1.1K ·GitHub
What is this?
Static, memory-efficient trie-like data structures for Python, built on the marisa-trie C++ library. Strings stored inside can take up to 50x-100x less memory than in a standard Python dict, with comparable lookup speed and fast prefix search.
Static, memory-efficient trie-like data structures for Python, built on the marisa-trie C++ library. Strings stored inside can take up to 50x-100x less memory than in a standard Python dict, with comparable lookup speed and fast prefix search.Is this a library I add to my Python project, or a separate tool?A pip-installable library you import into Python code. It provides Cython-based bindings as an alternative to the official SWIG bindings that ship with the C++ library distribution.
A pip-installable library you import into Python code. It provides Cython-based bindings as an alternative to the official SWIG bindings that ship with the C++ library distribution.Which Python versions does it support?Python 3.9 or newer; the GitHub topics also list PyPy3 support.
Python 3.9 or newer; the GitHub topics also list PyPy3 support.What are the current limitations?read() and write() only work with real files, not file-like objects — use pickling there instead. The .prefixes() method of BytesTrie and RecordTrie is quite slow with no iterator counterpart, and the trie classes expose keys() and items() but no values(). Mingw32 builds are also untested.
read() and write() only work with real files, not file-like objects — use pickling there instead. The .prefixes() method of BytesTrie and RecordTrie is quite slow with no iterator counterpart, and the trie classes expose keys() and items() but no values(). Mingw32 builds are also untested.Is the license straightforward for commercial use?Mostly: the wrapper code is MIT-licensed, but the bundled marisa-trie C++ library is dual-licensed under LGPL or BSD 2-clause, so check the LGPL terms if you redistribute the C++ part.
Mostly: the wrapper code is MIT-licensed, but the bundled marisa-trie C++ library is dual-licensed under LGPL or BSD 2-clause, so check the LGPL terms if you redistribute the C++ part.What kinds of tries does the package provide?The main documented class is StringTrie for immutable str-to-str mappings, implemented as two internal tries plus a compact ID mapping table. Trie, BytesTrie and RecordTrie classes are also available.
The main documented class is StringTrie for immutable str-to-str mappings, implemented as two internal tries plus a compact ID mapping table. Trie, BytesTrie and RecordTrie classes are also available.The structures are described as 'static' — does that mean read-only?Yes — you build the trie once and then only query it. StringTrie is explicitly for immutable str-to-str mappings, so you trade updatability for the compact memory footprint and fast lookups.
Yes — you build the trie once and then only query it. StringTrie is explicitly for immutable str-to-str mappings, so you trade updatability for the compact memory footprint and fast lookups.What is MARISA?It is the C++ library underneath this package: marisa-trie implements the compact trie structures and memory layout, while this repo adds the Cython-based Python bindings and publishes them on PyPI.
It is the C++ library underneath this package: marisa-trie implements the compact trie structures and memory layout, while this repo adds the Cython-based Python bindings and publishes them on PyPI.What are similar projects?pydantic, C-Plus-Plus, cython, leetcode, algorithms, annoy, CtCI-6th-Edition, JavaScript
pydantic, C-Plus-Plus, cython, leetcode, algorithms, annoy, CtCI-6th-Edition, JavaScriptWhich topics is it in?cython, pypi, text-processing, compression, python3, low-level, data-structures, algorithms, search, computer-science, cpp, c++
cython, pypi, text-processing, compression, python3, low-level, data-structures, algorithms, search, computer-science, cpp, c++