_ _
___| |__ _____ _(_)_ __ __ _
/ __| '_ \ / _ \ \ /\ / / | '_ \ / _` |
| (__| | | | __/\ V V /| | | | | (_| |
\___|_| |_|\___| \_/\_/ |_|_| |_|\__, |
|___/
http://chewing.im/
The Chewing (酷音) is an intelligent phonetic (Zhuyin/Bopomofo) input method, one of the most popular choices for Traditional Chinese users. Chewing was inspired by other proprietary intelligent Zhuyin input methods under Microsoft Windows, namely, Wang-Xin by Eten, Microsoft New Zhuyin, and Nature Zhuyin.
The Chewing core team extended their work and actively maintains the project as full open source efforts.
- Website: http://chewing.im/
- Issue tracker: https://github.com/chewing/libchewing/issues
- Mailing lists:
- Development: https://groups.google.com/group/chewing-devel
- General: https://groups.google.com/group/chewing
- Matrix:
- Development: https://matrix.to/#/#libchewing:matrix.org
- General: https://matrix.to/#/#chewing-users:matrix.org
- Build Status:
- Travis-Ci:
- Coverity Scan:
- Coveralls:
- Travis-Ci:
libchewing is derived from the original Chewing input method, a xcin module focusing on intelligent phonetic (Bopomofo/Zhuyin) processing by Lu-chuan Kung (lckung) and Kang-pen Chen (kpchen) sponsored by Tsan-sheng Hsu from Academia Sinica during 1999-2001. However, the original authors of Chewing dropped its development, and Chewing was highly coupled with xcin, which prevents from comprehensive applications. There was a similar input method, bimsphone, which was bundled in XCIN server. However, it did not provide a convenient API for further development, either.
Jim Huang, et al. formed the Chewing core team and extended Kung and Chen's work since 2002. The chewing core team renamed the project as "new" chewing to differentiate their work from the original. Nevertheless, the English name has remained "chewing", which is identified by various input method framework as well.
In 2004, Chewing core team successfully ported the input method to several platforms and framework along with community collaboration.
- Split logic and view.
- Support multiple operating systems, and input framework.
- Provide a universal API for input framework and further development.
Chewing has been adopted by various input frameworks in Unix-like systems and even Microsoft Windows. On these systems, the chewing package is usually split into two parts: libchewing, which handles the actual character selection logic; and input framework interface for display and preference setting.
- The active integration:
- JMCCE
- SCIM
- standalone Microsoft Windows 32/64-bit (windows-chewing)
- Text Services Framework for Microsoft Windows
- UIM
- ibus
- UCIMF
- mozc
- gcin/HIME
- fcitx
- The inactive one:
- xcin
- IIIMF
- MacOS X (SpaceChewing)
- Sun's Java Desktop System Input Method Framework
- OpenVanilla Input Method Framework (previous than version 1.0)
- OXIM
- Hsu
- IBM
- Gin-Yieh
- Eten
- Eten 26 keys
- Dvorak
- Dvorak Hsu
- HanYu PinYin
- Taiwan Huayu Luomapinyin
- MPS2 Pinyin
libchewing provides straightforward API and design, which enables third-party projects to deploy the innovative. Here are the examples:
-
FreeArray: utilizing libchewing for selecting phrases of Array input method. https://github.com/shaform/libfreearray https://github.com/shaform/ibus-freearray
-
CitC (Chewing in the Cloud): extending cloud services for real-time training and phrase selection based on Chewing IM.
-
NTNU-Master Input Method: A Faster Chinese IM based on windows-chewing
The following tools are used to build libchewing. Not all tools are necessary during building. For example, if the compiler you used is clang, gcc & Visual Studio are not needed. The versions listed here is the minimal version known to build libchewing. If any tools you use below this version, libchewing might not be built.
- Build tools:
- autoconf >= 2.65
- automake >= 1.11.6
- libtool >= 2.4.2
- cmake >= 2.8.8 (optional)
- Toolchain / IDE:
- clang >= 3.2
- gcc >= 4.6.3
- Visual Studio Express 2012
- Documentation tools:
- texinfo >= 4.12
libchewing uses the following CI services:
- https://travis-ci.org/chewing/libchewing (testing)
- https://drone.io/github.com/chewing/libchewing (testing)
- https://coveralls.io/r/chewing/libchewing (coverage report)
- https://launchpad.net/~chewing/+archive/chewing (ubuntu PPA)
# ./configure --prefix=/usr
(If you checkout from GIT, make sure running ./autogen.sh
before this.)
# make
# make install
For macOS:
# brew install libchewing # latest release version
# brew install --HEAD libchewing # development, git master branch
see "INSTALL" for details.
Example cross-build instructions:
# CC_FOR_BUILD=gcc ./configure \
--host=arm-none-linux-gnueabi \
--disable-shared --enable-static
# make
To build libchewing on Windows, you need to setup MinGW and MSYS in your system. The installer of MinGW and MSYS is in the following link:
https://sourceforge.net/projects/mingw/files/Installer/mingw-get-inst/
In "Select Components" during installing, please select the following items:
- MinGW Compiler Suite -> C Compiler
- MSYS Basic System
After installing, execute \msys\1.0\msys.bat (default is C:\MinGW\msys\1.0\msys.bat) to enter MSYS shell.
If you get the source from the git repository you need additional step to make the source buildable; use the following commands to install necessary packages.
- mingw-get install automake
- mingw-get install autoconf
- mingw-get install libtool
Now you have the build environment for libchewing. However, you need to check the line end of source code is LF instead of CR/LF before running autogen.sh. The easily way to do this is using git:
- git config core.autocrlf input
- rm -rf *
- git reset --hard
Now you can run the following commands in MSYS shell to build libchewing:
If you get the source from the git repository, run:
- autogen.sh
Then
- configure
- make
If you run testchewing.exe (available after `make check') you will find that testchewing.exe cannot print the correct string. This is because testchewing.exe prints the UTF-8 string, while Windows cannot print it to console.
libchewing supports cmake (https://www.cmake.org/) build system. You can use the following command to build with cmake:
- cmake .
- make
cmake is also the preferred way to build libchewing on Windows platform due to better Windows integrated. You can use the following command to create Visual Studio project in 32-bits Windows platform:
- cmake . -G "Visual Studio 11"
or the following command to create Visual Studio project in 64-bits Windows platform:
- cmake . -G "Visual Studio 11 Win64" (64-bits Windows)
To build libchewing on OS X, you will need tools listed in the requirement. Since OS X does not ship with those tools, building those tools from source could be a tricky task.
A simple way to install those tools is by using Homebrew, a package manager for OS X. You can learn more about Homebrew or see the installation instruction from
Once Homebrew is installed, run the following commands to install the tools you need:
# brew install autoconf automake
# brew install libtool
# brew install cmake
# brew install texinfo
If you get the source from the git repository, run:
# ./autogen.sh
Because OS X uses an older version of makeinfo, you have to set MAKEINFO
manually to where Homebrew installed makeinfo. For example:
# ./configure MAKEINFO=/usr/local/Cellar/texinfo/5.2/bin/makeinfo
then
# make
Because OS X uses an older version of makeinfo, you have to set MAKEINFO
manually to where Homebrew installed makeinfo. For example:
# cmake -DMAKEINFO=/usr/local/Cellar/texinfo/5.2/bin/makeinfo .
then
# make
( modified from https://code.google.com/p/ibus/wiki/ChewingUserGuide )
Chewing guides the user to input Chinese by its pronunciation, in the form of either Bopomofo/Zhuyin or Hanyu pinyin, as well as Chinese punctuation marks and normal and full shape number and English alphabets.
The following sections assume you are using the default configuration, that is, with default/DaChen Bopomofo keyboard layout, on an en_US keyboard, and default key-binding.
Preedit buffer: The places for storing your typing before sending to the applications (such as firefox) you are using.
Mode: Mode determines how Chewing reacts on keyboard input.
This mode is for normal Chinese character and punctuation typing. You are more likely working on this mode.
In this mode, alpha-numberic and punctuation marks are interpreted as either Bopomofo symbols or punctuation marks. If the symbols can form Chinese characters, it will choose the most suitable character according to the context in preedit buffer.
It is recommended that you enter the whole Chinese sentences, so Chewing can do auto-correction for you. If you like what you see, you can press Enter to commit the characters in preedit buffer.
If something goes wrong, you can select the character by move your cursor using {Left} or {Right}, then press {Down} to enter Candidate Selection mode to choose words.
To prevent the auto-correction on certain phrase, you may press {Tab} to break the auto-correction on the end of the sentence.
You can also remember the 2,3,4 words phrases by pressing {Ctrl-2}, {Ctrl-3}, and {Ctrl-4} on the end of the phrase.
Also note that Shift's behavior changes in this mode. Shift-(alpha number) outputs corresponding Full shape Chinese symbols if "Easy symbol input" is enabled; Or outputs corresponding number half shape lowercase English alphabets if "Easy symbol input" is disabled.
Talking about inputing Chinese symbols, other then enable "Easy symbol input" mode, you can also press {Ctrl-1} or {`} to open up a symbol selection dialog, select the category, then use {Down} key to choose the symbols as you would do for characters.
Key binding API name Functionality
----------- -------- -------------
Caps Lock chewing_handle_Capslock Toggle Temporary English sub-mode
Down chewing_handle_Down Enter Candidate Selection mode
Shift-Space chewing_handle_ShiftSpace Toggle Half/Full Shape sub-mode
Enter chewing_handle_Enter Commit the content in preedit buffer
to active application window
Tab chewing_handle_Tab Break the auto-correction.
Ctrl-1 chewing_handle_CtrlNum Open symbol selection dialog
Ctrl-2 chewing_handle_CtrlNum Remember 2-word phrase.
Ctrl-3 chewing_handle_CtrlNum Remember 3-word phrase.
Ctrl-4 chewing_handle_CtrlNum Remember 4-word phrase.
This sub-mode is for inputing half shape and full shape characters. Half shape characters are essentially normal English characters; while the full shape characters are full-width pretty symbols that resemble English characters.
Key binding API name Functionality
----------- -------- -------------
Shift-Space chewing_handle_ShiftSpace Toggle Half/Full Shape sub-mode
This sub-mode is for temporary English inputting.
Key binding API name Functionality
----------- -------- -------------
Caps Lock chewing_handle_Capslock Toggle Temporary English sub-mode
This mode is for Choosing the candidate. Firstly, the longest phrases that fit the pronunciation are shown, then the second shortest, till the single characters. Pressing {Down} again to return to the longest phrases.
For example, after entering "w91o3g4". Pressing {Down} makes Chewing displays 3-word candidate "台北市"; pressing {Down} again makes Chewing displays 2-word candidate "北市"; pressing {Down} again makes Chewing displays 1-word candidates "市","是"; pressing {Down} again makes Chewing displays back the 3-word candidates "台北市";
Key binding API name Functionality
----------- -------- -------------
Down chewing_handle_Down Next bunch of candidates in
different length
Left chewing_handle_Left Previous page of candidates
Right chewing_handle_Right Next page of candidates
1, 2, ...0 chewing_handle_Default Select 1st, 2nd, ... 10th candidate
This mode is on whenever the preedit buffer is empty. This allows you to use movement keys (such as cursor keys and page up/page down) and popular key binding (Ctrl-A, Ctrl-S).
For the brief usage of libchewing APIs, please check the simplified example implemented in file contrib/simple-select.c
Except the following source code:
-
thirdparty/sqlite-amalgamation/ contains sqlite3 source which is in public domain. See https://www.sqlite.org/copyright.html for more information.
-
cmake/FindCurses.cmake is modified from CMake source, which is licensed under BSD 3-Clause.
All source code are licensed under GNU LGPL v2.1 (Lesser General Public License v2.1). See "COPYING" for details.
See "AUTHORS" for details.