ruby/enc/trans/utf8_mac.trans

#include "transcode_data.h"

<%
  require 'utf8_mac-tbl'

  def charlen(v)
    v.gsub(/[0-7].|[c-d].{3}|e.{5}/, '.').size
  end

  map = {}
  MAC_DECOMPOSE_TBL.each do |c, d|
    v = map[c]
    next if v && charlen(v) > charlen(d)
    map[c] = d
  end
  transcode_tblgen("UTF-8", "UTF8-MAC",
    map.to_a + [
    ["{00-7F}", :nomap],
    ["{c2-df}{80-bf}", :nomap0],
    ["e0{a0-bf}{80-bf}", :nomap0],
    ["{e1-ec}{80-bf}{80-bf}", :nomap0],
    ["ed{80-9f}{80-bf}", :nomap0],
    ["{ee-ef}{80-bf}{80-bf}", :nomap0],
    ["f0{90-bf}{80-bf}{80-bf}", :nomap0],
    ["{f1-f3}{80-bf}{80-bf}{80-bf}", :nomap0],
  ])

  map = {}
  map["{00-7f}"] = :func_so
  map["{c2-df}{80-bf}"] = :func_so
  map["e0{a0-bf}{80-bf}"] = :func_so
  map["{e1-ec}{80-bf}{80-bf}"] = :func_so
  map["ed{80-9f}{80-bf}"] = :func_so
  map["{ee-ef}{80-bf}{80-bf}"] = :func_so
  map["f0{90-bf}{80-bf}{80-bf}"] = :func_so
  map["{f1-f3}{80-bf}{80-bf}{80-bf}"] = :func_so
  map["f4{80-8f}{80-bf}{80-bf}"] = :func_so
  transcode_generate_node(ActionMap.parse(map), "from_UTF8_MAC")

  # http://www.unicode.org/Public/UNIDATA/CompositionExclusions.txt
  composition_exclusions = [
    0x0958,0x0959,0x095A,0x095B,0x095C,0x095D,0x095E,0x095F,
    0x09DC,0x09DD,0x09DF,0x0A33,0x0A36,0x0A59,0x0A5A,0x0A5B,
    0x0A5E,0x0B5C,0x0B5D,0x0F43,0x0F4D,0x0F52,0x0F57,0x0F5C,
    0x0F69,0x0F76,0x0F78,0x0F93,0x0F9D,0x0FA2,0x0FA7,0x0FAC,
    0x0FB9,0xFB1D,0xFB1F,0xFB2A,0xFB2B,0xFB2C,0xFB2D,0xFB2E,
    0xFB2F,0xFB30,0xFB31,0xFB32,0xFB33,0xFB34,0xFB35,0xFB36,
    0xFB38,0xFB39,0xFB3A,0xFB3B,0xFB3C,0xFB3E,0xFB40,0xFB41,
    0xFB43,0xFB44,0xFB46,0xFB47,0xFB48,0xFB49,0xFB4A,0xFB4B,
    0xFB4C,0xFB4D,0xFB4E,0x2ADC,
#    0x1D15E,0x1D15F,0x1D160,0x1D161,0x1D162,0x1D163,0x1D164,
#    0x1D1BB,0x1D1BC,0x1D1BD,0x1D1BE,0x1D1BF,0x1D1C0,
    0x0340..0x0341,0x0343,0x0374,0x037E,0x0387,
    0x1F71,0x1F73,0x1F75,0x1F77,0x1F79,0x1F7B,0x1F7D,0x1FBB,
    0x1FBE,0x1FC9,0x1FCB,0x1FD3,0x1FDB,0x1FE3,0x1FEB,0x1FEE..0x1FEF,
    0x1FF9,0x1FFB,0x1FFD,0x2000..0x2001,0x2126,0x212A..0x212B,0x2329,0x232A,
    0xF900..0xFA0D,0xFA10,0xFA12,0xFA15..0xFA1E,0xFA20,0xFA22,0xFA25..0xFA26,
    0xFA2A..0xFA6D,0xFA70..0xFAD9,
#    0x2F800..0x2FA1D,
    0x0344,0x0F73,0x0F75,0x0F81
  ]
  extbl = {}
  composition_exclusions.each do |x|
    case x
    when Range
      x.each do |n|
        extbl[[n].pack("U").unpack("H*")[0]] = true
      end
    when Integer
      extbl[[x].pack("U").unpack("H*")[0]] = true
    end
  end
  ary = MAC_DECOMPOSE_TBL.reject{|k,v|charlen(v)!=2||extbl[k]}.map{|k,v|[v,k]}
  transcode_generate_node(ActionMap.parse(ary), "from_utf8_mac_nfc2")
%>

<%= transcode_generated_code %>

#define BYTE_ADDR(index) (<%= OUTPUT_PREFIX %>byte_array + (index))
#define WORD_ADDR(index) (<%= OUTPUT_PREFIX %>word_array + INFO2WORDINDEX(index))
#define BL_BASE BYTE_ADDR(BYTE_LOOKUP_BASE(WORD_ADDR(next_info)))
#define BL_INFO WORD_ADDR(BYTE_LOOKUP_INFO(WORD_ADDR(next_info)))
#define BL_MIN_BYTE     (BL_BASE[0])
#define BL_MAX_BYTE     (BL_BASE[1])
#define BL_OFFSET(byte) (BL_BASE[2+(byte)-BL_MIN_BYTE])
#define BL_ACTION(byte) (BL_INFO[BL_OFFSET((byte))])

#define STATUS_BUF_SIZE 16
struct from_utf8_mac_status {
    unsigned char buf[STATUS_BUF_SIZE];
    int beg;
    int end;
};
#define buf_empty_p(p) ((p)->beg == (p)->end)
#define buf_bytesize(p) (((p)->end - (p)->beg + STATUS_BUF_SIZE) % STATUS_BUF_SIZE)
#define utf8_trailbyte(c) (((c) & 0xC0) == 0x80)

static void
buf_push(struct from_utf8_mac_status *sp, const unsigned char *p, ssize_t l)
{
    const unsigned char *pend = p + l;
    while (p < pend) {
        /* if (sp->beg == sp->end) */
        sp->buf[sp->end++] = *p++;
        sp->end %= STATUS_BUF_SIZE;
    }
}

static unsigned char
buf_shift(struct from_utf8_mac_status *sp)
{
    /* if (sp->beg == sp->end) */
    unsigned char c = sp->buf[sp->beg++];
    sp->beg %= STATUS_BUF_SIZE;
    return c;
}

static void
buf_clear(struct from_utf8_mac_status *sp)
{
    sp->beg = sp->end = 0;
}

static unsigned char
buf_at(struct from_utf8_mac_status *sp, int pos)
{
    pos += sp->beg;
    pos %= STATUS_BUF_SIZE;
    return sp->buf[pos];
}

static size_t
buf_output_char(struct from_utf8_mac_status *sp, unsigned char *o)
{
    size_t n = 0;
    while (!buf_empty_p(sp)) {
        o[n++] = buf_shift(sp);
        if (!utf8_trailbyte(sp->buf[sp->beg])) break;
    }
    return n;
}

static size_t
buf_output_all(struct from_utf8_mac_status *sp, unsigned char *o)
{
    size_t n = 0;
    while (!buf_empty_p(sp)) {
        o[n++] = buf_shift(sp);
    }
    return n;
}

static VALUE
get_info(VALUE next_info, struct from_utf8_mac_status *sp) {
    int pos = 0;
    while (pos < buf_bytesize(sp)) {
        unsigned char next_byte = buf_at(sp, pos++);
        if (next_byte < BL_MIN_BYTE || BL_MAX_BYTE < next_byte)
            next_info = INVALID;
        else {
            next_info = (VALUE)BL_ACTION(next_byte);
        }
        if ((next_info & 3) == 0) continue;
        break;
    }
    return next_info;
}

static size_t
buf_apply(struct from_utf8_mac_status *sp, unsigned char *o)
{
    size_t n = 0;
    VALUE next_info;
    unsigned char buf[3];
    if (buf_bytesize(sp) < 3 || (buf_bytesize(sp) == 3 && buf_at(sp, 0) >= 0xE0)) {
        /* char length is less than 2 */
        return 0;
    }
    next_info = get_info(from_utf8_mac_nfc2, sp);
    switch (next_info & 0x1F) {
      case THREEbt:
      case TWObt:
        buf[n++] = getBT1(next_info);
        buf[n++] = getBT2(next_info);
        if (THREEbt == (next_info & 0x1F))
            buf[n++] = getBT3(next_info);
        buf_clear(sp);
        buf_push(sp, buf, n);
        return 0;
        break;
      default:
        return buf_output_char(sp, o);
        break;
    }
}

static int
from_utf8_mac_init(void *statep)
{
    struct from_utf8_mac_status *sp = statep;
    buf_clear(sp);
    return 0;
}

static ssize_t
from_utf8_mac_finish(void *statep,
        unsigned char *o, size_t osize)
{
    struct from_utf8_mac_status *sp = statep;
    return buf_output_all(sp, o);
}

static ssize_t
fun_so_from_utf8_mac(void *statep,
        const unsigned char *s, size_t l,
        unsigned char *o, size_t osize)
{
    struct from_utf8_mac_status *sp = statep;
    ssize_t n = 0;

    switch (l) {
      case 1:
        n = from_utf8_mac_finish(sp, o, osize);
        break;
      case 4:
        n = from_utf8_mac_finish(sp, o, osize);
        o[n++] = *s++;
        o[n++] = *s++;
        o[n++] = *s++;
        o[n++] = *s++;
        return n;
    }

    buf_push(sp, s, l);
    n += buf_apply(sp, o);
    return n;
}

static const rb_transcoder
rb_from_UTF8_MAC = {
    "UTF8-MAC", "UTF-8", from_UTF8_MAC,
    TRANSCODE_TABLE_INFO,
    1, /* input_unit_length */
    4, /* max_input */
    10, /* max_output */
    asciicompat_encoder, /* asciicompat_type */
    sizeof(struct from_utf8_mac_status), from_utf8_mac_init, from_utf8_mac_init,
    NULL, NULL, NULL, fun_so_from_utf8_mac,
    from_utf8_mac_finish
};

TRANS_INIT(utf8_mac)
{
<%= transcode_register_code %>
    rb_register_transcoder(&rb_from_UTF8_MAC);
}