2 minute read

In this article we will discuss the difference in reading binary numerical data between SAS and Python.

Byte Order

Integer values for integer binary data are typically stored in one of three sizes: 1 byte, 2 bytes, or 4 bytes. The ordering of the bytes for the integer varies depending on the platform (operating environment) on which the integers were produced. There are two types of byte orders, “big-endian” and “little-endian”. In short, the main difference lies in where most significant byte is stored. For example, if an integer is stored in 4 bytes, then 2 is stored as 10 00 00 00 in little-endian and as 00 00 00 10 in big-endian. In this article we will just focus on the little-endian, since it is used in Windows and Linux.


SAS uses informats to read the binary data and the corresponding formats for writing the binary data. The common informats are IBw.d, PIBw.d.

IBw.d: for integer binary values. w specifies the width of the input fields and d is a scaling factor and specifies the power of 10 that divides the integer.

PIBw.d: for positive integer binary values.

Note that when using infile statement to open a binary file, one has to specify option recfm to value n so that SAS knows the file is a binary file.


Use open(file_name, 'rb') to open the binary file, remember to include b to indicate a binary file. There are two ways commonly used in Python to read binary numerical data. The first one is to use function int.from_bytes(). It is simple and quick. The following example converts the bytes b’\xf8’ to the integer 248.


However, it needs a loop if one is converting a series of bytes. In such case, it is easier to use the built in package struct. The specific function used to read binary data is struct.unpack. Let’s say there is a stream of bytes containing 4 integers each of which is stored in 4 bytes. The following one command can read all 4 integers at the same time.

import struct
struct.unpack('iiii', b'\xf8\x17\x00\x00\x00\x00\x00\x00\x01\x00\x00\x00\xf6\x0b\x00\x00')

It returns a tuple of 4 integers. Here iiii is the format string similar to informats/formats in SAS, which tells Python how the integers are stored. It can also be written as 'i'*4. The following table compares some formats with SAS informats.

Format Python Type Standard Size SAS Informat/Format
b integer 1 IB1.
B integer 1 PIB1.
h integer 2 IB2.
H integer 2 PIB2.
i integer 4 IB4.
I integer 4 PIB4.
q integer 8 IB8.
Q integer 8 PIB8.


We will be using this binary file. Feel free to download the file for your own test. The file has 21 bytes or 9 integers. The first 4 integers are stored in 4 bytes each and the last 5 integers are store in 1 byte each.

  • SAS
data test;
  infile 'binary_integer' recfm=n;
  array y{5} y1-y5;
  input (x1 x2 x3 x4) (ib4.);
  do j = 1 to 5;
    input y{j} pib1.;
  drop j;
  • Python
import struct
with open('binary_inter', 'rb') as f:
    x = f.read() # read all the content
    x1, x2, x3, x4 = struct.unpack('i'*4, x[:16])
    y1, y2, y3, y4, y5 = struct.unpack('b'*5, x[17:])