pycrc32

pycrc32 is a Python module for SIMD-accelerated CRC32 checksum computation.
Big thanks to rust-crc32fast - this project just provides Python bindings for their Rust implementation.

Installation

pip install pycrc32

Usage

from pycrc32 import crc32

data = b"123456789"
print(f"crc32 for {data!r} is {crc32(data)}")

Advanced Checksum Calculation with Hasher

For scenarios that require more flexibility, such as processing large amounts of data or computing the checksum in stages, you can use the Hasher class:

from pycrc32 import Hasher

# Create a new Hasher instance
hasher = Hasher()

# Update the hasher with data chunks
hasher.update(b"123456")
hasher.update(b"789")

# Finalize the computation and get the checksum
checksum = hasher.finalize()
print(f"Checksum: {checksum}")

# Reset the hasher to compute another checksum
hasher.reset()
hasher.update(b"The quick brown fox jumps over the lazy dog")
new_checksum = hasher.finalize()
print(f"New checksum: {new_checksum}")

You can also initialize a Hasher with a specific initial CRC32 state:

initial_crc = 12345678
hasher = Hasher.with_initial(initial_crc)

hasher.update(b"additional data")
final_checksum = hasher.finalize()
print(f"Final checksum with initial state: {final_checksum}")

To combine checksums from different data blocks without needing to concatenate the data, use the combine method:

hasher1 = Hasher()
hasher1.update(b"Data block 1")
checksum1 = hasher1.finalize()

hasher2 = Hasher()
hasher2.update(b"Data block 2")
checksum2 = hasher2.finalize()

# Combine checksums from hasher1 into hasher2
hasher1.combine(hasher2)  # Combine the state of hasher2 into hasher1

# The final checksum after combination
combined_checksum = hasher1.finalize()
print(f"Combined checksum: {combined_checksum}")

Speed

The performance of pycrc32 has been benchmarked on a trusty old Intel i7-8550U using 32MB of random input data. Below is a comparison of the median computation times across different libraries:

Library	Median Time (s)
pycrc32	0.002703
zlib	0.019796
fastcrc	0.071426

We reach almost 10x performance improvements compared to the zlib baseline implementation.
See scripts/bench.py for more details.