CODEWITHSUNDEEP
pythonbinary-data
abelenky
abelenky

Reputation: 64710

How to loop over a binary file in Python in chunks

I'm trying to use Python to loop over a long binary file filled with 8-byte records.

Each record has the format [ uint16 | uint16 | uint32 ]
(which is "HHI" in struct-formatting)

Apparently each 8-byte block is getting treated as an int, instead of an array of 8-bytes, then causing the struct.unpack call to fail

with open(fname, "rb") as f:
    sz=struct.calcsize("HHI")
    print(sz)                # This shows 8, as expected 
    for raw in f.read(sz):   # Expect this should read 8 bytes into raw
        print(type(raw))     # This says raw is an 'int', not a byte-array
        record=struct.unpack("HHI", raw ) # "TypeError: a bytes-like object is required, not 'int'"
        print(record)

How can I read my file as a series of structures, and print them each out?

Upvotes: 2

Views: 4405

Answers (4)

deeenes
deeenes

Reputation: 4586

You can also do this using the walrus operator (:=), and I find that more concise and readable:

fname = '/tmp/foobar.txt'
size = 2

with open(fname, 'rb') as fp:
    while chunk := fp.read(size):
        print(chunk)

echo 'foobar' > /tmp/foobar.txt

python iter-chunks.py

b'fo'
b'ob'
b'ar'
b'\n'

This implements the solution the OP asked for:

I want the first 8bytes, then iterate to get the next 8, and the following 8, etc, until the full file has been processed

Upvotes: 0

snakecharmerb
snakecharmerb

Reputation: 55844

The iter builtin, if passed a callable and a sentinel value will call the callable repeatedly until the sentinel value is returned.

So you can create a partial function with functools.partial (or use a lambda) and pass it to iter, like this:

with open('foo.bin', 'rb') as f:
    chunker = functools.partial(f.read, 8)
    for chunk in iter(chunker, b''):      # Read 8 byte chunks until empty byte returned
        # Do stuff with chunk

Upvotes: 4

gdlmx
gdlmx

Reputation: 6789

f.read(len) only returns a byte string. Then raw will be a single byte.

The correct way of looping is:

with open(fname, 'rb') as f:
    while True:
        raw = f.read(8)
        if len(raw)!=8:
            break # ignore the incomplete "record" if any
        record = struct.unpack("HHI", raw )
        print(record)

Upvotes: 3

Harvey
Harvey

Reputation: 339

I've never used this before, but it looks like an initialization issue:

   with open(fname, "rb") as f:
        fmt = 'HHI'
        raw=struct.pack(fmt,1,2,3)
        len=struct.calcsize(fmt)
        print(len)               # This shows 8, as expected 
        for raw in f.read(len):  # Expect this should read 8 bytes into raw
            print(type(raw))     # This says raw is an 'int', not a byte-array
            record=struct.unpack(fmt, raw ) # "TypeError: a bytes-like object is required, not 'int'"
            print(record)

You may want to look at iter_unpack() for optimization if you have adequate ram.

Note that in 3.7, the default value changes from bytes to string. see near end of page https://docs.python.org/3/library/struct.html#struct.pack

Upvotes: 0

Related Questions