I want to pack a 10-bit integer and a 54-bit integer into a 64-bit structure. But this code ends up using 128 bits.
from ctypes import *
class ABC(Structure):
_fields_ = [("a", c_int, 10), ("b", c_uint64, 54)]
print(sizeof(ABC) * 8) # 128
-
Do you consider only unsigned integers?Maciej Gol– Maciej Gol2015-10-04 17:23:44 +00:00Commented Oct 4, 2015 at 17:23
-
No, signed and unsigned. But if you have a solution for unsigned only, I would like to see it.mcu– mcu2015-10-04 17:25:53 +00:00Commented Oct 4, 2015 at 17:25
Add a comment
|
2 Answers
For whatever reason, ctypes doesn't do bit packing properly when you mix different width objects into the same byte. Something like this appears to work for me:
class ABC(Structure):
_pack_ = 1
_fields_ = [("a",c_int64,10),("b",c_uint64,54)]
4 Comments
mcu
What exactly does the
_pack_ attribute do and what are valid values for it?
Mostly Harmless
stackoverflow.com/questions/14771150/… - in essence it emulates C's
pragma pack, but since your structure is byte aligned to a standard width already it probably affects nothing hereMostly Harmless
As for valid values, docs.python.org/2/library/ctypes.html#ctypes.Structure._pack_ and other comments while searching around indicate that it'll take small positive powers of 2
mcu
I think this would apply to Python. msdn.microsoft.com/en-us/library/2e70t5y1.aspx Specifies the value, in bytes, to be used for packing. ... Valid values are 1, 2, 4, 8, and 16. The alignment of a member will be on a boundary that is either a multiple of n or a multiple of the size of the member, whichever is smaller.
Considering int_10 and int_54 are integers with desired bitness:
>>> def pack(int_10, int_54):
... return (int_10 & 0x3FF) | (int_54 & 0x3FFFFFFFFFFFFF) << 10
>>> bin(pack(1, 1))
'0b10000000001'
>>> bin(pack(1, 2 ** 54 - 1))
'0b1111111111111111111111111111111111111111111111111111110000000001'
>>> bin(pack(2 ** 10, 2 ** 54))
'0b0'
You can then pack the resulting integer using struct.pack to get a byte string.
>>> import struct
>>> struct.pack('Q', pack(2 ** 10 - 1, 2 ** 54 - 1))
'\xff\xff\xff\xff\xff\xff\xff\xff'
1 Comment
mcu
This is ok for compacting data for storage and such. But it won't let me manipulate the bit fields easily, like:
x=ABC(10,20);x.a=-15;x.b=50 I am really looking for Python's built-in functionality, if it exists.