differences between py2 and py3

[Skaperen]

this code, using new division in both Python2 and Python3, shows there are still differences.

from __future__ import division,print_function
efs = 9007199254739968 * 1024
print(str(efs/(2**10)),'kB kilo bytes')
print(str(efs/(2**20)),'MB Mega bytes')
print(str(efs/(2**30)),'GB Giga bytes')
print(str(efs/(2**40)),'TB Tera bytes')
print(str(efs/(2**50)),'PB Peta bytes')
print(str(efs/(2**60)),'EB Exa bytes')

FYI, this is the size of a filesystem available in the AWS cloud. i have never been able to fill it up.

[gjenkinslb]

import __future__ import division, print_function
in python 2 without the future import, / does integer divide so 3/2 is 1
with the future import / does a float divide so 3/2 is 1.5
you need to use the // operator: elf//(2**10) for example. this does integer divide in both.

from __future__ import print_function # don't need the division now
efs = 9007199254739968 * 1024
print(str(efs//(2**10)),'kB kilo bytes')
print(str(efs//(2**20)),'MB Mega bytes')
print(str(efs//(2**30)),'GB Giga bytes')
print(str(efs//(2**40)),'TB Tera bytes')
print(str(efs//(2**50)),'PB Peta bytes')
print(str(efs//(2**60)),'EB Exa bytes')

[Skaperen]

actually, i wanted to do float, not int, else i'd get 7 for exabytes. i did a little more digging and found that it is the str() function, or whatever it uses, that is different.

Output:
lt1/forums /home/forums 1> py3 
Python 3.5.2 (default, Nov 23 2017, 16:37:01) 
[GCC 5.4.0 20160609] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> str(7.9999999999990905)
'7.9999999999990905'
>>> 
lt1/forums /home/forums 2> py2
Python 2.7.12 (default, Nov 20 2017, 18:23:56) 
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> str(7.9999999999990905)
'8.0'
>>> 
lt1/forums /home/forums 3>

---

still more digging: the float arithmetic is the same. it's the conversion that is different.

---

it could appear that Python2 has only 38 bits of float precision, but in fact it has 53 (on i386/amd64) just as Python3. i presume the real difference is the configured number of digits in the conversion to decimal (w/o reading the docs, yet). i think i will make a Python version of the float analyzer i once wrote in C. it could be a useful tool if we ever get to use Python on an architecture with more bits of precision than x86/amd64. there are some out there, and gcc can be made to compile double to a larger size on many platforms (you can go up to 64 bits on i386/amd64 and up to 112 bits on S/370).

[casevh]

There were changes made to the output formatting of floats in Python 2.7 and more changes in Python 3.2.

Here is an online presentation by the author of changes:

https://www.slideshare.net/dickinsm/magical-float-repr

[Skaperen]

that's probably the difference i'm seeing.