404359845775746 98574803934567774824230985421074605062371141877954 18215304647498358194126739876755916554394607706291 45711964776865421676604298316526243868372056680693 76L

>>> math.pow(2, 1000)

1.0715086071862673e+301

Incredible as it may sound, having integer powers is actually very useful. When doing arithmetic for example.

kroq-gar78

September 25th, 2010, 10:09 PM

Thanks Bachstelze. I didn't know there was just pow() and that there was a difference between pow() and math.pow().

Marking thread as solved.

Bachstelze

September 26th, 2010, 11:59 AM

Thanks Bachstelze. I didn't know there was just pow() and that there was a difference between pow() and math.pow().

Marking thread as solved.

Yup, this is why you should never do

from math import *

CptPicard

September 26th, 2010, 01:01 PM

pow is integer, math.pow is float.

Well, unless you give pow float parameters.

That math.pow is specifically a float is interesting; could be one of these rare cases where I might actually agree with worksofcraft's objections to Python's handling of number types.

Bachstelze

September 26th, 2010, 01:14 PM

Well, unless you give pow float parameters.

That math.pow is specifically a float is interesting; could be one of these rare cases where I might actually agree with worksofcraft's objections to Python's handling of number types.

It makes sense to me. As I said, in some applications you really want full-precision exponentiation of integers (e.g. when doing RSA crypto), but on the other hand, sometimes you don't need the precision, and having long integers is just a bother.

Having both is a nice thing about Python IMO.

CptPicard

September 26th, 2010, 01:36 PM

Having both is a nice thing about Python IMO.

Yeah, but it should then be pythonically generic, output according to input, as in pow, and not inconsistent in math.pow with that idea with a requirement to just know that this is how math.pow works. That's my only objection to it really.

Bachstelze

September 26th, 2010, 02:55 PM

I don't know, it's intuitive enough to me: math.pow always returns float, and pow returns int when it makes sense (like addition, you can't really expect it to return int when you give it float arguments). I must admit I never use pow, though, I use ** for integer exponentiation, and math.pow for floats.

Wybiral

September 26th, 2010, 09:59 PM

Yeah, but it should then be pythonically generic, output according to input, as in pow, and not inconsistent in math.pow with that idea with a requirement to just know that this is how math.pow works. That's my only objection to it really.

I agree. It's "import math" not "import floatmath" (which is more like what it really is). Though, I still feel like integer overflow should result in bignum :) But, I also feel like floating point overflow should result in decimal/rational representation... Not range exceptions.

I must admit I never use pow, though, I use ** for integer exponentiation, and math.pow for floats.

I use "pow" all the time. It works as you'd expect it to work, pow(2.0, 2) == 4.0, pow(2, 2) == 4. The math module to me is just for trig functions and floor/ceil.

jpkotta

September 27th, 2010, 02:44 AM

From http://docs.python.org/library/math.html:

This module is always available. It provides access to the mathematical functions defined by the C standard.

It makes sense then, because the C math library functions return double (ignoring the variants that take floats and longs). math.pow() works like C's pow() instead of Python's pow().

nvteighen

September 27th, 2010, 08:21 AM

It makes sense then, because the C math library functions return double (ignoring the variants that take floats and longs). math.pow() works like C's pow() instead of Python's pow().

No, it doesn't make sense: it adds a weird inconsistency in the language. While the rest works generically, lots of times hiding the details of the C Standard Library, math.pow doesn't and brings in requirements that are out of place in a duck-typed language. Ok, you may say it's a minor issue because there's pow(), but it confused the OP, who is a beginner... how many other beginners may have fallen into this pitfall?

Bachstelze

September 27th, 2010, 08:49 AM

No, it doesn't make sense: it adds a weird inconsistency in the language.

What's the inconsistency? All the functions in math return a float, seems pretty consistent to me. The confusing thing is that there's pow and math.pow, which are different, but that's a questionable design choice, not an inconsistency.

nvteighen

September 27th, 2010, 04:58 PM

What's the inconsistency? All the functions in math return a float, seems pretty consistent to me. The confusing thing is that there's pow and math.pow, which are different, but that's a questionable design choice, not an inconsistency.

The inconsistency lies in that where everything in Python is meant to work as generically as possible, there you have math.pow as a float-specific power function and pow as a generic power function, when only one should actually exist (the second). If someone needs to make a float of an int, then he should cast it, not the language.

jpkotta

September 28th, 2010, 01:22 AM

What's the inconsistency? All the functions in math return a float, seems pretty consistent to me. The confusing thing is that there's pow and math.pow, which are different, but that's a questionable design choice, not an inconsistency.

The inconsistency lies in that where everything in Python is meant to work as generically as possible, there you have math.pow as a float-specific power function and pow as a generic power function, when only one should actually exist (the second). If someone needs to make a float of an int, then he should cast it, not the language.

To me, math.pow is not part of Python, but pow is. math.pow is a library function that happens to come with Python, and thus doesn't have all of the expectations of pow. But it definitely seems silly to have two functions that do the same thing, especially when one does it poorly.

BTW, pow is really a method call, like len and repr.

Powered by vBulletin® Version 4.2.2 Copyright © 2019 vBulletin Solutions, Inc. All rights reserved.