"OSError: [Errno 22] Invalid argument" when read()ing a huge file

Question

I'm trying to write a small script that prints the checksum of a file (using some code from https://gist.github.com/Zireael-N/ed36997fd1a967d78cb2):

import sys
import os
import hashlib

file = '/Users/Me/Downloads/2017-11-29-raspbian-stretch.img'

with open(file, 'rb') as f:
    contents = f.read()
    print('SHA256 of file is %s' % hashlib.sha256(contents).hexdigest())

But I'm getting the following error message:

Traceback (most recent call last):
  File "checksum.py", line 8, in <module>
    contents = f.read()
OSError: [Errno 22] Invalid argument

What am I doing wrong? I'm using python 3 on macOS High Sierra

Answer 1

There have been several issues over the history of Python (most fixed in recent versions) reading more than 2-4 GB at once from a file handle (an unfixable version of the problem also occurs on 32 bit builds of Python, where they simply lack the virtual address space to allocate the buffer; not I/O related, but seen most frequently slurping large files). A workaround available for hashing is to update the hash in fixed size chunks (which is a good idea anyway, since counting on RAM being greater than file size is a poor idea). The most straightforward approach is to change your code to:

with open(file, 'rb') as f:
    hasher = hashlib.sha256()  # Make empty hasher to update piecemeal
    while True:
        block = f.read(64 * (1 << 20)) # Read 64 MB at a time; big, but not memory busting
        if not block:  # Reached EOF
            break
        hasher.update(block)  # Update with new block
print('SHA256 of file is %s' % hasher.hexdigest())  # Finalize to compute digest

If you're feeling fancy, you can "simplify" the loop using two-arg iter and some functools magic, replacing the whole of the while loop with:

for block in iter(functools.partial(f.read, 64 * (1 << 20)), b''):
    hasher.update(block)

Or on Python 3.8+, with the walrus operator, := it's simpler without the need for imports or unreadable code:

while block := f.read(64 * (1 << 20)):  # Assigns and tests result in conditional!
    hasher.update(block)