Skip to main content

Python wrapper for MurmurHash (MurmurHash3), a set of fast and robust hash functions.

Project description


GitHub Super-Linter Build passing PyPi Version Python Versions License: CC0-1.0 Total Downloads Recent Downloads Conda Version

mmh3 is a Python wrapper for MurmurHash (MurmurHash3), a set of fast and robust non-cryptographic hash functions invented by Austin Appleby.

Combined with probabilistic techniques like a Bloom filter, MinHash, and feature hashing, mmh3 allows you to develop high-performance systems in fields such as data mining, machine learning, and natural language processing.

How to use


pip install mmh3 # for macOS, use "pip3 install mmh3" and python3


>>> import mmh3
>>> mmh3.hash("foo") # returns a 32-bit signed int
>>> mmh3.hash("foo", 42) # uses 42 as a seed
>>> mmh3.hash("foo", signed=False) # returns a 32-bit unsigned int

Other functions:

>>> mmh3.hash64("foo") # two 64 bit signed ints (by using the 128-bit algorithm as its backend)
(-2129773440516405919, 9128664383759220103)
>>> mmh3.hash64("foo", signed=False) #  two 64 bit unsigned ints
(16316970633193145697, 9128664383759220103)
>>> mmh3.hash128("foo", 42) # 128 bit unsigned int
>>> mmh3.hash128("foo", 42, signed=True) # 128 bit signed int
>>> mmh3.hash_bytes("foo") # 128 bit value as bytes
>>> import numpy as np
>>> a = np.zeros(2 ** 32, dtype=np.int8)
>>> mmh3.hash_bytes(a)

Beware that hash64 returns two values, because it uses the 128-bit version of MurmurHash3 as its backend.

hash_from_buffer hashes byte-likes without memory copying. The method is suitable when you hash a large memory-view such as numpy.ndarray.

>>> mmh3.hash_from_buffer(numpy.random.rand(100))
>>> mmh3.hash_from_buffer(numpy.random.rand(100), signed=False)

hash64, hash128, and hash_bytes have the third argument for architecture optimization. Use True for x64 and False for x86 (default: True):

>>> mmh3.hash64("foo", 42, True) 
(-840311307571801102, -6739155424061121879)


3.0.0 (2021-02-23)

  • Python wheels are now available, thanks to the power of cibuildwheel.
    • Supported platforms are manylinux1_x86_64, manylinux2010_x86_64, manylinux2014_aarch64, win32, win_amd64, macosx_10_9_x86_64, and macosx_11_0_arm64 (Apple Silicon).
  • Add support for newer macOS environments. Thanks Matthew Honnibal!
  • Drop support for Python 2.7, 3.3, 3.4, and 3.5.
  • Add support for Python 3.7, 3.8, and 3.9.
  • Migrate Travis CI and AppVeyor to GitHub Actions.

2.5.1 (2017-10-31)

  • Bug fix for hash_bytes. Thanks doozr!

2.5 (2017-10-28)

  • Add hash_from_buffer. Thanks Dimitri Vorona!
  • Add a keyword argument signed.

2.4 (2017-05-27)

  • Support seeds with 32-bit unsigned integers; thanks Alexander Maznev!
  • Support 64-bit data (under 64-bit environments)
  • Fix compile errors for Python 3.6 under Windows systems.
  • Add unit testing and continuous integration with Travis CI and AppVeyor.

2.3.2 (2017-05-26)

  • Relicensed from public domain to CC0-1.0.

2.3.1 (2015-06-07)

  • Fix compile errors for gcc >=5.

2.3 (2013-12-08)

  • Add hash128, which returns a 128-bit signed integer.
  • Fix a misplaced operator which could cause memory leak in a rare condition.
  • Fix a malformed value to a Python/C API function which may cause runtime errors in recent Python 3.x versions.

The first two commits are from Derek Wilson. Thanks!

2.2 (2013-03-03)

  • Improve portability to support systems with old gcc (version < 4.4) such as CentOS/RHEL 5.x. (Commit from Micha Gorelick. Thanks!)

2.1 (2013-02-25)

  • Add __version__ constant. Check if it exists when the following revision matters for your application.
  • Incorporate the revision r147, which includes robustness improvement and minor tweaks.

Beware that due to this revision, the result of 32-bit version of 2.1 is NOT the same as that of 2.0. E.g.,:

>>> mmh3.hash("foo") # in mmh3 2.0
>>> mmh3.hash("foo") # in mmh3 2.1

The results of hash64 and hash_bytes remain unchanged. Austin Appleby, the author of Murmurhash, ensured this revision was the final modification to MurmurHash3's results and any future changes would be to improve performance only.



Known Issues

Getting different results from other MurmurHash3-based libraries

By default, mmh3 returns signed values for 32-bit and 64-bit versions and unsigned values for hash128, due to historical reasons. Please use the keyword argument signed to obtain a desired result.

For compatibility with Google Guava (Java), see

Unexpected results when given non 32-bit seeds

Version 2.4 changed the type of seeds from signed 32-bit int to unsigned 32-bit int. The resulting values with signed seeds still remain the same as before, as long as they are 32-bit.

>>> mmh3.hash("aaaa", -1756908916) # signed representation for 0x9747b28c
>>> mmh3.hash("aaaa", 2538058380) # unsigned representation for 0x9747b28c

Be careful so that these seeds do not exceed 32-bit. Unexpected results may happen with invalid values.

>>> mmh3.hash("foo", 2 ** 33)
>>> mmh3.hash("foo", 2 ** 34)


MurmurHash3 was originally developed by Austin Appleby and distributed under public domain.

Ported and modified for Python by Hajime Senuma.

See also


The following textbooks and tutorials are great sources to learn how to use mmh3 (and other hash algorithms in general) for high-performance computing.

Similar libraries

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mmh3-3.0.0.tar.gz (10.4 kB view hashes)

Uploaded Source

Built Distributions

mmh3-3.0.0-cp39-cp39-win_amd64.whl (15.2 kB view hashes)

Uploaded CPython 3.9 Windows x86-64

mmh3-3.0.0-cp39-cp39-win32.whl (14.7 kB view hashes)

Uploaded CPython 3.9 Windows x86

mmh3-3.0.0-cp39-cp39-manylinux2014_aarch64.whl (49.9 kB view hashes)

Uploaded CPython 3.9

mmh3-3.0.0-cp39-cp39-manylinux2010_x86_64.whl (49.3 kB view hashes)

Uploaded CPython 3.9 manylinux: glibc 2.12+ x86-64

mmh3-3.0.0-cp39-cp39-manylinux1_x86_64.whl (49.3 kB view hashes)

Uploaded CPython 3.9

mmh3-3.0.0-cp39-cp39-macosx_11_0_arm64.whl (13.2 kB view hashes)

Uploaded CPython 3.9 macOS 11.0+ ARM64

mmh3-3.0.0-cp39-cp39-macosx_10_9_x86_64.whl (12.6 kB view hashes)

Uploaded CPython 3.9 macOS 10.9+ x86-64

mmh3-3.0.0-cp38-cp38-win_amd64.whl (15.3 kB view hashes)

Uploaded CPython 3.8 Windows x86-64

mmh3-3.0.0-cp38-cp38-win32.whl (14.8 kB view hashes)

Uploaded CPython 3.8 Windows x86

mmh3-3.0.0-cp38-cp38-manylinux2014_aarch64.whl (50.7 kB view hashes)

Uploaded CPython 3.8

mmh3-3.0.0-cp38-cp38-manylinux2010_x86_64.whl (50.0 kB view hashes)

Uploaded CPython 3.8 manylinux: glibc 2.12+ x86-64

mmh3-3.0.0-cp38-cp38-manylinux1_x86_64.whl (50.0 kB view hashes)

Uploaded CPython 3.8

mmh3-3.0.0-cp38-cp38-macosx_10_9_x86_64.whl (12.6 kB view hashes)

Uploaded CPython 3.8 macOS 10.9+ x86-64

mmh3-3.0.0-cp37-cp37m-win_amd64.whl (15.2 kB view hashes)

Uploaded CPython 3.7m Windows x86-64

mmh3-3.0.0-cp37-cp37m-win32.whl (14.7 kB view hashes)

Uploaded CPython 3.7m Windows x86

mmh3-3.0.0-cp37-cp37m-manylinux2014_aarch64.whl (51.6 kB view hashes)

Uploaded CPython 3.7m

mmh3-3.0.0-cp37-cp37m-manylinux2010_x86_64.whl (50.9 kB view hashes)

Uploaded CPython 3.7m manylinux: glibc 2.12+ x86-64

mmh3-3.0.0-cp37-cp37m-manylinux1_x86_64.whl (50.9 kB view hashes)

Uploaded CPython 3.7m

mmh3-3.0.0-cp37-cp37m-macosx_10_9_x86_64.whl (12.5 kB view hashes)

Uploaded CPython 3.7m macOS 10.9+ x86-64

mmh3-3.0.0-cp36-cp36m-win_amd64.whl (15.2 kB view hashes)

Uploaded CPython 3.6m Windows x86-64

mmh3-3.0.0-cp36-cp36m-win32.whl (14.7 kB view hashes)

Uploaded CPython 3.6m Windows x86

mmh3-3.0.0-cp36-cp36m-manylinux2014_aarch64.whl (49.9 kB view hashes)

Uploaded CPython 3.6m

mmh3-3.0.0-cp36-cp36m-manylinux2010_x86_64.whl (49.2 kB view hashes)

Uploaded CPython 3.6m manylinux: glibc 2.12+ x86-64

mmh3-3.0.0-cp36-cp36m-manylinux1_x86_64.whl (49.2 kB view hashes)

Uploaded CPython 3.6m

mmh3-3.0.0-cp36-cp36m-macosx_10_9_x86_64.whl (12.5 kB view hashes)

Uploaded CPython 3.6m macOS 10.9+ x86-64

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page