/* Mapping tables for JOHAB handling. Copyright (C) 1998-2016 Free Software Foundation, Inc. This file is part of the GNU C Library. Contributed by Jungshik Shin <jshin@pantheon.yale.edu> and Ulrich Drepper <drepper@cygnus.com>, 1998. The GNU C Library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation; either version 2.1 of the License, or (at your option) any later version. The GNU C Library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. You should have received a copy of the GNU Lesser General Public License along with the GNU C Library; if not, see <http://www.gnu.org/licenses/>. */ #include <dlfcn.h> #include <stdint.h> #include <ksc5601.h> /* The table for Bit pattern to Hangul Jamo 5 bits each are used to encode leading consonants(19 + 1 filler), medial vowels(21 + 1 filler) and trailing consonants(27 + 1 filler). KS C 5601-1992 Annex 3 Table 2 0 : Filler, -1: invalid, >= 1 : valid */ static const int init[32] = { -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1 }; static const int mid[32] = { -1, -1, 0, 1, 2, 3, 4, 5, -1, -1, 6, 7, 8, 9, 10, 11, -1, -1, 12, 13, 14, 15, 16, 17, -1, -1, 18, 19, 20, 21, -1, -1 }; static const int final[32] = { -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, -1, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, -1, -1 }; /* Hangul Jamo in Johab to Unicode 2.0 : Unicode 2.0 defines 51 Hangul Compatibility Jamos in the block [0x3131,0x314e] It's to be considered later which Jamo block to use, Compatibility block [0x3131,0x314e] or Hangul Conjoining Jamo block, [0x1100,0x11ff] */ static const uint32_t init_to_ucs[19] = { 0x3131, 0x3132, 0x3134, 0x3137, 0x3138, 0x3139, 0x3141, 0x3142, 0x3143, 0x3145, 0x3146, 0x3147, 0x3148, 0x3149, 0x314a, 0x314b, 0x314c, 0x314d, 0x314e }; static const uint32_t final_to_ucs[31] = { L'\0', L'\0', 0x3133, L'\0', 0x3135, 0x3136, L'\0', L'\0', 0x313a, 0x313b, 0x313c, 0x313d, 0x313e, 0x313f, 0x3140, L'\0', L'\0', 0x3144, L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0', L'\0' }; /* The following three arrays are used to convert precomposed Hangul syllables in [0xac00,0xd???] to Jamo bit patterns for Johab encoding cf. : KS C 5601-1992, Annex3 Table 2 Arrays are used to speed up things although it's possible to get the same result arithmetically. */ static const int init_to_bit[19] = { 0x8800, 0x8c00, 0x9000, 0x9400, 0x9800, 0x9c00, 0xa000, 0xa400, 0xa800, 0xac00, 0xb000, 0xb400, 0xb800, 0xbc00, 0xc000, 0xc400, 0xc800, 0xcc00, 0xd000 }; static const int mid_to_bit[21] = { 0x0060, 0x0080, 0x00a0, 0x00c0, 0x00e0, 0x0140, 0x0160, 0x0180, 0x01a0, 0x01c0, 0x1e0, 0x0240, 0x0260, 0x0280, 0x02a0, 0x02c0, 0x02e0, 0x0340, 0x0360, 0x0380, 0x03a0 }; static const int final_to_bit[28] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d }; /* The conversion table from UCS4 Hangul Compatibility Jamo in [0x3131,0x3163] to Johab cf. 1. KS C 5601-1992 Annex 3 Table 2 2. Unicode 2.0 manual */ static const uint16_t jamo_from_ucs_table[51] = { 0x8841, 0x8c41, 0x8444, 0x9041, 0x8446, 0x8447, 0x9441, 0x9841, 0x9c41, 0x844a, 0x844b, 0x844c, 0x844d, 0x844e, 0x844f, 0x8450, 0xa041, 0xa441, 0xa841, 0x8454, 0xac41, 0xb041, 0xb441, 0xb841, 0xbc41, 0xc041, 0xc441, 0xc841, 0xcc41, 0xd041, 0x8461, 0x8481, 0x84a1, 0x84c1, 0x84e1, 0x8541, 0x8561, 0x8581, 0x85a1, 0x85c1, 0x85e1, 0x8641, 0x8661, 0x8681, 0x86a1, 0x86c1, 0x86e1, 0x8741, 0x8761, 0x8781, 0x87a1 }; static uint32_t johab_sym_hanja_to_ucs (uint_fast32_t idx, uint_fast32_t c1, uint_fast32_t c2) { if (idx <= 0xdefe) return (uint32_t) __ksc5601_sym_to_ucs[(c1 - 0xd9) * 188 + c2 - (c2 > 0x90 ? 0x43 : 0x31)]; else return (uint32_t) __ksc5601_hanja_to_ucs[(c1 - 0xe0) * 188 + c2 - (c2 > 0x90 ? 0x43 : 0x31)]; } /* Definitions used in the body of the `gconv' function. */ #define CHARSET_NAME "JOHAB//" #define FROM_LOOP from_johab #define TO_LOOP to_johab #define DEFINE_INIT 1 #define DEFINE_FINI 1 #define MIN_NEEDED_FROM 1 #define MAX_NEEDED_FROM 2 #define MIN_NEEDED_TO 4 #define ONE_DIRECTION 0 /* First define the conversion function from JOHAB to UCS4. */ #define MIN_NEEDED_INPUT MIN_NEEDED_FROM #define MAX_NEEDED_INPUT MAX_NEEDED_FROM #define MIN_NEEDED_OUTPUT MIN_NEEDED_TO #define LOOPFCT FROM_LOOP #define BODY \ { \ uint32_t ch = *inptr; \ \ if (ch <= 0x7f) \ { \ /* Plain ISO646-KR. */ \ if (ch == 0x5c) \ ch = 0x20a9; /* half-width Korean Currency WON sign */ \ ++inptr; \ } \ /* Johab : 1. Hangul \ 1st byte : 0x84-0xd3 \ 2nd byte : 0x41-0x7e, 0x81-0xfe \ 2. Hanja & Symbol : \ 1st byte : 0xd8-0xde, 0xe0-0xf9 \ 2nd byte : 0x31-0x7e, 0x91-0xfe \ 0xd831-0xd87e and 0xd891-0xd8fe are user-defined area */ \ else \ { \ if (__builtin_expect (ch > 0xf9, 0) \ || __builtin_expect (ch == 0xdf, 0) \ || (__builtin_expect (ch > 0x7e, 0) && ch < 0x84) \ || (__builtin_expect (ch > 0xd3, 0) && ch < 0xd9)) \ { \ /* These are illegal. */ \ STANDARD_FROM_LOOP_ERR_HANDLER (1); \ } \ else \ { \ /* Two-byte character. First test whether the next \ character is also available. */ \ uint32_t ch2; \ uint_fast32_t idx; \ \ if (__glibc_unlikely (inptr + 1 >= inend)) \ { \ /* The second character is not available. Store the \ intermediate result. */ \ result = __GCONV_INCOMPLETE_INPUT; \ break; \ } \ \ ch2 = inptr[1]; \ idx = ch * 256 + ch2; \ if (__glibc_likely (ch <= 0xd3)) \ { \ /* Hangul */ \ int_fast32_t i, m, f; \ \ i = init[(idx & 0x7c00) >> 10]; \ m = mid[(idx & 0x03e0) >> 5]; \ f = final[idx & 0x001f]; \ \ if (__builtin_expect (i == -1, 0) \ || __builtin_expect (m == -1, 0) \ || __builtin_expect (f == -1, 0)) \ { \ /* This is illegal. */ \ STANDARD_FROM_LOOP_ERR_HANDLER (1); \ } \ else if (i > 0 && m > 0) \ ch = ((i - 1) * 21 + (m - 1)) * 28 + f + 0xac00; \ else if (i > 0 && m == 0 && f == 0) \ ch = init_to_ucs[i - 1]; \ else if (i == 0 && m > 0 && f == 0) \ ch = 0x314e + m; /* 0x314f + m - 1 */ \ else if (__builtin_expect ((i | m) == 0, 1) \ && __builtin_expect (f > 0, 1)) \ ch = final_to_ucs[f - 1]; /* round trip?? */ \ else \ { \ /* This is illegal. */ \ STANDARD_FROM_LOOP_ERR_HANDLER (1); \ } \ } \ else \ { \ if (__builtin_expect (ch2 < 0x31, 0) \ || (__builtin_expect (ch2 > 0x7e, 0) && ch2 < 0x91) \ || __builtin_expect (ch2, 0) == 0xff \ || (__builtin_expect (ch, 0) == 0xd9 && ch2 > 0xe8) \ || (__builtin_expect (ch, 0) == 0xda \ && ch2 > 0xa0 && ch2 < 0xd4) \ || (__builtin_expect (ch, 0) == 0xde && ch2 > 0xf1)) \ { \ /* This is illegal. */ \ STANDARD_FROM_LOOP_ERR_HANDLER (1); \ } \ else \ { \ ch = johab_sym_hanja_to_ucs (idx, ch, ch2); \ /* if (idx <= 0xdefe) \ ch = __ksc5601_sym_to_ucs[(ch - 0xd9) * 192 \ + ch2 - (ch2 > 0x90 \ ? 0x43 : 0x31)]; \ else \ ch = __ksc5601_hanja_to_ucs[(ch - 0xe0) *192 \ + ch2 - (ch2 > 0x90 \ ?0x43 : 0x31)];\ */ \ } \ } \ } \ \ if (__glibc_unlikely (ch == 0)) \ { \ /* This is an illegal character. */ \ STANDARD_FROM_LOOP_ERR_HANDLER (2); \ } \ \ inptr += 2; \ } \ \ put32 (outptr, ch); \ outptr += 4; \ } #define LOOP_NEED_FLAGS #define ONEBYTE_BODY \ { \ if (c <= 0x7f) \ return (c == 0x5c ? 0x20a9 : c); \ else \ return WEOF; \ } #include <iconv/loop.c> /* Next, define the other direction. */ #define MIN_NEEDED_INPUT MIN_NEEDED_TO #define MIN_NEEDED_OUTPUT MIN_NEEDED_FROM #define MAX_NEEDED_OUTPUT MAX_NEEDED_FROM #define LOOPFCT TO_LOOP #define BODY \ { \ uint32_t ch = get32 (inptr); \ /* \ if (ch >= (sizeof (from_ucs4_lat1) / sizeof (from_ucs4_lat1[0]))) \ { \ if (ch >= 0x0391 && ch <= 0x0451) \ cp = from_ucs4_greek[ch - 0x391]; \ else if (ch >= 0x2010 && ch <= 0x9fa0) \ cp = from_ucs4_cjk[ch - 0x02010]; \ else \ break; \ } \ else \ cp = from_ucs4_lat1[ch]; \ */ \ \ if (ch <= 0x7f && ch != 0x5c) \ *outptr++ = ch; \ else \ { \ if (ch >= 0xac00 && ch <= 0xd7a3) \ { \ if (__glibc_unlikely (outptr + 2 > outend)) \ { \ result = __GCONV_FULL_OUTPUT; \ break; \ } \ \ ch -= 0xac00; \ \ ch = (init_to_bit[ch / 588] /* 21 * 28 = 588 */ \ + mid_to_bit[(ch / 28) % 21]/* (ch % (21 * 28)) / 28 */ \ + final_to_bit[ch % 28]); /* (ch % (21 * 28)) % 28 */ \ \ *outptr++ = ch / 256; \ *outptr++ = ch % 256; \ } \ /* KS C 5601-1992 Annex 3 regards 0xA4DA(Hangul Filler : U3164) \ as symbol */ \ else if (ch >= 0x3131 && ch <= 0x3163) \ { \ ch = jamo_from_ucs_table[ch - 0x3131]; \ \ if (__glibc_unlikely (outptr + 2 > outend)) \ { \ result = __GCONV_FULL_OUTPUT; \ break; \ } \ \ *outptr++ = ch / 256; \ *outptr++ = ch % 256; \ } \ else if ((ch >= 0x4e00 && ch <= 0x9fa5) \ || (ch >= 0xf900 && ch <= 0xfa0b)) \ { \ size_t written; \ uint32_t temp; \ \ written = ucs4_to_ksc5601_hanja (ch, outptr, outend - outptr); \ if (__builtin_expect (written, 1) == 0) \ { \ result = __GCONV_FULL_OUTPUT; \ break; \ } \ if (__glibc_unlikely (written == __UNKNOWN_10646_CHAR)) \ { \ STANDARD_TO_LOOP_ERR_HANDLER (4); \ } \ \ outptr[0] -= 0x4a; \ outptr[1] -= 0x21; \ \ temp = outptr[0] * 94 + outptr[1]; \ \ outptr[0] = 0xe0 + temp / 188; \ outptr[1] = temp % 188; \ outptr[1] += outptr[1] >= 78 ? 0x43 : 0x31; \ \ outptr += 2; \ } \ else if (ch == 0x20a9) \ *outptr++ = 0x5c; \ else \ { \ size_t written; \ uint32_t temp; \ \ written = ucs4_to_ksc5601_sym (ch, outptr, outend - outptr); \ if (__builtin_expect (written, 1) == 0) \ { \ result = __GCONV_FULL_OUTPUT; \ break; \ } \ if (__builtin_expect (written == __UNKNOWN_10646_CHAR, 0) \ || (outptr[0] == 0x22 && outptr[1] > 0x68)) \ { \ UNICODE_TAG_HANDLER (ch, 4); \ STANDARD_TO_LOOP_ERR_HANDLER (4); \ } \ \ temp = (outptr[0] < 0x4a ? outptr[0] + 0x191 : outptr[0] + 0x176);\ outptr[1] += (temp % 2 ? 0x5e : 0); \ outptr[1] += (outptr[1] < 0x6f ? 0x10 : 0x22); \ outptr[0] = temp / 2; \ \ outptr += 2; \ } \ } \ \ inptr += 4; \ } #define LOOP_NEED_FLAGS #include <iconv/loop.c> /* Now define the toplevel functions. */ #include <iconv/skeleton.c>