In Files

  • nkf/nkf.c


Class/Module Index [+]



NKF - Ruby extension for Network Kanji Filter


This is a Ruby Extension version of nkf (Network Kanji Filter). It converts the first argument and returns converted result. Conversion details are specified by flags as the first argument.

Nkf is a yet another kanji code converter among networks, hosts and terminals. It converts input kanji code to designated kanji code such as ISO-2022-JP, Shift_JIS, EUC-JP, UTF-8 or UTF-16.

One of the most unique faculty of nkf is the guess of the input kanji encodings. It currently recognizes ISO-2022-JP, Shift_JIS, EUC-JP, UTF-8 and UTF-16. So users needn’t set the input kanji code explicitly.

By default, X0201 kana is converted into X0208 kana. For X0201 kana, SO/SI, SSO and ESC-(-I methods are supported. For automatic code detection, nkf assumes no X0201 kana in Shift_JIS. To accept X0201 in Shift_JIS, use -X, -x or -S.


-b -u

Output is buffered (DEFAULT), Output is unbuffered.

-j -s -e -w -w16 -w32

Output code is ISO-2022-JP (7bit JIS), Shift_JIS, EUC-JP, UTF-8N, UTF-16BE, UTF-32BE. Without this option and compile option, ISO-2022-JP is assumed.

-J -S -E -W -W16 -W32

Input assumption is JIS 7 bit, Shift_JIS, EUC-JP, UTF-8, UTF-16, UTF-32.


Assume JIS input. It also accepts EUC-JP. This is the default. This flag does not exclude Shift_JIS.


Assume Shift_JIS and X0201 kana input. It also accepts JIS. EUC-JP is recognized as X0201 kana. Without -x flag, X0201 kana (halfwidth kana) is converted into X0208.


Assume EUC-JP input. It also accepts JIS. Same as -J.


No conversion.


Output sequence to designate JIS-kanji. (DEFAULT B)


Output sequence to designate ASCII. (DEFAULT B)


{de/en}crypt ROT13/47

-h –hiragana –katakana –katakana-hiragana

-h1 –hiragana

Katakana to Hiragana conversion.

-h2 –katakana

Hiragana to Katakana conversion.

-h3 –katakana-hiragana

Katakana to Hiragana and Hiragana to Katakana conversion.


Text mode output (MS-DOS)


ISO8859-1 (Latin-1) support

-f[m [- n]]

Folding on m length with n margin in a line. Without this option, fold length is 60 and fold margin is 10.


New line preserving line folding.


Convert X0208 alphabet (Fullwidth Alphabets) to ASCII.

-Z -Z0

Convert X0208 alphabet to ASCII.


Converts X0208 kankaku to single ASCII space.


Converts X0208 kankaku to double ASCII spaces.


Replacing Fullwidth >, <, “, & into ‘&gt;’, ‘&lt;’, ‘&quot;’, ‘&amp;’ as in HTML.

-X -x

Assume X0201 kana in MS-Kanji. With -X or without this option, X0201 is converted into X0208 Kana. With -x, try to preserve X0208 kana and do not convert X0201 kana to X0208. In JIS output, ESC-(-I is used. In EUC output, SSO is used.


Assume broken JIS-Kanji input, which lost ESC. Useful when your site is using old B-News Nihongo patch.


allows any char after ESC-( or ESC-$.


forces ASCII after NL.


Replacing non iso-2022-jp char into a geta character (substitute character in Japanese).

-d -c

Delete r in line feed, Add r in line feed.


MIME ISO-2022-JP/ISO8859-1 decode. (DEFAULT) To see ISO8859-1 (Latin-1) -l is necessary.


Decode MIME base64 encoded stream. Remove header or other part before



Decode MIME quoted stream. ‘_’ in quoted stream is converted to space.


Non-strict decoding.

It allows line break in the middle of the base64 encoding.


No MIME decode.


MIME encode. Header style. All ASCII code and control characters are intact. Kanji conversion is performed before encoding, so this cannot be used as a picture encoder.


MIME encode Base64 stream.


Perfome quoted encoding.


Input and output code is ISO8859-1 (Latin-1) and ISO-2022-JP. -s, -e and -x are not compatible with this option.


new line mode Without this option, nkf doesn’t convert line breaks.


unix (LF)


windows (CRLF)


mac (CR)

–fj –unix –mac –msdos –windows

convert for these system

–jis –euc –sjis –mime –base64

convert for named code

–jis-input –euc-input –sjis-input –mime-input –base64-input

assume input system

–ic=input codeset –oc=output codeset

Set the input or output codeset. NKF supports following codesets and those codeset name are case insensitive.


a.k.a. RFC1468, 7bit JIS, JUNET

EUC-JP (eucJP-nkf)

a.k.a. AT&T JIS, Japanese EUC, UJIS


a.k.a. x-eucjp-open-19970715-ascii


a.k.a. x-eucjp-open-19970715-ms


Microsoft Version of EUC-JP.


SJIS, MS-Kanji


a.k.a. CP932


same as UTF-8N


UTF-8 without BOM


UTF-8 with BOM


same as UTF-16BE


UTF-16 Big Endian without BOM


UTF-16 Big Endian with BOM


UTF-16 Little Endian without BOM


UTF-16 Little Endian with BOM


same as UTF-32BE


UTF-32 Big Endian without BOM


UTF-32 Big Endian with BOM


UTF-32 Little Endian without BOM


UTF-32 Little Endian with BOM


NKDed UTF-8, a.k.a. UTF8-NFD (input only)

–fb-{skip, html, xml, perl, java, subchar}

Specify the way that nkf handles unassigned characters. Without this option, –fb-skip is assumed.

–prefix= escape character target character ..

When nkf converts to Shift_JIS, nkf adds a specified escape character to specified 2nd byte of Shift_JIS characters. 1st byte of argument is the escape character and following bytes are target characters.


Handle the characters extended in CP932 as unassigned characters.


When Unicode to Encoded byte conversion, don’t convert characters which is not round trip safe. When Unicode to Unicode conversion, with this and -x option, nkf can be used as UTF converter. (In other words, without this and -x option, nkf doesn’t save some characters)

When nkf convert string which related to path, you should use this opion.


Decode hex encoded characters.


Unescape percent escaped characters.

Ignore rest of -option.



Release date of nkf


Version of nkf


Full version string of nkf

Public Class Methods

guess(str) => encoding click to toggle source

Returns guessed encoding of str by nkf routine.

               static VALUE
rb_nkf_guess(VALUE obj, VALUE src)

    input_ctr = 0;
    input = (unsigned char *)StringValuePtr(src);
    i_len = RSTRING_LENINT(src);

    guess_f = TRUE;
    kanji_convert( NULL );
    guess_f = FALSE;

    return rb_enc_from_encoding(rb_nkf_enc_get(get_guessed_code()));
nkf(opt, str) => string click to toggle source

Convert str and return converted result. Conversion details are specified by opt as String.

require 'nkf'
output = NKF.nkf("-s", input)
               static VALUE
rb_nkf_convert(VALUE obj, VALUE opt, VALUE src)
    VALUE tmp;
    if (!output_encoding) rb_raise(rb_eArgError, "no output encoding given");

    switch (nkf_enc_to_index(output_encoding)) {
    case UTF_8_BOM:    output_encoding = nkf_enc_from_index(UTF_8); break;
    case UTF_16BE_BOM: output_encoding = nkf_enc_from_index(UTF_16BE); break;
    case UTF_16LE_BOM: output_encoding = nkf_enc_from_index(UTF_16LE); break;
    case UTF_32BE_BOM: output_encoding = nkf_enc_from_index(UTF_32BE); break;
    case UTF_32LE_BOM: output_encoding = nkf_enc_from_index(UTF_32LE); break;
    output_bom_f = FALSE;

    incsize = INCSIZE;

    input_ctr = 0;
    input = (unsigned char *)StringValuePtr(src);
    i_len = RSTRING_LENINT(src);
    tmp = rb_str_new(0, i_len*3 + 10);

    output_ctr = 0;
    output     = (unsigned char *)RSTRING_PTR(tmp);
    o_len      = RSTRING_LENINT(tmp);
    *output    = '\0';

    /* use _result_ begin*/
    result = tmp;
    result = Qnil;
    /* use _result_ end */

    rb_str_set_len(tmp, output_ctr);

    if (mimeout_f)
        rb_enc_associate(tmp, rb_usascii_encoding());
        rb_enc_associate(tmp, rb_nkf_enc_get(nkf_enc_name(output_encoding)));

    return tmp;