Bram Moolenaar wrote:
> Please have a look and give me suggestions for improvement. Especially check
> if something is unclear or incorrect.
How about this?
Takuhiro Nishioka mailto:takuhiro@...
*multibyte.txt* For Vim version 5.6. Last change: 2000 Apr 25
VIM REFERENCE MANUAL by Bram Moolenaar et al.
Multi-byte support *multibyte* *multi-byte*
*Chinese* *Japanese* *Korean*
There are languages which have many characters that can not be represented
using one byte (one octet). These are Chinese (simplified or traditional),
Japanese and Korean. These languages uses more than one byte to represent a
This is limited information on the support in Vim to edit files that use more
than one byte per character. Actually, only two-byte codes are currently
Also see |+multi_byte| and |'fileencoding'|.
1. Introduction |multibyte-intro|
2. Compiling |multibyte-compiling|
3. Display (X fontset support) |multibyte-display|
4. Input (XIM support) |multibyte-input|
5. UTF-8 in XFree86 xterm |UTF8-xterm|
1. Introduction *multibyte-intro*
There are a number of languages in the world. And there are different
cultures and environments at least as much as the number of languages. A
linguistic environment corresponding to an area is called "|locale|". The
POSIX standard defines a concept of |locale|, which includes a lot of
information about |charset|, collating order for sorting, date format,
currency format and so on.
Your system need to support the |locale| system and the language |locale| of
your choice. Some system has a few language |locale|s, so the |locale| of the
language which you want to use may not be on your system. If so, you have to
add the language |locale|. But on some systems, it is not possible to add
other |locale|s. In this case, install X |locale|s by installing X compiled
with X_LOCALE. Add "-DX_LOCALE" to the CFLAGS if your X lib support X_LOCALE.
For example, When you are using Linux system and you want to use Japanese, set
up your system one of the followings.
- libc5 + X compiled with X_LOCALE
- glibc-2.0 + libwcsmbs + X compiled without X_LOCALE
- glibc-2.1 + locale-ja + X compiled without X_LOCALE
The location in which the |locale|s are installed varies system to system.
For example, "/usr/share/locale", "/usr/lib/locale", etc. See your system's
setlocale() man page.
The format of |locale| name is:
Territory means the country, codeset means the |charset|. For example, the
|locale| name "ja_JP.eucJP" means the language is Japanese, the country is
Japan, the codeset is EUC-JP. But it also could be "ja", "ja_JP.EUC",
"ja_JP.ujis", etc. And unfortunately, the |locale| name for a specific
language, territory and codeset is not unified and depends on your system.
This name is used for the LANG environment value. When you want to use Korean
and the |locale| name is "ko", do this:
sh: export LANG=ko
csh: setenv LANG ko
Examples of locale name:
|charset| language |locale-name|
GB2312 Chinese (simplified) zh_CN.EUC, zh_CN.GB2312
Big5 Chinese (traditional) zh_TW.BIG5, zh_TW.Big5
CNS-11643 Chinese (traditional) zh_TW
EUC-JP Japanese ja, ja_JP.EUC, ja_JP.ujis, ja_JP.eucJP
Shift_JIS Japanese ja_JP.SJIS, ja_JP.Shift_JIS
EUC-KR Korean ko, ko_KR.EUC
Even if your system does not have the multibyte language |locale| of your
choice, or does not have a enough implementation of the locale, Vim can
somehow handle the multibyte languages. Add "--enable-broken-locale" flag at
CODED CHARACTER SET (CCS)
|CCS| is a mapping from a set of characters to a set of integers. For
example, ((65, A), (66, B), (67, C)) is a |CCS| and ((0x41, A), (0x42, B),
(0x43, C)) is also a |CCS|. Examples of |CCS| are ISO 10646, US-ASCII,
ISO-8859 series, JIS X 0208, JIS X 0201, KS C 5601 (KS X 1001) and KS C 5636
(KS X 1003).
The term "integer" means code point or character number and is different from
octets or bit combination.
Typically, a |CCS| is a character table. Representing the column/line as
hexadecimal number becomes the code point of the character. For example,
US-ASCII CCS has 8x16 character table, the column number start with 0 and end
with 7, the line number start with 0 end with F. The code point of the
character at 4/1 is 0x41.
CHARACTER ENCODING SCHEME (CES)
|CES| is a mapping from a sequence of elements in one or more |CCS|es to a
sequence of octets. Examples of |CES| are EUC-JP, EUC-KR, EUC-CN (GB 2312),
EUC-TW (CNS-11643), ISO-2022-JP, ISO-2022-KR, ISO-2022-CN, UTF-8, etc.
|charset| is a method of converting a sequence of octets into a sequence of
characters, the combination of one or more |CCS|es and a |CES|. For example,
ISO-2022-JP |charset| is the combination of ASCII, JIS X 0201, JIS X 0208
|CCS|es and ISO-2022-JP |CES|. Examples of |charset| are US-ASCII, ISO-8859
series, GB2312, EUC-JP, EUC-KR, Shift_JIS, Big5, UTF-8, etc.
Note that this is not a term used by other standards bodies, such as ISO, but
a term defined in RFC 2130. The term "codeset" in POSIX has the same meaning
as |charset| here. |charset| does not mean character set (a set of
characters) and the term "character repertoire" means a collection of distinct
characters. There are historical reasons, see RFC 2130.
One language could have some |charset|s. For example, Japanese has
ISO-2022-JP, EUC-JP and Shift_JIS |charset|s. ISO-2022-JP |charset| is used
for Internet Messages, because it is encoded in 7-bit format. EUC-JP is used
on Unix, Shift_JIS is used on Windows and MacOS. Korean has ISO-2022-KR and
EUC-KR, ISO-2022-KR is for Internet Messages, EUC-KR is used on all three
major platforms, MacOS, Unix and MS-Windows. Chinese has HZ, GB2312 (EUC-CN),
ISO-2022-CN, Big5 and CNS-11643 (EUC-TW). In PRC, People's Republic of China,
simplified Chinese characters are used in daily life and GB2312 is the
national standard charset. In Taiwan, traditional Chinese characters are used
in daily life. Though CNS-11643 (EUC-TW) is the formal charset, Big5 is the
Note: CNS-11643 or EUC-TW is not registered in IANA.
Vim does not convert automatically to the locale's |charset| at display time.
So, if a file's |charset| differs from your locale's |charset|, the file is
not displayed correctly. So, you must know the file's |charset| by anyway,
guessing, using some utilities, etc, and convert the |charset| to the locale's
Useful utilities for converting the |charset|:
Nkf is "Network Kanji code conversion Filter". One of the most unique
facility of nkf is the guess of the input Kanji code. So, you don't
need to know what the inputting file's |charset| is. When convert to
EUC-JP from ISO-2022-JP or Shift_JIS, simply do the following command
Nkf can be found at:
Hc is "Hanzi Converter". Hc convert a GB file to a Big5 file, or Big5
file to GB file. Hc can be found at:
Hmconv is Korean code conversion utility for especially for E-mail. It
can convert between EUC-KR and ISO-2022-KR. Hmconv can be found at:
Lv is a Powerful Multilingual File Viewer. And it can be worked as
|charset| converter. Supported |charset|: ISO-2022-CN, ISO-2022-JP,
ISO-2022-KR, GB2312 (EUC-CN), EUC-JP, EUC-KR, CNS-11643 (EUC-TW),
UTF-7, UTF-8, ISO-8859 series, Shift_JIS, Big5 and HZ. Lv can be found
X LOGICAL FONT DESCRIPTION (XLFD)
XLFD is the X font name and contains the information about the font size,
|CCS|, etc. The name is in this format:
Each field means:
- FOUNDRY: FOUNDRY field. The company that created the font.
- FAMILY: FAMILY_NAME field. Basic font family name. (helvetica, gothic,
- WEIGHT: WEIGHT_NAME field. How thick the letters are. (light, medium,
- SLANT: SLANT field.
ri: Reverse Italic
ro: Reverse Oblique
number: Scaled font
- WIDTH: SETWIDTH_NAME field. Width of characters. (normal, condensed,
narrow, double wide)
- STYLE: ADD_STYLE_NAME field. Extra info to describe font. (Serif, Sans
Serif, Informal, Decorated, etc)
- PIXEL: PIXEL_SIZE field. Height, in pixels, of characters.
- POINT: POINT_SIZE field. Ten times height of characters in points.
- X: RESOLUTION_X field. X resolution (dots per inch).
- Y: RESOLUTION_Y field. Y resolution (dots per inch).
- SPACE: SPACING field.
- AVE: AVERAGE_WIDTH field. Ten times average width in pixels.
- CR: CHARSET_REGISTRY field. Indicates the name of the font |CCS| name.
- CE: CHARSET_ENCODING field. In some CCSes, such as ISO-8859 series,
this field is the part of |CCS| name. In other CCSes, such as JIS
X 0208, if this field is 0, code points has the same value as GL,
and GR if 1.
For example, in case of a 14 dots font corresponding to JIS X 0208, it is
A |CCS| typically associated with one font. The languages which must manage
multiple |CCS|es needs to manage multiple font. In X11R5, for the
internationalization of output API, FontSet was introduced. By using this,
Xlib takes care of switching of fonts and the display. Till X11R4, the
application themselves had to manage this.
|locale| database has the information about the |charset| of the |locale|,
which |CCS|(es) is needed and which |CES| the locale uses. When you use the
locale which must manage multiple |CCS|es, you have to specify the each
|CCS|'s font in 'guifontset' option.
|charset| language |CCS|es
GB2312 Chinese (simplified) ISO-8859-1 and GB 2312
Big5 Chinese (traditional) ISO-8859-1 and Big5
CNS-11643 Chinese (traditional) ISO-8859-1, CNS 11643-1 and CNS 11643-2
EUC-JP Japanese JIS X 0201 and JIS X 0208
EUC-KR Korean ISO-8859-1 and KS C 5601 (KS X 1001)
The |XLFD| contains the information of |CCS|. So, by searching in fonts.dir,
you can find the |CCS|'s font. The fonts.dir is in the fonts directory (e.g.
/usr/X11R6/lib/X11/fonts/*), the format of the file is:
First line: the number of fonts which are contained in this fonts.dir
other line: FILENAME |XLFD|
Or, you can search fonts using xlsfonts command. For example, when you're
searching for the font for KS C 5601:
xlsfonts | grep ksc5601
will show you the list of it.
In 'guifontset' option and ~/.Xdefaults, you specify the
|base_font_name_list|, which is a list of |XLFD| font names that Xlib uses to
load the fonts needed for the |locale|. The base font names are a
For example, when you use the ja_JP.eucJP |locale|, which require JIS X 0201
and JIS X 0208 |CCS|es. You could supply a |base_font_name_list| that
explicitly specifies the charsets, like:
Alternatively, the user could supply a base font name list that omits the
|CCS| name, letting Xlib select font characters required for the locale. For
Alternatively, the user could supply a single base font name that allows Xlib
to select from all available fonts. For example:
Alternatively, the user could specify the alias name. See fonts.alias in
the fonts directory.
Note that in East Asian fonts, the standard character cell is square. When
mixing Latin font and East Asian font, East Asian font width should be twice
the Latin font width. And GVIM needs fixed width font.
X INPUT METHOD (XIM) *XIM* *xim* *x-input-method*
XIM (X Input Method) is an international input module for X. There are two
kind of structures, Xlib unit type and |IM-server| (Input-Method server) type.
|IM-server| type is suitable for complex inputting, like CJK inputting.
In |IM-server| type input structures, the input event is handled by either
of the two ways: FrontEnd system and BackEnd system. In the FrontEnd
system, input events are snatched by the |IM-server| first, then |IM-server|
give the application the result of input. On the other hand, the BackEnd
system works reverse order. MS Windows adopt BackEnd system. In X, most of
|IM-server|s adopt FrontEnd system. The demerit of BackEnd system is the
large overhead in communication, but it provides safe synchronization with
no restrictions on applications.
For example, there are xwnmo and kinput2 Japanese |IM-server|, both are
FrontEnd system. Xwnmo is distributed with Wnn (see below), kinput2 can be
found at: ftp://ftp.sra.co.jp/pub/x11/kinput2/
- Conversion Server
Some system needs additional server: conversion server. Most of Japanese
|IM-server|s need it, Kana-Kanji conversion server. For Chinese inputting,
it depends on the method of inputting, in some methods, PinYin or ZhuYin to
HanZi conversion server is needed. For Korean inputting, if you want to
input Hanja, Hangul-Hanja conversion server is needed.
For example, the Japanese inputting process is divided into 2 steps. First
we pre-input Hira-gana, second Kana-Kanji conversion. There are so many
Kanji characters (6349 Kanji characters are defined in JIS X 0208) and the
number of Hira-gana characters are 76. So, first, we pre-input text as
pronounced in Hira-gana, second, we convert Hira-gana to Kanji or Kata-Kana,
if needed. There are some Kana-Kanji conversion server: jserver
(distributed with Wnn, see below) and canna. Canna can be found at:
There is a good input system: Wnn4.2. Wnn 4.2 contains,
xwnmo (|multilingualized| |IM-server|)
jserver (Japanese Kana-Kanji conversion server)
cserver (Chinese PinYin or ZhuYin to simplified HanZi conversion server)
tserver (Chinese PinYin or ZhuYin to traditional HanZi conversion server)
kserver (Hangul-Hanja conversion server)
Wnn 4.2 can be found at:
For Chinese, there's a great XIM server named "xcin", you can input both
Traditional and Simplified Chinese characters. And it can accept other locale
if you make a correct input table. Xcin can be found at:
- Input Style
When inputting CJK, there needs four areas.
1. The area to perform display of input in the midst
2. The area to display input mode.
3. The area to display the next candidate for the selection.
4. The area to display other tools.
The third area is needed when converting. For example, in Japanese
inputting, multiple Kanji characters could have the same pronunciation, so
a sequence of Hira-gana characters could map to a distinct sequence of Kanji
The first and second areas are defined in international input of X with the
names of "Preedit Area", "Status Area" respectively. The third and fourth
areas are not defined and are left to be managed by the |IM-server|. In the
international input, four input styles have been defined using combinations
of Preedit Area and Status Area: |OnTheSpot|, |OffTheSpot|, |OverTheSpot|
Currently, GUI Vim support three style, |OverTheSpot|, |OffTheSpot| and
*. on-the-spot *OnTheSpot*
Preedit Area and Status Area are performed by the client application in
the area of application. The client application is directed by the
|IM-server| to display all pre-edit data at the location of text
insertion. The client registers callbacks invoked by the input method
*. over-the-spot *OverTheSpot*
Status Area is created in a fixed position within the area of application,
in case of Vim, the position is the additional status line. Preedit Area
is made at present input position of application. The input method
displays pre-edit data in a window which it brings up directly over the
text insertion position.
*. off-the-spot *OffTheSpot*
Preedit Area and Status Area are performed in the area of application, in
case of Vim, the area is additional status line. The client application
provides display windows for the pre-edit data to the input method which
displays into them directly.
*. root-window *Root*
Preedit Area and Status Area are performed outside of the area of
application. The input method displays all pre-edit data in a separate
area of the screen in a window specific to the input method.
LOCALIZATION, INTERNATIONALIZATION AND MULTILINGUALIZATION
*localized* *Localization* *L10N*
Localization (L10N) To fit a system or an application with a
*internationalized* *Internationalization* *I18N*
Internationalization (I18N) To enable a system or an application to fit
with a specific language according to the
*multilingualized* *Multilingualization* *M17N*
Multilingualization (M17N) To enable a system or an application to be
able to use multiple languages at the same
For example, JVim (Japanized version Vim 3.0) is a |localized| application for
Japanese. Cxterm (|localized| xterm for Chinese), kterm (|localized| xterm
for Japanese) and hanterm (|localized| xterm for Korean) is also a |localized|
application. Gnome is an |internationalized| application. It can be
|localized| for many languages according to the |locale|. Mule (Multilingual
Enhancement for GNU Emacs) is a |multilingualized| application. It can handle
multiple |charset|s and can maintain a mixture of languages in a single
Vim is an |internationalized| application. So, you can change the language
specifying the |locale| and some options at start time.
2. Compiling *multibyte-compiling*
-. Before you start to compile Vim, be sure that your system has the language
|locale| of your choice. You might need to add "-DX_LOCALE" to CFLAGS.
-. Compiling Vim:
> ./configure --with-x --enable-multibyte --enable-fontset --enable-xim
-. You can use multi-byte in the Vim GUI, which fully supports the
|+multi_byte| feature. If you only use console Vim, low-level multibyte
input/output depends on your console. For example, if you run Vim in an
xterm, you should use a |localized| xterm or an xterm which support |XIM|.
|localized| xterms are kterm (Kanji term) or hanterm (for korean) for
example. Known |XIM| supporting xterms are Eterm (Enlightened terminal),
3. Display *multibyte-display*
Note that Display and Input are independent. It is possible to see your
language even though you have no input method for it.
Multibyte output uses |xfontset| feature.
-. Be sure that your system has the fonts corresponding to the |CCS|es, which
the |locale| needs to manage. See: |xfontset|.
-. Following are requirements to use multibyte language.
If needed, insert the lines below in your $HOME/.Xdefaults file.
The GTK+ version of GUI Vim does not use .Xdefaults, thus this change is
not needed for the GTK+ version.
These 3 lines are specific for Vim:
Note: Vim.font is for text area.
Vim*fontSet is for menu.
Vim*fontList is for menu (for Motif GUI)
For example, when you are using Japanese and 14 dots font,
> Vim.font: -misc-fixed-medium-r-normal--14-*
> Vim*fontSet: -misc-fixed-medium-r-normal--14-*
> Vim*fontList: -misc-fixed-medium-r-normal--14-*
> Vim.font: k14,r14
> Vim.fontSet: k14,r14
> Vim.fontList: k14
You should set the 'guifontset' option to display a multi-byte language.
For example, when you are using Japanese and 14 dots font,
> set guifontset=-misc-fixed-medium-r-normal--14-*
> set guifontset=k14,r14
Note: You can not use IM unless you specify 'guifontset'.
Therefore, Latin users, you have to also use 'guifontset'
if you use IM.
You should not set 'guifont'. If it is set, Vim ignores 'guifontset'.
It means Vim runs without fontset support, you can see only English. The
multi-byte characters are displayed corrupted.
After the |+fontset| feature is enabled as explained above, Vim does not
allow using 'font'. For example, if you use:
> :set guifontset=eng_font,your_font
in your .gvimrc, then you should use for highlighting:
> :hi Comment font=another_eng_font,another_your_font
If you would do
> :hi Comment font=another_eng_font
VIM will also try to use it as a fontset. So, if it cannot display your
|locale| dependent codeset, you will see a error message.
-. In your .vimrc, add this
> set fileencoding=korea
You can change "korea" to the some other name such as japan, taiwan.
See |'fileencoding'| for the supported encodings.
-. If a file's charset is different from your |locale|'s charset, you need to
convert the charset. See |charset-conversion|.
4. Input (XIM, X Input Method support) *multibyte-input*
Note that Display and Input are independent. It is possible to see your
language even though you have no input method for it. But when your Display
method doesn't match your Input method, the text will be displayed wrong.
-. To input your language you should run the |IM-server| which supports your
language and |conversion-server| if needed. Multibyte input uses |XIM|
Next 3 lines are common for all X applications which uses |XIM|.
If you already use |XIM|, don't care.
> *international: True
> *.inputMethod: your_input_server_name
> *.preeditType: your_input_style
Note: input_server_name is your |IM-server| name (check your
your_input_style is one of |OverTheSpot|, |OffTheSpot|, |Root|.
See also |xim-input-style|.
*international may not necessary if you use X11R6.
*.inputMethod and *.preeditType is a optional if you use X11R6.
For example, when you are using kinput2 as |IM-server|,
> *international: True
> *.inputMethod: kinput2
> *.preeditType: OverTheSpot
When using |OverTheSpot|, GUI Vim always connects to the IM Server even in
Normal mode, so you can input your language with commands like "f" and
"r". But when using one of the other two methods, GUI Vim connects to the
IM Server only if it is not in Normal mode.
If your IM Server does not support |OverTheSpot|, and if you want to use
your language with some Normal mode command like "f" or "r", then you
should use a |localized| xterm or an xterm which supports |XIM|
-. If needed, you can set the XMODIFIERS env. var.
sh: export XMODIFIERS="@im=input_server_name"
csh: setenv XMODIFIERS "@im=input_server_name"
For example, when you are using kinput2 as |IM-server| and sh,
> export XMODIFIERS="@im=kinput2"
Contributions specifically for the multi-byte features by:
Chi-Deok Hwang <hwang@...
Sung-Hyun Nam <namsh@...
Taro Muraoka <koron@...
Yasuhiro Matsumoto <mattn@...
5. UTF-8 in XFree86 xterm *UTF8-xterm*
This is a short explanation of how to use UTF-8 character encoding in the
xterm that comes with XFree86 by Thomas Dickey (text by Markus Kuhn).
NOTE: Editing and viewing UTF-8 text in Vim does not work as expected yet!
Get the latest xterm version which has now UTF-8 support:
Compile it with "./configure --enable-wide-chars ; make"
Also get the ISO 10646-1 version of the 6x13 font, which is available on
and install the font as described in the README file.
Now start xterm with
> xterm -u8 -fn -misc-fixed-medium-r-semicondensed--13-120-75-75-c-60-iso10646-1
and you will have a working UTF-8 terminal emulator. Try both
> cat utf-8-demo.txt
> vim utf-8-demo.txt
with the demo text that comes with ucs-fonts.tar.gz in order to see
whether there are any problems with UTF-8 in your xterm.