pyahocorasick 1.4.0

Download and source code

You can fetch pyahocorasick from:

• GitHub https://github.com/WojciechMula/pyahocorasick/

• Pypi https://pypi-hypernode.com/pypi/pyahocorasick/

• Conda-Forge https://github.com/conda-forge/pyahocorasick-feedstock/

Quick start

This module is written in C. You need a C compiler installed to compile native CPython extensions. To install:

pip install pyahocorasick

Then create an Automaton:

>>> import ahocorasick
>>> A = ahocorasick.Automaton()

You can use the Automaton class as a trie. Add some string keys and their associated value to this trie. Here we associate a tuple of (insertion index, original string) as a value to each key string we add to the trie:

>>> for idx, key in enumerate('he her hers she'.split()):
...   A.add_word(key, (idx, key))

Then check if some string exists in the trie:

>>> 'he' in A
True
>>> 'HER' in A
False

And play with the get() dict-like method:

>>> A.get('he')
(0, 'he')
>>> A.get('she')
(3, 'she')
>>> A.get('cat', 'not exists')
'not exists'
>>> A.get('dog')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
KeyError

Now convert the trie to an Aho-Corasick automaton to enable Aho-Corasick search:

>>> A.make_automaton()

Then search all occurrences of the keys (the needles) in an input string (our haystack).

Here we print the results and just check that they are correct. The Automaton.iter() method return the results as two-tuples of the end index where a trie key was found in the input string and the associated value for this key. Here we had stored as values a tuple with the original string and its trie insertion order:

>>> for end_index, (insert_order, original_value) in A.iter(haystack):
...     start_index = end_index - len(original_value) + 1
...     print((start_index, end_index, (insert_order, original_value)))
...     assert haystack[start_index:start_index + len(original_value)] == original_value
...
(1, 2, (0, 'he'))
(1, 3, (1, 'her'))
(1, 4, (2, 'hers'))
(4, 6, (3, 'she'))
(5, 6, (0, 'he'))

You can also create an eventually large automaton ahead of time and pickle it to re-load later. Here we just pickle to a string. You would typically pickle to a file instead:

>>> import cPickle
>>> pickled = cPickle.dumps(A)
>>> B = cPickle.loads(pickled)
>>> B.get('he')
(0, 'he')

See also:

• FAQ and Who is using pyahocorasick? https://github.com/WojciechMula/pyahocorasick/wiki/FAQ#who-is-using-pyahocorasick

Documentation

The full documentation including the API overview and reference is published on readthedocs.

Overview

With an Aho-Corasick automaton you can efficiently search all occurrences of multiple strings (the needles) in an input string (the haystack) making a single pass over the input string. With pyahocorasick you can eventually build large automatons and pickle them to reuse them over and over as an indexed structure for fast multi pattern string matching.

One of the advantages of an Aho-Corasick automaton is that the typical worst-case and best-case runtimes are about the same and depends primarily on the size of the input string and secondarily on the number of matches returned. While this may not be the fastest string search algorithm in all cases, it can search for multiple strings at once and its runtime guarantees make it rather unique. Because pyahocorasick is based on a Trie, it stores redundant keys prefixes only once using memory efficiently.

A drawback is that it needs to be constructed and “finalized” ahead of time before you can search strings. In several applications where you search for several pre-defined “needles” in a variable “haystacks” this is actually an advantage.

Aho-Corasick automatons are commonly used for fast multi-pattern matching in intrusion detection systems (such as snort), anti-viruses and many other applications that need fast matching against a pre-defined set of string keys.

Internally an Aho-Corasick automaton is typically based on a Trie with extra data for failure links and an implementation of the Aho-Corasick search procedure.

Behind the scenes the pyahocorasick Python library implements these two data structures: a Trie and an Aho-Corasick string matching automaton. Both are exposed through the Automaton class.

In addition to Trie-like and Aho-Corasick methods and data structures, pyahocorasick also implements dict-like methods: The pyahocorasick Automaton is a Trie a dict-like structure indexed by string keys each associated with a value object. You can use this to retrieve an associated value in a time proportional to a string key length.

pyahocorasick is available in two flavors:

• a CPython C-based extension, compatible with Python 2 and 3.

• a simpler pure Python module, compatible with Python 2 and 3. This is only available in the source repository (not on Pypi) under the py/ directory and has a slightly different API.

Build and install from PyPi

To install for common operating systems, use pip. Pre-built wheels should be available on Pypi at some point in the future:

pip install pyahocorasick

To build from sources you need to have a C compiler installed and configured which should be standard on Linux and easy to get on MacOSX.

On Windows and Python 2.7 you need the Microsoft Visual C++ Compiler for Python 2.7 (aka. Visual Studio 2008). There have been reports that pyahocorasick does not build yet with MinGW. It may build with cygwin but this has not been tested. If you get this working with these platforms, please report in a ticket!

To build from sources, clone the git repository or download and extract the source archive.

Install pip (and its setuptools companion) and then run (in a virtualenv of course!):

pip install .

If compilation succeeds, the module is ready to use.

Support

Support is available through the GitHub issue tracker to report bugs or ask questions.

Contributing

You can submit contributions through GitHub pull requests.

Authors

The initial author and maintainer is Wojciech Muła. Philippe Ombredanne, the current co-owner, rewrote documentation, setup CI servers and did a whole lot of work to make this module better accessible to end users.

Alphabetic list of authors:

• Andrew Grigorev

• Bogdan

• David Woakes

• Edward Betts

• Frankie Robertson

• Frederik Petersen

• gladtosee

• INADA Naoki

• Jan Fan

• Pastafarianist

• Philippe Ombredanne

• Renat Nasyrov

• Sylvain Zimmer

• Xiaopeng Xu

This library would not be possible without help of many people, who contributed in various ways. They created pull requests, reported bugs as GitHub issues or via direct messages, proposed fixes, or spent their valuable time on testing.

Thank you.

License

This library is licensed under very liberal BSD-3-Clause license. Some portions of the code are dedicated to the public domain such as the pure Python automaton and test code.

Full text of license is available in LICENSE file.

Other Aho-Corasick implementations for Python you can consider

While pyahocorasick tries to be the finest and fastest Aho Corasick library for Python you may consider these other libraries:

• py_aho_corasick by Jan

• Written in pure Python.

• Poor performance.

• ahocorapy by abusix

• Written in pure Python.

• Better performance than py-aho-corasick.

• Using pypy, ahocorapy’s search performance is only slightly worse than pyahocorasick’s.

• Performs additional suffix shortcutting (more setup overhead, less search overhead for suffix lookups).

• Includes visualization tool for resulting automaton (using pygraphviz).

• MIT-licensed, 100% test coverage, tested on all major python versions (+ pypy)

• noaho by Jeff Donner

• Written in C. Does not return overlapping matches.

• Does not compile on Windows (July 2016).

• No support for the pickle protocol.

• acora by Stefan Behnel

• Written in Cython.

• Large automaton may take a long time to build (July 2016)

• No support for a dict-like protocol to associate a value to a string key.

• ahocorasick by Danny Yoo

• Written in C.

• seems unmaintained (last update in 2005).

• GPL-licensed.