在python中读取二进制文件

Question

I wrote a python script to create a binary file of integers.

import struct  
pos = [7623, 3015, 3231, 3829]  
inh = open('test.bin', 'wb')  
for e in pos:  
    inh.write(struct.pack('i', e))  
inh.close()

It worked well, then I tried to read the 'test.bin' file using the below code.

import struct  
inh = open('test.bin', 'rb')  
for rec in inh:  
    pos = struct.unpack('i', rec)  
    print pos  
inh.close()

But it failed with an error message:

Traceback (most recent call last):   
   File "readbinary.py", line 10, in <module>  
   pos = struct.unpack('i', rec)  
   File "/usr/lib/python2.5/struct.py", line 87, in unpack  
   return o.unpack(s)  
struct.error: unpack requires a string argument of length 4

I would like to know how I can read these file using struct.unpack .
Many thanks in advance, Vipin

Answer 1

for rec in inh: reads one line at a time -- not what you want for a binary file. Read 4 bytes at a time (with a while loop and inh.read(4) ) instead (or read everything into memory with a single .read() call, then unpack successive 4-byte slices). The second approach is simplest and most practical as long as the amount of data involved isn't huge:

import struct
with open('test.bin', 'rb') as inh:
    indata = inh.read()
for i in range(0, len(data), 4):
    pos = struct.unpack('i', data[i:i+4])  
    print(pos)  

If you do fear potentially huge amounts of data (which would take more memory than you have available), a simple generator offers an elegant alternative:

import struct
def by4(f):
    rec = 'x'  # placeholder for the `while`
    while rec:
        rec = f.read(4)
        if rec: yield rec           
with open('test.bin', 'rb') as inh:
    for rec in by4(inh):
        pos = struct.unpack('i', rec)  
        print(pos)  

A key advantage to this second approach is that the by4 generator can easily be tweaked (while maintaining the specs: return a binary file's data 4 bytes at a time) to use a different implementation strategy for buffering, all the way to the first approach (read everything then parcel it out) which can be seen as "infinite buffering" and coded:

def by4(f):
    data = inf.read()
    for i in range(0, len(data), 4):
        yield data[i:i+4]

while leaving the "application logic" (what to do with that stream of 4-byte chunks) intact and independent of the I/O layer (which gets encapsulated within the generator).

Answer 2

I think "for rec in inh" is supposed to read 'lines', not bytes. What you want is:

while True:
    rec = inh.read(4) # Or inh.read(struct.calcsize('i'))
    if len(rec) != 4:
        break
    (pos,) = struct.unpack('i', rec)
    print pos

Or as others have mentioned:

while True:
    try:
        (pos,) = struct.unpack_from('i', inh)
    except (some_exception...):
        break

Answer 3

Check the size of the packed integers:

>>> pos
[7623, 3015, 3231, 3829]
>>> [struct.pack('i',e) for e in pos]
['\xc7\x1d\x00\x00', '\xc7\x0b\x00\x00', '\x9f\x0c\x00\x00', '\xf5\x0e\x00\x00']

We see 4-byte strings, it means that reading should be 4 bytes at a time:

>>> inh=open('test.bin','rb')
>>> b1=inh.read(4)
>>> b1
'\xc7\x1d\x00\x00'
>>> struct.unpack('i',b1)
(7623,)
>>> 

This is the original int! Extending into a reading loop is left as an exercise .

Answer 4

You can probably use array as well if you want:

import array  
pos = array.array('i', [7623, 3015, 3231, 3829]) 
inh = open('test.bin', 'wb')  
pos.write(inh)
inh.close()

Then use array.array.fromfile or fromstring to read it back.

Answer 5

This function reads all bytes from file

def read_binary_file(filename):
try:
    f = open(filename, 'rb')
    n = os.path.getsize(filename)
    data = array.array('B')
    data.read(f, n)
    f.close()
    fsize = data.__len__()
    return (fsize, data)

except IOError:
    return (-1, [])

# somewhere in your code
t = read_binary_file(FILENAME)
fsize = t[0]

if (fsize > 0):
    data = t[1]
    # work with data
else:
    print 'Error reading file'

Answer 6

Your iterator isn't reading 4 bytes at a time so I imagine it's rather confused. Like SilentGhost mentioned, it'd probably be best to use unpack_from().