flier / pyfasthash

Python Non-cryptographic Hash Library
Apache License 2.0
281 stars 52 forks source link

Introduction pypi Travis CI Status codecov

pyhash is a python non-cryptographic hash library.

It provides several common hash algorithms with C/C++ implementation for performance and compatibility.

>>> import pyhash
>>> hasher = pyhash.fnv1_32()

>>> hasher('hello world')
2805756500L

>>> hasher('hello', ' ', 'world')
2805756500L

>>> hasher('world', seed=hasher('hello '))
2805756500L

It also can be used to generate fingerprints without seed.

>>> import pyhash
>>> fp = pyhash.farm_fingerprint_64()

>>> fp('hello')
>>> 13009744463427800296L

>>> fp('hello', 'world')
>>> [13009744463427800296L, 16436542438370751598L]

Notes

hasher('hello', ' ', 'world') is a syntax sugar for hasher('world', seed=hasher(' ', seed=hasher('hello'))), and may not equals to hasher('hello world'), because some hash algorithms use different hash and seed size.

For example, metro hash always use 32bit seed for 64/128 bit hash value.

>>> import pyhash
>>> hasher = pyhash.metro_64()

>>> hasher('hello world')
>>> 5622782129197849471L

>>> hasher('hello', ' ', 'world')
>>> 16402988188088019159L

>>> hasher('world', seed=hasher(' ', seed=hasher('hello')))
>>> 16402988188088019159L

Installation

$ pip install pyhash

Notes

If pip install failed with similar errors, #27

/usr/lib/gcc/x86_64-linux-gnu/6/include/smmintrin.h:846:1: error: inlining failed in call to always_inline 'long long unsigned int _mm_crc32_u64(long long unsigned int, long long unsigned int)': target specific option mismatch
 _mm_crc32_u64 (unsigned long long __C, unsigned long long __V)
 ^~~~~~~~~~~~~
src/smhasher/metrohash64crc.cpp:52:34: note: called from here
             v[0] ^= _mm_crc32_u64(v[0], read_u64(ptr)); ptr += 8;
                     ~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~~~~~

Please upgrade pip and setuptools to latest version and try again

$ pip install --upgrade pip setuptools

Notes

If pip install failed on MacOS with similar errors #28

   creating build/temp.macosx-10.6-intel-3.6
   ...
   /usr/bin/clang -fno-strict-aliasing -Wsign-compare -fno-common -dynamic -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -arch i386 -arch x86_64 -g -c src/smhasher/metrohash64crc.cpp -o build/temp.macosx-10.6-intel-3.6/src/smhasher/metrohash64crc.o -msse4.2 -maes -mavx -mavx2
    src/smhasher/metrohash64crc.cpp:52:21: error: use of undeclared identifier '_mm_crc32_u64'
                v[0] ^= _mm_crc32_u64(v[0], read_u64(ptr)); ptr += 8;
                        ^

You may try to

$ CFLAGS="-mmacosx-version-min=10.13" pip install pyhash

Notes

pyhash only support pypy v6.0 or newer, please download and install the latest pypy.

Algorithms

pyhash supports the following hash algorithms

String and Bytes literals

Python has two types can be used to present string literals, the hash values of the two types are definitely different.

For example,

$ python2
Python 2.7.15 (default, Jun 17 2018, 12:46:58)
[GCC 4.2.1 Compatible Apple LLVM 9.1.0 (clang-902.0.39.2)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyhash
>>> hasher = pyhash.murmur3_32()
>>> hasher('foo')
4138058784L
>>> hasher(u'foo')
2085578581L
>>> hasher(b'foo')
4138058784L
$ python3
Python 3.7.0 (default, Jun 29 2018, 20:13:13)
[Clang 9.1.0 (clang-902.0.39.2)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import pyhash
>>> hasher = pyhash.murmur3_32()
>>> hasher('foo')
2085578581
>>> hasher(u'foo')
2085578581
>>> hasher(b'foo')
4138058784

You can also import unicode_literals to use unicode literals in Python 2.x

from __future__ import unicode_literals

In general, it is more compelling to use unicode_literals when back-porting new or existing Python 3 code to Python 2/3 than when porting existing Python 2 code to 2/3. In the latter case, explicitly marking up all unicode string literals with u'' prefixes would help to avoid unintentionally changing the existing Python 2 API. However, if changing the existing Python 2 API is not a concern, using unicode_literals may speed up the porting process.