README for I18N Package

o Name and location of package

Name:           php-3.0.18-i18n-ja-2
Location:       http://www.happysize.co.jp/techie/php-ja-jp/

Currently, this I18N version of PHP only adds Japanese support to base
PHP.  It allows you to use Japanese in scripts, as well as conversion
between various Japanese encodings.  It will work perfectly fine with 
ASCII with i18n option enabled.  (note: executable is bit larger due 
to UNICODE table).  The basic design aproach is to allow for other 
languages to be added in the future.  Developers are encourage to join

For more information on Japanese encodings, please refer to the 
section "Additional Notes."

o What is this package?

This package allows you to handle multiple Japanese encodings (SJIS, EUC,
UTF-8, JIS) in PHP.  If you find any bugs in this package, please report
them to the appropriate mailing list.  For now, the PHP-jp mailing list 
is the best place for this.

PHP-jp ML       mailto:PHP-jp@sidecar.ics.es.osaka-u.ac.jp
                (discussions are in Japanese)

o Who should use this

Due to lack of documentation, it's not intended for beginners.  If
something goes wrong, be prepared to fix it on your own.

o Warranty and Copyright

There is no warranty with this package.  Use it at your own risk.

Please refer to the source code for the copyrights.  In general, each
program's copyright is owned by the programmer.  Unless you obey the
copyright holders restrictions, you are not allowed to use it in any

o Redistribution

As described in the source code, this package and the components are
allowed to be redistributed with certain restrictions.

Due to this package being still in beta, please try to redistribute
it as an entire package.  Please try not to distribute it as a form
of patch.  Because we would prefer to have this package distributed
as one single package (not patch of patch of patch), avoid releasing
any patch to this package.

o Who made this

A team of volunteers, PHP3 Internationalization, has been contributing
their free time producing it.  Although we are not related to the core
PHP programmers, we are hoping to have our modifications merged into the
core distribution in the near future.  Thus, we did not call this a
"Japanese Patch" (or distribution).  Our final goal is to have true
i18nized PHP!

For anyone interested in this project, please drop us a line.

Contact Address:
        (Discussions are in Japanese, but feel free to write us in English)

Webpage (English and Japanese):

Project Outline (Japanese):

        Hironori Sato 
        Shigeru Kanemoto 
        Tsukada Takuya 
        U. Kenkichi 
        Other gracious contributors

o Future plans

- fulfilling what's written in outline
- support for other languages other than Japanese
- make the character conversion as a library (?)
- more testing

o Special Thanks to

PHP Japanese webpage maintainer, Hirokawa-san
PHP-JP ML's Yamamoto-san
Previous jp-patch developers

  Advantages of using I18N package

- allows you to use various character encodings for script files and 
  http output
- distinguish character encoding in POST/GET/COOKIE
- proper mail output using JIS as body and MIME/Base64/JIS subject
- if http output's Content-Type is text/html, it will set proper charset
- stable character encoding conversion
- multibyte regex


o Summary

Add --enable-i18n option when running configure.  For your own setup,
add any other appropriate options as well.

Don't forget to copy php3.ini-dist to desired location.
(ex. /usr/local/lib/php3.ini)

If you have already installed PHP3, copy all the entries in php3.ini-dist
which start with "i18n.xxxx" to php3.ini.

o configure option
      include i18n features

      include multibyte regex library
      (without i18n enabled, mbregex functions will not function)

o creating cgi version

    % tar xvzf php-3.0.18-i18n-ja-2.tar.gz
    % cd php-3.0.18-i18n-ja-2
    % ./configure --enable-i18n --enable-mbregex
    % make

o creating Apache version (regular module)

    % tar xvzf php-3.0.18-i18n-ja-2.tar.gz
    % tar xvzf apache_1.3.x.tar.gz
    % cd apache_1.3.x
    % ./configure
    % cd ../php-3.0.18-i18n-ja-2
    % ./configure --with-apache=../apache_1.3.x --enable-i18n --enable-mbregex
    % make
    % make install
    % cd ../apache_1.3.x
    % ./configure --activate-module=src/modules/php3/libphp3.a
    % make
    % make install

o creating Apache DSO version

    create DSO capable Apache first
    % tar xvzf apache_1.3.x.tar.gz
    % cd apache-1.3.x
    % ./configure --enable-shared=max
    % make
    % make install

    now create php3
    % cd php-3.0.18-i18n-ja-2
    % ./configure --with-apxs=/usr/local/apache/bin/apxs --enable-i18n \
    % make
    % make install

  Additional Notes

o Multibyte regex library

From beta4, we have included the multibyte (mb) regex library which comes with
Ruby.  With this addition, you can now use regex in EUC, SJIS and UTF-8
encoding.  To avoid any conflicts with HSREGEX included with Apache,
each function name has been changed.  Therefore, mb regex functions are
named differently from the original ereg functions in PHP.  The character
encoding used in mb regex is configured in i18n.internal_encoding.

o Binary Output

If http output encoding is set to other than 'pass', conversion of encoding
from internal encoding to http output is done automatically.  Thus,
if you prefer to spit out anything in raw binary format, your data
may be corrupted.  In such event, set http_output to 'pass'.


o Content-Type

Depending on the setting of http_output, PHP will output the proper charset.
ex. Content-Type: text/html; charset="..."

Be aware of following:

- If you set Content-Type header using header() function, that will
  override the automatic addition of charset.
- Be cautious when you set i18n_http_output, since if any output is
  made prior to this, proper header may have been sent out to the
  client already.

o In the event of trouble

If you find any bugs or trouble, please contact us at the above address.
It may help us to track the problem if you send us the script as well.

If you encounter any memory related error such as segmentation violation,
add --enable-debug when you run configure.  This will give you more
detail information on where error has occurred.  The error is stored
in the server log or regular http output in CGI mode.

o About Japanese encodings

Due to historical reason, there are multiple character encodings used
for Japanese.  The most common encodings are: SJIS, EUC, JIS, and UTF-8.  
Here are (very) brief description of them:

  commonly used in UNIX environment
  8bit-8bit combo
  always >=0x80

  commonly used in Mac or PCs
  similar to EUC
  mostly 8bit-8bit (some 8bit-7bit)
  mostly >=0x80
  there are some halfwidth (size of ASCII) multibytes

  commonly used in 7bit environment (nntp and smtp)
  starts with escaping char, \033 and a few more characters

  16bit+ encoding
  defines many languages existing in this world
  see http://www.unicode.org/ for more detail

Because of having all these character encodings, PHP needs to translate
between these encodings on the fly.  Also, the addition of the mb regex 
library allows you to handle mb strings without fear of getting mb char 
chopped in half.

Since Japanese is not the only language with multiple encodings, we
encourage other developers to modify our code to suit your needs.  We
definitely need people to work with Korean, Chinese (both traditional
and simplified), and Russian.  Let us know if you are interested in
this project!

  php3.ini setting

The following init options will allow you to change the default settings.
Define these settings in the global section of php3.ini.

All keywords are case-insensitive.

o Encoding naming

    For each encoding, there are three names: standarized, alias, MIME

    - UTF-8
         standard: UTF-8
         alias: N/A 
         mime: UTF-8

    - ASCII
         standard: ASCII
         alias: N/A
         mime: US-ASCII

    - Japanese EUC
         standard: EUC-JP
         alias: EUC, EUC_JP, eucJP, x-euc-jp
         mime: EUC-JP

    - Shift JIS
         standard: SJIS
         alias: x-sjis, MS_Kanji
         mime: Shift_JIS

    - JIS
         standard: JIS
         alias: N/A 
         mime: ISO-2022-JP

    - Quoted-Printable
         standard: Quoted-Printable
         alias: qprint
         mime: N/A

    - BASE64
         standard: BASE64
         alias: N/A
         mime: N/A

    - no conversion
         standard: pass
         alias: none
         mime: N/A

    - auto encoding detection
         standard: auto
         alias: unknown
         mime: N/A

    * N/A - Not Applicapable

o i18n.http_output - default http output encoding

    i18n.http_output = EUC-JP|SJIS|JIS|UTF-8|pass
        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        pass: no conversion

    The default is pass (internal encoding is used)
    It can be re-configured on the fly using i18n_http_output().

o i18n.internal_encoding - internal encoding

    i18n.internal_encoding = EUC-JP|SJIS|UTF-8
        EUC-JP : EUC
        SJIS: SJIS
        UTF-8: UTF-8

    The default is EUC-JP.  

    PHP parser is designed based on using ISO-8859-1.  For other
    encodings, following conditions have to be satisfied in order
    to use them:
       - per byte encoding
       - single byte charactor in range of 00h-7fh which is compatible 
         with ASCII
       - multibyte without 00h-7fh
    In case of Japanese, EUC-JP and UTF-8 are the only encoding that
    meets this criteria.

    If i18n.internal_encoding and i18n.http_output differs, conversion
    takes place at the time of output.  If you convert any data within
    PHP scripts to URL encoding, BASE64 or Quoted-Printable, encoding
    stays as defined in i18n.internal_encoding.  Thus, if you would
    prefer to encode in compliance with i18n.http_output, you need
    to manually convert encoding.

    ex. $str = urlencode( i18n_convert($str, i18n_http_output()) );

    Encoding such as ISO-2022-** and HZ encoding which uses escape
    sequences can not be used as internal encoding.  If used, they
    result in following errors:
       - parser pukes funky error
       - magic_quotes_*** breaks encoding (SJIS may have similar problem)
       - string manipulation and regex will malfunction

o i18n.script_encoding - script encoding

    i18n.script_encoding = auto|EUC-JP|SJIS|JIS|UTF-8
        auto: automatic
        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

    The default is auto.
    The script's encoding is converted to i18n.internal_encoding before
    entering the script parser.

    Be aware that auto detection may fail under some conditions.
    For best auto detection, add multibyte charactor at begining of

o i18n.http_input - handling of http input (GET/POST/COOKIE)

    i18n.http_input = pass|auto
        auto: auto conversion
        pass: no conversion

    The default is auto.
    If set to pass, no conversion will take place.
    If set to auto, it will automatically detect the encoding.  If
    detection is successful, it will convert to the proper internal
    encoding.  If not, it will assume the input as defined in

o i18n.http_input_default - default http input encoding

    i18n.http_input_default = pass|EUC-JP|SJIS|JIS|UTF-8
        pass: no conversion
        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

    The default is pass.
    This option is only effective as long as i18n.http_input is set to
    auto.  If the auto detection fails, this encoding is used as an
    assumption to convert the http input to the internal encoding.
    If set to pass, no conversion will take place.

o sample settings

    1) For most flexibility, we recommend using following example.
         i18n.http_output = SJIS
         i18n.internal_encoding = EUC-JP
         i18n.script_encoding = auto
         i18n.http_input = auto
         i18n.http_input_default = SJIS

    2) To avoid unexpected encoding problems, try these:

         i18n.http_output = pass
         i18n.internal_encoding = EUC-JP
         i18n.script_encoding = pass
         i18n.http_input = pass
         i18n.http_input_default = pass

  PHP functions

The following describes the additional PHP functions.

All keywords are case-insensitive.

o i18n_http_output(encoding)
o encoding = i18n_http_output()

    This will set the http output encoding.  Any output following this
    function will be controlled by this function.  If no argument is given,
    the current http output encode setting is returned.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        pass: no conversion

    NONE is not allowed

o encoding = i18n_internal_encoding()

    Returns the current internal encoding as a string.

    internal encoding
        EUC-JP : EUC
        SJIS: SJIS
        UTF-8: UTF-8

o encoding = i18n_http_input()

    Returns http input encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        pass: no conversion (only if i18n.http_input is set to pass)

o string = i18n_convert(string, encoding)
  string = i18n_convert(string, encoding, pre-conversion-encoding)

    Returns converted string in desired encoding.  If
    pre-conversion-encoding is not defined, the given 
    string is assumed to be in internal encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        pass: no conversion

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        pass: no conversion
        auto: auto detection

o encoding = i18n_discover_encoding(string)

    Encoding of the given string is returned (as a string).

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        ASCII: ASCII (only 09h, 0Ah, 0Dh, 20h-7Eh)
        pass: unable to determine (text is too short to determine)
        unknown: unknown or possible error

o int = mbstrlen(string)
o int = mbstrlen(string, encoding)

    Returns character length of a given string.  If no encoding is defined,
    the encoding of string is assumed to be the internal encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8
        auto: automatic

o int = mbstrpos(string1, string2)
o int = mbstrpos(string1, string2, start)
o int = mbstrpos(string1, string2, start, encoding)

    Same as strpos.  If no encoding is defined, the encoding of string
    is assumed to be the internal encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

o int = mbstrrpos(string1, string2)
o int = mbstrrpos(string1, string2, encoding)

    Same as strrpos.  If no encoding is defined, the encoding of string
    is assumed to be the internal encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

o string = mbsubstr(string, position)
o string = mbsubstr(string, position, length)
o string = mbsubstr(string, position, length, encoding)

    Same as substr.  If no encoding is defined, the encoding of string
    is assumed to be the internal encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

o string = mbstrcut(string, position)
o string = mbstrcut(string, position, length)
o string = mbstrcut(string, position, length, encoding)

    Same as subcut.  If position is the 2nd byte of a mb character, it will cut
    from the first byte of that character.  It will cut the string without
    chopping a single byte from a mb character.  In another words, if you
    set length to 5, you will only get two mb characters.  If no encoding
    is defined, the encoding of string is assumed to be the internal encoding.

        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

o string = i18n_mime_header_encode(string)
    MIME encode the string in the format of =?ISO-2022-JP?B?[string]?=.

o string = i18n_mime_header_decode(string)
    MIME decodes the string.

o string = i18n_ja_jp_hantozen(string)
o string = i18n_ja_jp_hantozen(string, option)
o string = i18n_ja_jp_hantozen(string, option, encoding)

    Conversion between full width character and halfwidth character.

    The following options are allowed.  The default is "KV".
    Acronym: FW = fullwidth, HW = halfwidth

    "r" :  FW alphabet -> HW alphabet

    "R" :  HW alphabet -> FW alphabet

    "n" :  FW number -> HW number

    "N" :  HW number -> FW number

    "a" :  FW alpha numeric (21h-7Eh) -> HW alpha numeric

    "A" :  HW alpha numeric (21h-7Eh) -> FW alpha numeric

    "k" :  FW katakana -> HW katakana

    "K" :  HW katakana -> FW katakana

    "h" :  FW hiragana -> HW hiragana

    "H" :  HW hiragana -> FW katakana

    "c" :  FW katakana -> FW hiragana

    "C" :  FW hiragana -> FW katakana

    "V" :  merge dakuon character.  only works with "K" and "H" option

    If no encoding is defined, the encoding of string is assumed to be
    the internal encoding.
        EUC-JP : EUC
        SJIS: SJIS
        JIS : JIS
        UTF-8: UTF-8

int = mbereg(regex_pattern, string, string)
int = mberegi(regex_pattern, string, string)
    mb version of ereg() and eregi()

string = mbereg_replace(regex_pattern, string, string)
string = mberegi_replace(regex_pattern, string, string)
    mb version of ereg_replace() and eregi_replace()

string_array = mbsplit(regex, string, limit)
    mb version of split()


Here, we have gathered some commonly asked questions on PHP-jp mailing

o To use Japanese in GET method

If you need to assign Japanese text in GET method with argument, such as;
xxxx.php?data=, use urlencode function in PHP.  If not,
text may not be passed onto action php properly.

ex: Link

o When passing data via GET/POST/COOKIE, \ character sneaks in

When using SJIS as internal encoding, or passed-on data includes '"\, 
PHP automatically inserts escaping character, \.  Set magic_quotes_gpc
in php3.ini from On to Off.  An alternative work around to this problem 
is to use StripSlashes().

If $quote_str is in SJIS and you would like to extract Japanese text,
use ereg_replace as follows:


This will effectively extract Japanese text out of $quote_str.

o Sometimes, encoding detection fails

If i18n_http_input() returns 'pass', it's likely that PHP failed to
detect whether it's SJIS or EUC.  In such case, use  to properly detect the incoming text's 

  Japanese Manual
Translated manual done by "PHP Japanese Manual Project" :


Starting 3.0.18-i18n-ja, we have removed doc-jp from tarball package.

  Change Logs

o 2000-10-28, Rui Hirokawa 

This patch is derived from php-3.0.15-i18n-ja as well as php-3.0.16 by 
Kuwamura applied to original php-3.0.18.  It also includes following fixes:

1) allows you to set charset in mail().
2) fixed mbregex definitions to avoid conflicts with system regex
3) php3.ini-dist now uses PASS for http_output instead of SJIS

o 2000-11-24, Hironori Sato 

Applied above patched and added detection for gdImageStringTTF in configure.
Following setups are known to work:

gd-1.3-6, gd-devel-1.3-6, freetype-1.3.1-5, freetype-devel-1.3.1-5
        i18n_convert("日本語", "UTF-8"));

gd-1.7.3-1k1, gd-devel-1.7.3-1k1, freetype-1.3.1-5, freetype-devel-1.3.1-5
    * i18n_internal_encoding = EUC 又は SJIS

For any gd libraries before 1.6.2, you need to use i18n_convert.  For
gd-1.5.2/3, upgrade to anything above 1.7 to use ImageTTFText without
using i18n_convert.  As long as you have internal_encoding set to EUC or
SJIS, ImageTTFText should work without mojibake.  Again, make sure you 
have i18n_http_output("pass") before calling ImageGif, ImagePng, ImageJpeg!

o 2000-12-09, Rui Hirokawa 

Fixed mail() which was causing segmentation fault when header was null.