uniseg

====== uniseg

A Python package to determine Unicode text segmentations.

• uniseg · PyPI <https://pypi.org/project/uniseg/>_
• emptypage / uniseg-py — Bitbucket <https://bitbucket.org/emptypage/uniseg-py/>_
• uniseg documentation — Read the Docs <https://uniseg-py.readthedocs.io/>_

News

We released the version 0.9.0 on November, 2024, and this is the first release ever which passes all the Unicode breaking tests (congrats!). And now I'm going to make its release number to 1.0, with some breaking changes for the APIs soon. Thank you.

Features

This package provides:

• Functions to get Unicode Character Database (UCD) properties concerned with text segmentations.
• Functions to determine segmentation boundaries of Unicode strings.
• Classes that help implement Unicode-aware text wrapping on both console (monospace) and graphical (monospace / proportional) font environments.

Supporting segmentations are:

code point Code point <https://www.unicode.org/glossary/#code_point>_ is "any value in the Unicode codespace." It is the basic unit for processing Unicode strings.

Historically, units per Unicode string object on elder versions of Python was build-dependent. Some builds uses UTF-16 as an implementation for that and treat each code point greater than U+FFFF as a "surrogate pair", which is a pair of the special two code points. The uniseg package had provided utility functions in order to treat Unicode strings per proper code points on every platform.

Since Python 3.3, The Unicode string is implemented with "flexible string representation", which gives access to full code points and space-efficiency [PEP 393]_. So you don't need to worry about treating complex multi-code-points issue any more. If you want to treat some Unicode string per code point, just iterate that like: for c in s:. So uniseg.codepoint module has been deprecated and deleted.

.. _[PEP 393]: https://peps.python.org/pep-0393/

grapheme cluster Grapheme cluster <https://www.unicode.org/glossary/#grapheme_cluster>_ approximately represents "user-perceived character." They may be made up of single or multiple Unicode code points. e.g. "g̈", "g" + combining diaeresis is a single user-perceived character, while which represents with two code points, U+0067 LATIN SMALL LETTER G and U+0308 COMBINING DIAERESIS.

word break Word boundaries are familiar segmentation in many common text operations. e.g. Unit for text highlighting, cursor jumping etc. Note that words are not determinable only by spaces or punctuations in text in some languages. Such languages like Thai or Japanese require dictionaries to determine appropriate word boundaries. Though the package only provides simple word breaking implementation which is based on the scripts and doesn't use any dictionaries, it also provides ways to customize its default behavior.

sentence break Sentence breaks are also common in text processing but they are more contextual and less formal. The sentence breaking implementation (which is specified in UAX: Unicode Standard Annex) in the package is simple and formal too. But it must be still useful in some usages.

line break Implementing line breaking algorithm is one of the key features of this package. The feature is important in many general text presentations in both CLI and GUI applications.

Requirements

Python 3.9 or later.

Install

.. code:: console

$ pip install uniseg

Changes

0.9.1 (2025-01-16)

• Fix ambiguous_as_wide options are not working on uniseg.wrap.

0.9.0 (2024-11-07)

• Unicode 16.0.0.
• Rule-based grapheme cluster segmentation is back.
• And, this is the first release ever that passes the entire Unicode breaking tests!

0.8.1 (2024-08-13)

• Fix sentence_break('/') raised an exception. (Thanks to Nathaniel Mills)

0.8.0 (2024-02-08)

• Unicode 15.0.0.
• Regex-based grapheme cluster segmentation.
• Quit supporting Python versions < 3.8.

0.7.2 (2022-09-20)

• Improve performance of Unicode lookups. PR by Max Bachmann <https://bitbucket.org/emptypage/uniseg-py/pull-requests/1>_.

0.7.1 (2015-05-02)

• CHANGE: wrap.Wrapper.wrap(): returns the count of lines now.
• Separate LICENSE from README.txt for the packaging-related reason in some environments.

0.7.0 (2015-02-27)

• CHANGE: Quitted gathering all submodules's members on the top, uniseg module.
• CHANGE: Reform uniseg.wrap module and sample scripts.
• Maintained uniseg.wrap module, and sample scripts work again.

0.6.4 (2015-02-10)

• Add uniseg-dbpath console command, which just print the path of ucd.sqlite3.
• Include sample scripts under the package's subdirectory.

0.6.3 (2015-01-25)

• Python 3.4
• Support modern setuptools, pip and wheel.

0.6.2 (2013-06-09)

• Python 3.3

0.6.1 (2013-06-08)

• Unicode 6.2.0

References

• UAX #29: Unicode Text Segmentation (16.0.0) <https://www.unicode.org/reports/tr29/tr29-45.html>_
• UAX #14: Unicode Line Breaking Algorithm (16.0.0) <https://www.unicode.org/reports/tr14/tr14-53.html>_

Related / Similar Projects

PyICU <https://pypi.python.org/pypi/PyICU>_ - Python extension wrapping the ICU C++ API PyICU is a Python extension wrapping International Components for Unicode library (ICU). It also provides text segmentation supports and they just perform richer and faster than those of ours. PyICU is an extension library so it requires ICU dynamic library (binary files) and compiler to build the extension. Our package is written in pure Python; it runs slower but is more portable.

pytextseg <https://pypi.python.org/pypi/pytextseg>_ - Python module for textsegmentation pytextseg package focuses very similar goal to ours; it provides Unicode-aware text wrapping features. They designed and uses their original string class (not built-in unicod / str classes) for the purpose. We use strings as just ordinary built-in unicode / str objects for text processing in our modules.