chr() equivalent returning a bytes object, in py3k

Question

Python 2.x has chr(), which converts a number in the range 0-255 to a byte string with one character with that numeric value, and unichr(), which converts a number in the range 0-0x10FFFF to a Unicode string with one character with that Unicode codepoint. Python 3.x replaces unichr() with chr(), in keeping with its "Unicode strings are default" policy, but I can't find anything that does exactly what the old chr() did. The 2to3 utility (from 2.6) leaves chr calls alone, which is not right in general :(

(This is for parsing and serializing a file format which is explicitly defined in terms of 8-bit bytes.)

Solution 1

Consider using bytearray((255,)) which works the same in Python2 and Python3. In both Python generations the resulting bytearray-object can be converted to a bytes(obj) which is an alias for a str() in Python2 and real bytes() in Python3.

# Python2
>>> x = bytearray((32,33))
>>> x
bytearray(b' !')
>>> bytes(x)
' !'

# Python3
>>> x = bytearray((32,33))
>>> x
bytearray(b' !')
>>> bytes(x)
b' !'

Solution 2

Try the following:

b = bytes([x])

For example:

>>> bytes([255])
b'\xff'

Solution 3

In case you want to write Python 2/3 compatible code, use six.int2byte

Solution 4

Yet another alternative (Python 3.5+):

>>> b'%c' % 65
b'A'

Solution 5

>>> import struct
>>> struct.pack('B', 10)
b'\n'
>>> import functools
>>> bchr = functools.partial(struct.pack, 'B')
>>> bchr(10)
b'\n'

View more solutions

Comments

[zwol]

I get a little twitchy about throwing around scratch arrays but probably I shouldn't. It does the job, anyway.

[malthe]

@Zack: You could use bytes((255, )) as a variation.

[Guido U. Draheim]

bytes((255,)) in Python2 will NOT give you b'\xff' ... it returns '(255,)' instead.

[jfs]

@GuidoDraheim2013: it is Python 3 code i.e., don't use bytes([255]) on Python 2, use chr(255) there.

[zwol]

I don't see why that would be better than Guido's answer, particularly if I have no other need for six.

[youfu]

@zwol: For Python 3.2+, int2byte = operator.methodcaller("to_bytes", 1, "big"). According to the comment, this is about 2x faster than bytes((...)). Anyway, int2byte(x) looks better than bytes(bytearray((x,))) for me.

[zwol]

Speed is good, but in the thing that provoked the original question, no dependencies outside the standard library was an overriding concern.

[Perkins]

bytes takes a tuple for a constructor directly, so you can just use bytes((x,)). Only need to use bytearray if you want it to be mutable.

[zwol]

Annoying that b'{:c}'.format(65) doesn't work as well, but thanks, this could be quite handy for the thing I originally wanted this for (and never got around to finishing).

[martineau]

@malthe: bytes((255, )) is still creates a scratch array.