int(math.log(2**i, 2))

[AlexRhatushnyak]

BPO	17958
Nosy	@mdickinson

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2013-05-12.10:33:03.997>
created_at = <Date 2013-05-12.05:36:48.141>
labels = ['type-bug', 'invalid']
title = 'int(math.log(2**i, 2))'
updated_at = <Date 2013-05-12.10:56:55.393>
user = 'https://bugs.python.org/AlexRhatushnyak'

bugs.python.org fields:

activity = <Date 2013-05-12.10:56:55.393>
actor = 'mark.dickinson'
assignee = 'none'
closed = True
closed_date = <Date 2013-05-12.10:33:03.997>
closer = 'mark.dickinson'
components = []
creation = <Date 2013-05-12.05:36:48.141>
creator = 'Alex.Rhatushnyak'
dependencies = []
files = []
hgrepos = []
issue_num = 17958
keywords = []
message_count = 4.0
messages = ['188976', '189005', '189006', '189012']
nosy_count = 2.0
nosy_names = ['mark.dickinson', 'Alex.Rhatushnyak']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue17958'
versions = ['Python 2.7']

[AlexRhatushnyak]

In Python 2.7.4:

import math
for i in range(40, 55):  print int(math.log(2**i, 2)),

output:

40 41 42 43 44 45 46 46 48 49 49 50 52 53 53

[mdickinson]

Sorry: this is not a bug, but a difficult-to-avoid consequence of floating-point imprecision: math.log(n, 2) is computed as log(n) / log(2), and each of the log computations and the division can introduce small errors.

For what it's worth, Python 3.3 has a log2 function, which has accuracy that's a little bit better than math.log(n, 2), and gives the 'correct' answer exact powers of 2.

Closing as invalid.

[mdickinson]

Postscript: depending on what you're doing, you might also find the int.bit_length method helpful.

[mdickinson]

For tracker archaeologists: see also issue bpo-11888, issue bpo-9959.