Archive Pages Categories Tags

Unsigned integers and Python

28 June 2013

The easiest (and portable!) way to get unsigned integers in Python is to import the required types from the ctypes module. However, sometimes you need to convert Pythons ‘normal’ integers to unsigned values. As will be explained, the values are still signed integers (or long), but this has little effect on the application.

If you google python unsigned integer or similar terms, you will find several posts telling you that in order to convert a signed to an unsigned value, it is sufficient to do something similar to:

val & 0xff (val is a 8 bit variable)

However, few explain why this works. In order to understand why, you first have to know how Python stores a number. Numbers in Python are stored as either signed integers or long, depending on which Python version you use (follow the links here for more information). For example, the following is the case on my machine (which is 64 bit):

>>> val = 9223372036854775807 (maximum value of int 64)
>>> type(val)
<type 'int'>
>>> val += 1
>>> type(val)
<type 'long'>

By increasing the value of val by 1, I exceed the limit of a signed 64 bit integer and the value is converted to a long. If Python had used or converted to an unsigned integer, val would still have been an int. Or, not long.

Signed integers are represented by a bit, usually the most significant bit, being set to 0 for positive numbers or 1 for negative numbers. What val & 0xff does is actually val & 0x000000ff (on a 32 bit machine). In other words, the signed bit is set to 0 and an unsigned value is emulated.

An example:

>>> val = -1
>>> val & 0xff
blog comments powered by Disqus
Fork me on GitHub