X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=src%2Fsiconv.c;h=26f36780d449a36e9223abe6b79ed5ce4c4df4a5;hb=be821514c869d68186361b5aab6bbfd1aa60e087;hp=a10fd3d26785a8bccc20a79b2e82a8f4682b75e7;hpb=af30aafd05572139098f984ddf13ba12072afa47;p=yaz-moved-to-github.git diff --git a/src/siconv.c b/src/siconv.c index a10fd3d..26f3678 100644 --- a/src/siconv.c +++ b/src/siconv.c @@ -1,16 +1,27 @@ /* - * Copyright (c) 1997-2003, Index Data + * Copyright (C) 1995-2007, Index Data ApS * See the file LICENSE for details. * - * $Id: siconv.c,v 1.2 2004-03-11 10:09:11 oleg Exp $ + * $Id: siconv.c,v 1.37 2007-03-20 21:37:32 adam Exp $ + */ +/** + * \file siconv.c + * \brief Implements simple ICONV + * + * This implements an interface similar to that of iconv and + * is used by YAZ to interface with iconv (if present). + * For systems where iconv is not present, this layer + * provides a few important conversions: UTF-8, MARC-8, Latin-1. + * + * MARC-8 reference: + * http://www.loc.gov/marc/specifications/speccharmarc8.html */ - -/* mini iconv and wrapper for system iconv library (if present) */ #if HAVE_CONFIG_H #include #endif +#include #include #include #include @@ -22,11 +33,48 @@ #include #endif + #include -unsigned long yaz_marc8_conv (unsigned char *inp, size_t inbytesleft, - size_t *no_read); - +unsigned long yaz_marc8_1_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_2_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_3_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_4_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_5_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_6_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_7_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_8_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8_9_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); + + +unsigned long yaz_marc8r_1_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_2_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_3_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_4_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_5_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_6_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_7_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_8_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); +unsigned long yaz_marc8r_9_conv(unsigned char *inp, size_t inbytesleft, + size_t *no_read, int *combining); + struct yaz_iconv_struct { int my_errno; int init_flag; @@ -36,9 +84,97 @@ struct yaz_iconv_struct { size_t inbytesleft, size_t *no_read); size_t (*write_handle)(yaz_iconv_t cd, unsigned long x, char **outbuf, size_t *outbytesleft); + size_t (*flush_handle)(yaz_iconv_t cd, + char **outbuf, size_t *outbytesleft); + int marc8_esc_mode; + + int comb_offset; + int comb_size; + unsigned long comb_x[8]; + size_t comb_no_read[8]; + size_t no_read_x; + unsigned long unget_x; #if HAVE_ICONV_H iconv_t iconv_cd; #endif + unsigned long compose_char; + + unsigned long write_marc8_comb_ch[8]; + size_t write_marc8_comb_no; + unsigned write_marc8_second_half_char; + unsigned long write_marc8_last; + const char *write_marc8_page_chr; +}; + +static struct { + unsigned long x1, x2; + unsigned y; +} latin1_comb[] = { + { 'A', 0x0300, 0xc0}, /* LATIN CAPITAL LETTER A WITH GRAVE */ + { 'A', 0x0301, 0xc1}, /* LATIN CAPITAL LETTER A WITH ACUTE */ + { 'A', 0x0302, 0xc2}, /* LATIN CAPITAL LETTER A WITH CIRCUMFLEX */ + { 'A', 0x0303, 0xc3}, /* LATIN CAPITAL LETTER A WITH TILDE */ + { 'A', 0x0308, 0xc4}, /* LATIN CAPITAL LETTER A WITH DIAERESIS */ + { 'A', 0x030a, 0xc5}, /* LATIN CAPITAL LETTER A WITH RING ABOVE */ + /* no need for 0xc6 LATIN CAPITAL LETTER AE */ + { 'C', 0x0327, 0xc7}, /* LATIN CAPITAL LETTER C WITH CEDILLA */ + { 'E', 0x0300, 0xc8}, /* LATIN CAPITAL LETTER E WITH GRAVE */ + { 'E', 0x0301, 0xc9}, /* LATIN CAPITAL LETTER E WITH ACUTE */ + { 'E', 0x0302, 0xca}, /* LATIN CAPITAL LETTER E WITH CIRCUMFLEX */ + { 'E', 0x0308, 0xcb}, /* LATIN CAPITAL LETTER E WITH DIAERESIS */ + { 'I', 0x0300, 0xcc}, /* LATIN CAPITAL LETTER I WITH GRAVE */ + { 'I', 0x0301, 0xcd}, /* LATIN CAPITAL LETTER I WITH ACUTE */ + { 'I', 0x0302, 0xce}, /* LATIN CAPITAL LETTER I WITH CIRCUMFLEX */ + { 'I', 0x0308, 0xcf}, /* LATIN CAPITAL LETTER I WITH DIAERESIS */ + { 'N', 0x0303, 0xd1}, /* LATIN CAPITAL LETTER N WITH TILDE */ + { 'O', 0x0300, 0xd2}, /* LATIN CAPITAL LETTER O WITH GRAVE */ + { 'O', 0x0301, 0xd3}, /* LATIN CAPITAL LETTER O WITH ACUTE */ + { 'O', 0x0302, 0xd4}, /* LATIN CAPITAL LETTER O WITH CIRCUMFLEX */ + { 'O', 0x0303, 0xd5}, /* LATIN CAPITAL LETTER O WITH TILDE */ + { 'O', 0x0308, 0xd6}, /* LATIN CAPITAL LETTER O WITH DIAERESIS */ + /* omitted: 0xd7 MULTIPLICATION SIGN */ + /* omitted: 0xd8 LATIN CAPITAL LETTER O WITH STROKE */ + { 'U', 0x0300, 0xd9}, /* LATIN CAPITAL LETTER U WITH GRAVE */ + { 'U', 0x0301, 0xda}, /* LATIN CAPITAL LETTER U WITH ACUTE */ + { 'U', 0x0302, 0xdb}, /* LATIN CAPITAL LETTER U WITH CIRCUMFLEX */ + { 'U', 0x0308, 0xdc}, /* LATIN CAPITAL LETTER U WITH DIAERESIS */ + { 'Y', 0x0301, 0xdd}, /* LATIN CAPITAL LETTER Y WITH ACUTE */ + /* omitted: 0xde LATIN CAPITAL LETTER THORN */ + /* omitted: 0xdf LATIN SMALL LETTER SHARP S */ + { 'a', 0x0300, 0xe0}, /* LATIN SMALL LETTER A WITH GRAVE */ + { 'a', 0x0301, 0xe1}, /* LATIN SMALL LETTER A WITH ACUTE */ + { 'a', 0x0302, 0xe2}, /* LATIN SMALL LETTER A WITH CIRCUMFLEX */ + { 'a', 0x0303, 0xe3}, /* LATIN SMALL LETTER A WITH TILDE */ + { 'a', 0x0308, 0xe4}, /* LATIN SMALL LETTER A WITH DIAERESIS */ + { 'a', 0x030a, 0xe5}, /* LATIN SMALL LETTER A WITH RING ABOVE */ + /* omitted: 0xe6 LATIN SMALL LETTER AE */ + { 'c', 0x0327, 0xe7}, /* LATIN SMALL LETTER C WITH CEDILLA */ + { 'e', 0x0300, 0xe8}, /* LATIN SMALL LETTER E WITH GRAVE */ + { 'e', 0x0301, 0xe9}, /* LATIN SMALL LETTER E WITH ACUTE */ + { 'e', 0x0302, 0xea}, /* LATIN SMALL LETTER E WITH CIRCUMFLEX */ + { 'e', 0x0308, 0xeb}, /* LATIN SMALL LETTER E WITH DIAERESIS */ + { 'i', 0x0300, 0xec}, /* LATIN SMALL LETTER I WITH GRAVE */ + { 'i', 0x0301, 0xed}, /* LATIN SMALL LETTER I WITH ACUTE */ + { 'i', 0x0302, 0xee}, /* LATIN SMALL LETTER I WITH CIRCUMFLEX */ + { 'i', 0x0308, 0xef}, /* LATIN SMALL LETTER I WITH DIAERESIS */ + /* omitted: 0xf0 LATIN SMALL LETTER ETH */ + { 'n', 0x0303, 0xf1}, /* LATIN SMALL LETTER N WITH TILDE */ + { 'o', 0x0300, 0xf2}, /* LATIN SMALL LETTER O WITH GRAVE */ + { 'o', 0x0301, 0xf3}, /* LATIN SMALL LETTER O WITH ACUTE */ + { 'o', 0x0302, 0xf4}, /* LATIN SMALL LETTER O WITH CIRCUMFLEX */ + { 'o', 0x0303, 0xf5}, /* LATIN SMALL LETTER O WITH TILDE */ + { 'o', 0x0308, 0xf6}, /* LATIN SMALL LETTER O WITH DIAERESIS */ + /* omitted: 0xf7 DIVISION SIGN */ + /* omitted: 0xf8 LATIN SMALL LETTER O WITH STROKE */ + { 'u', 0x0300, 0xf9}, /* LATIN SMALL LETTER U WITH GRAVE */ + { 'u', 0x0301, 0xfa}, /* LATIN SMALL LETTER U WITH ACUTE */ + { 'u', 0x0302, 0xfb}, /* LATIN SMALL LETTER U WITH CIRCUMFLEX */ + { 'u', 0x0308, 0xfc}, /* LATIN SMALL LETTER U WITH DIAERESIS */ + { 'y', 0x0301, 0xfd}, /* LATIN SMALL LETTER Y WITH ACUTE */ + /* omitted: 0xfe LATIN SMALL LETTER THORN */ + { 'y', 0x0308, 0xff}, /* LATIN SMALL LETTER Y WITH DIAERESIS */ + + { 0, 0, 0} }; static unsigned long yaz_read_ISO8859_1 (yaz_iconv_t cd, unsigned char *inp, @@ -49,6 +185,7 @@ static unsigned long yaz_read_ISO8859_1 (yaz_iconv_t cd, unsigned char *inp, return x; } + static size_t yaz_init_UTF8 (yaz_iconv_t cd, unsigned char *inp, size_t inbytesleft, size_t *no_read) { @@ -62,17 +199,16 @@ static size_t yaz_init_UTF8 (yaz_iconv_t cd, unsigned char *inp, cd->my_errno = YAZ_ICONV_EINVAL; return (size_t) -1; } - if (inp[1] != 0xbb || inp[2] != 0xbf) - { - cd->my_errno = YAZ_ICONV_EILSEQ; - return (size_t) -1; - } - *no_read = 3; + if (inp[1] != 0xbb && inp[2] == 0xbf) + *no_read = 3; + else + *no_read = 0; return 0; } -static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, - size_t inbytesleft, size_t *no_read) +unsigned long yaz_read_UTF8_char(unsigned char *inp, + size_t inbytesleft, size_t *no_read, + int *error) { unsigned long x = 0; @@ -84,7 +220,7 @@ static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, else if (inp[0] <= 0xbf || inp[0] >= 0xfe) { *no_read = 0; - cd->my_errno = YAZ_ICONV_EILSEQ; + *error = YAZ_ICONV_EILSEQ; } else if (inp[0] <= 0xdf && inbytesleft >= 2) { @@ -94,19 +230,19 @@ static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, else { *no_read = 0; - cd->my_errno = YAZ_ICONV_EILSEQ; + *error = YAZ_ICONV_EILSEQ; } } else if (inp[0] <= 0xef && inbytesleft >= 3) { x = ((inp[0] & 0x0f) << 12) | ((inp[1] & 0x3f) << 6) | - (inp[1] & 0x3f); + (inp[2] & 0x3f); if (x >= 0x800) *no_read = 3; else { *no_read = 0; - cd->my_errno = YAZ_ICONV_EILSEQ; + *error = YAZ_ICONV_EILSEQ; } } else if (inp[0] <= 0xf7 && inbytesleft >= 4) @@ -118,7 +254,7 @@ static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, else { *no_read = 0; - cd->my_errno = YAZ_ICONV_EILSEQ; + *error = YAZ_ICONV_EILSEQ; } } else if (inp[0] <= 0xfb && inbytesleft >= 5) @@ -131,7 +267,7 @@ static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, else { *no_read = 0; - cd->my_errno = YAZ_ICONV_EILSEQ; + *error = YAZ_ICONV_EILSEQ; } } else if (inp[0] <= 0xfd && inbytesleft >= 6) @@ -144,17 +280,23 @@ static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, else { *no_read = 0; - cd->my_errno = YAZ_ICONV_EILSEQ; + *error = YAZ_ICONV_EILSEQ; } } else { *no_read = 0; - cd->my_errno = YAZ_ICONV_EINVAL; + *error = YAZ_ICONV_EINVAL; } return x; } +static unsigned long yaz_read_UTF8 (yaz_iconv_t cd, unsigned char *inp, + size_t inbytesleft, size_t *no_read) +{ + return yaz_read_UTF8_char(inp, inbytesleft, no_read, &cd->my_errno); +} + static unsigned long yaz_read_UCS4 (yaz_iconv_t cd, unsigned char *inp, size_t inbytesleft, size_t *no_read) { @@ -213,16 +355,525 @@ static unsigned long yaz_read_wchar_t (yaz_iconv_t cd, unsigned char *inp, } #endif +static unsigned long yaz_read_advancegreek(yaz_iconv_t cd, unsigned char *inp, + size_t inbytesleft, size_t *no_read) +{ + unsigned long x = 0; + int shift = 0; + int tonos = 0; + int dialitika = 0; + + *no_read = 0; + while (inbytesleft > 0) + { + if (*inp == 0x9d) + { + tonos = 1; + } + else if (*inp == 0x9e) + { + dialitika = 1; + } + else if (*inp == 0x9f) + { + shift = 1; + } + else + break; + inp++; + --inbytesleft; + (*no_read)++; + } + if (inbytesleft == 0) + { + cd->my_errno = YAZ_ICONV_EINVAL; /* incomplete input */ + *no_read = 0; + return 0; + } + switch (*inp) { + case 0x81: + if (shift) + if (tonos) + x = 0x0386; + else + x = 0x0391; + else + if (tonos) + x = 0x03ac; + else + x = 0x03b1; + break; + case 0x82: + if (shift) + x = 0x0392; + else + x = 0x03b2; + + break; + case 0x83: + if (shift) + x = 0x0393; + else + x = 0x03b3; + break; + case 0x84: + if (shift) + x = 0x0394; + else + x = 0x03b4; + break; + case 0x85: + if (shift) + if (tonos) + x = 0x0388; + else + x = 0x0395; + else + if (tonos) + x = 0x03ad; + else + x = 0x03b5; + break; + case 0x86: + if (shift) + x = 0x0396; + else + x = 0x03b6; + break; + case 0x87: + if (shift) + if (tonos) + x = 0x0389; + else + x = 0x0397; + else + if (tonos) + x = 0x03ae; + else + x = 0x03b7; + break; + case 0x88: + if (shift) + x = 0x0398; + else + x = 0x03b8; + break; + case 0x89: + if (shift) + if (tonos) + x = 0x038a; + else + if (dialitika) + x = 0x9e; + else + x = 0x0399; + else + if (tonos) + if (dialitika) + x = 0x0390; + else + x = 0x03af; + + else + if (dialitika) + x = 0x03ca; + else + x = 0x03b9; + break; + case 0x8a: + if (shift) + x = 0x039a; + else + x = 0x03ba; + + break; + case 0x8b: + if (shift) + x = 0x039b; + else + x = 0x03bb; + break; + case 0x8c: + if (shift) + x = 0x039c; + else + x = 0x03bc; + + break; + case 0x8d: + if (shift) + x = 0x039d; + else + x = 0x03bd; + break; + case 0x8e: + if (shift) + x = 0x039e; + else + x = 0x03be; + break; + case 0x8f: + if (shift) + if (tonos) + x = 0x038c; + else + x = 0x039f; + else + if (tonos) + x = 0x03cc; + else + x = 0x03bf; + break; + case 0x90: + if (shift) + x = 0x03a0; + else + x = 0x03c0; + break; + case 0x91: + if (shift) + x = 0x03a1; + else + x = 0x03c1; + break; + case 0x92: + x = 0x03c2; + break; + case 0x93: + if (shift) + x = 0x03a3; + else + x = 0x03c3; + break; + case 0x94: + if (shift) + x = 0x03a4; + else + x = 0x03c4; + break; + case 0x95: + if (shift) + if (tonos) + x = 0x038e; + else + if (dialitika) + x = 0x03ab; + else + x = 0x03a5; + else + if (tonos) + if (dialitika) + x = 0x03b0; + else + x = 0x03cd; + + else + if (dialitika) + x = 0x03cb; + else + x = 0x03c5; + break; + case 0x96: + if (shift) + x = 0x03a6; + else + x = 0x03c6; + break; + case 0x97: + if (shift) + x = 0x03a7; + else + x = 0x03c7; + break; + case 0x98: + if (shift) + x = 0x03a8; + else + x = 0x03c8; + + break; + + case 0x99: + if (shift) + if (tonos) + x = 0x038f; + else + x = 0x03a9; + else + if (tonos) + x = 0x03ce; + else + x = 0x03c9; + break; + default: + x = *inp; + break; + } + (*no_read)++; + + return x; +} + +static size_t yaz_write_advancegreek(yaz_iconv_t cd, unsigned long x, + char **outbuf, size_t *outbytesleft) +{ + size_t k = 0; + unsigned char *out = (unsigned char*) *outbuf; + if (*outbytesleft < 3) + { + cd->my_errno = YAZ_ICONV_E2BIG; /* not room for output */ + return (size_t)(-1); + } + switch (x) + { + case 0x03ac : out[k++]=0x9d; out[k++]=0x81; break; + case 0x03ad : out[k++]=0x9d; out[k++]=0x85; break; + case 0x03ae : out[k++]=0x9d; out[k++]=0x87; break; + case 0x03af : out[k++]=0x9d; out[k++]=0x89; break; + case 0x03cc : out[k++]=0x9d; out[k++]=0x8f; break; + case 0x03cd : out[k++]=0x9d; out[k++]=0x95; break; + case 0x03ce : out[k++]=0x9d; out[k++]=0x99; break; + case 0x0390 : out[k++]=0x9d; out[k++]=0x9e; out[k++]=0x89; break; + case 0x03b0 : out[k++]=0x9d; out[k++]=0x9e; out[k++]=0x95; break; + case 0x0386 : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x81; break; + case 0x0388 : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x85; break; + case 0x0389 : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x87; break; + case 0x038a : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x89; break; + case 0x038c : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x8f; break; + case 0x038e : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x95; break; + case 0x038f : out[k++]=0x9d; out[k++]=0x9f; out[k++]=0x99; break; + case 0x03ca : out[k++]=0x9e; out[k++]=0x89; break; + case 0x03cb : out[k++]=0x9e; out[k++]=0x95; break; + case 0x03aa : out[k++]=0x9e; out[k++]=0x9f; out[k++]=0x89; break; + case 0x03ab : out[k++]=0x9e; out[k++]=0x9f; out[k++]=0x95; break; + case 0x0391 : out[k++]=0x9f; out[k++]=0x81; break; + case 0x0392 : out[k++]=0x9f; out[k++]=0x82; break; + case 0x0393 : out[k++]=0x9f; out[k++]=0x83; break; + case 0x0394 : out[k++]=0x9f; out[k++]=0x84; break; + case 0x0395 : out[k++]=0x9f; out[k++]=0x85; break; + case 0x0396 : out[k++]=0x9f; out[k++]=0x86; break; + case 0x0397 : out[k++]=0x9f; out[k++]=0x87; break; + case 0x0398 : out[k++]=0x9f; out[k++]=0x88; break; + case 0x0399 : out[k++]=0x9f; out[k++]=0x89; break; + case 0x039a : out[k++]=0x9f; out[k++]=0x8a; break; + case 0x039b : out[k++]=0x9f; out[k++]=0x8b; break; + case 0x039c : out[k++]=0x9f; out[k++]=0x8c; break; + case 0x039d : out[k++]=0x9f; out[k++]=0x8d; break; + case 0x039e : out[k++]=0x9f; out[k++]=0x8e; break; + case 0x039f : out[k++]=0x9f; out[k++]=0x8f; break; + case 0x03a0 : out[k++]=0x9f; out[k++]=0x90; break; + case 0x03a1 : out[k++]=0x9f; out[k++]=0x91; break; + case 0x03a3 : out[k++]=0x9f; out[k++]=0x93; break; + case 0x03a4 : out[k++]=0x9f; out[k++]=0x94; break; + case 0x03a5 : out[k++]=0x9f; out[k++]=0x95; break; + case 0x03a6 : out[k++]=0x9f; out[k++]=0x96; break; + case 0x03a7 : out[k++]=0x9f; out[k++]=0x97; break; + case 0x03a8 : out[k++]=0x9f; out[k++]=0x98; break; + case 0x03a9 : out[k++]=0x9f; out[k++]=0x99; break; + case 0x03b1 : out[k++]=0x81; break; + case 0x03b2 : out[k++]=0x82; break; + case 0x03b3 : out[k++]=0x83; break; + case 0x03b4 : out[k++]=0x84; break; + case 0x03b5 : out[k++]=0x85; break; + case 0x03b6 : out[k++]=0x86; break; + case 0x03b7 : out[k++]=0x87; break; + case 0x03b8 : out[k++]=0x88; break; + case 0x03b9 : out[k++]=0x89; break; + case 0x03ba : out[k++]=0x8a; break; + case 0x03bb : out[k++]=0x8b; break; + case 0x03bc : out[k++]=0x8c; break; + case 0x03bd : out[k++]=0x8d; break; + case 0x03be : out[k++]=0x8e; break; + case 0x03bf : out[k++]=0x8f; break; + case 0x03c0 : out[k++]=0x90; break; + case 0x03c1 : out[k++]=0x91; break; + case 0x03c2 : out[k++]=0x92; break; + case 0x03c3 : out[k++]=0x93; break; + case 0x03c4 : out[k++]=0x94; break; + case 0x03c5 : out[k++]=0x95; break; + case 0x03c6 : out[k++]=0x96; break; + case 0x03c7 : out[k++]=0x96; break; + case 0x03c8 : out[k++]=0x98; break; + case 0x03c9 : out[k++]=0x99; break; + default: + if (x > 255) + { + cd->my_errno = YAZ_ICONV_EILSEQ; + return (size_t) -1; + } + out[k++] = x; + break; + } + *outbytesleft -= k; + (*outbuf) += k; + return 0; +} + + +static unsigned long yaz_read_marc8_comb (yaz_iconv_t cd, unsigned char *inp, + size_t inbytesleft, size_t *no_read, + int *comb); + static unsigned long yaz_read_marc8 (yaz_iconv_t cd, unsigned char *inp, size_t inbytesleft, size_t *no_read) { - return yaz_marc8_conv(inp, inbytesleft, no_read); + unsigned long x; + if (cd->comb_offset < cd->comb_size) + { + *no_read = cd->comb_no_read[cd->comb_offset]; + x = cd->comb_x[cd->comb_offset]; + + /* special case for double-diacritic combining characters, + INVERTED BREVE and DOUBLE TILDE. + We'll increment the no_read counter by 1, since we want to skip over + the processing of the closing ligature character + */ + /* this code is no longer necessary.. our handlers code in + yaz_marc8_?_conv (generated by charconv.tcl) now returns + 0 and no_read=1 when a sequence does not match the input. + The SECOND HALFs in codetables.xml produces a non-existant + entry in the conversion trie.. Hence when met, the input byte is + skipped as it should (in yaz_iconv) + */ +#if 0 + if (x == 0x0361 || x == 0x0360) + *no_read += 1; +#endif + cd->comb_offset++; + return x; + } + + cd->comb_offset = 0; + for (cd->comb_size = 0; cd->comb_size < 8; cd->comb_size++) + { + int comb = 0; + + if (inbytesleft == 0 && cd->comb_size) + { + cd->my_errno = YAZ_ICONV_EINVAL; + x = 0; + *no_read = 0; + break; + } + x = yaz_read_marc8_comb(cd, inp, inbytesleft, no_read, &comb); + if (!comb || !x) + break; + cd->comb_x[cd->comb_size] = x; + cd->comb_no_read[cd->comb_size] = *no_read; + inp += *no_read; + inbytesleft = inbytesleft - *no_read; + } + return x; } -static size_t yaz_write_UTF8 (yaz_iconv_t cd, unsigned long x, - char **outbuf, size_t *outbytesleft) +static unsigned long yaz_read_marc8s(yaz_iconv_t cd, unsigned char *inp, + size_t inbytesleft, size_t *no_read) +{ + unsigned long x = yaz_read_marc8(cd, inp, inbytesleft, no_read); + if (x && cd->comb_size == 1) + { + /* For MARC8s we try to get a Latin-1 page code out of it */ + int i; + for (i = 0; latin1_comb[i].x1; i++) + if (cd->comb_x[0] == latin1_comb[i].x2 && x == latin1_comb[i].x1) + { + *no_read += cd->comb_no_read[0]; + cd->comb_size = 0; + x = latin1_comb[i].y; + break; + } + } + return x; +} + +static unsigned long yaz_read_marc8_comb(yaz_iconv_t cd, unsigned char *inp, + size_t inbytesleft, size_t *no_read, + int *comb) +{ + *no_read = 0; + while(inbytesleft >= 1 && inp[0] == 27) + { + size_t inbytesleft0 = inbytesleft; + inp++; + inbytesleft--; + while(inbytesleft > 0 && strchr("(,$!)-", *inp)) + { + inbytesleft--; + inp++; + } + if (inbytesleft <= 0) + { + *no_read = 0; + cd->my_errno = YAZ_ICONV_EINVAL; + return 0; + } + cd->marc8_esc_mode = *inp++; + inbytesleft--; + (*no_read) += inbytesleft0 - inbytesleft; + } + if (inbytesleft <= 0) + return 0; + else + { + unsigned long x; + size_t no_read_sub = 0; + *comb = 0; + + switch(cd->marc8_esc_mode) + { + case 'B': /* Basic ASCII */ + case 'E': /* ANSEL */ + case 's': /* ASCII */ + x = yaz_marc8_1_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case 'g': /* Greek */ + x = yaz_marc8_2_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case 'b': /* Subscripts */ + x = yaz_marc8_3_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case 'p': /* Superscripts */ + x = yaz_marc8_4_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case '2': /* Basic Hebrew */ + x = yaz_marc8_5_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case 'N': /* Basic Cyrillic */ + case 'Q': /* Extended Cyrillic */ + x = yaz_marc8_6_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case '3': /* Basic Arabic */ + case '4': /* Extended Arabic */ + x = yaz_marc8_7_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case 'S': /* Greek */ + x = yaz_marc8_8_conv(inp, inbytesleft, &no_read_sub, comb); + break; + case '1': /* Chinese, Japanese, Korean (EACC) */ + x = yaz_marc8_9_conv(inp, inbytesleft, &no_read_sub, comb); + break; + default: + *no_read = 0; + cd->my_errno = YAZ_ICONV_EILSEQ; + return 0; + } + *no_read += no_read_sub; + return x; + } +} + +static size_t yaz_write_UTF8(yaz_iconv_t cd, unsigned long x, + char **outbuf, size_t *outbytesleft) +{ + return yaz_write_UTF8_char(x, outbuf, outbytesleft, &cd->my_errno); +} + +size_t yaz_write_UTF8_char(unsigned long x, + char **outbuf, size_t *outbytesleft, + int *error) { unsigned char *outp = (unsigned char *) *outbuf; + if (x <= 0x7f && *outbytesleft >= 1) { *outp++ = (unsigned char) x; @@ -270,7 +921,7 @@ static size_t yaz_write_UTF8 (yaz_iconv_t cd, unsigned long x, } else { - cd->my_errno = YAZ_ICONV_E2BIG; /* not room for output */ + *error = YAZ_ICONV_E2BIG; /* not room for output */ return (size_t)(-1); } *outbuf = (char *) outp; @@ -280,26 +931,78 @@ static size_t yaz_write_UTF8 (yaz_iconv_t cd, unsigned long x, static size_t yaz_write_ISO8859_1 (yaz_iconv_t cd, unsigned long x, char **outbuf, size_t *outbytesleft) { + /* list of two char unicode sequence that, when combined, are + equivalent to single unicode chars that can be represented in + ISO-8859-1/Latin-1. + Regular iconv on Linux at least does not seem to convert these, + but since MARC-8 to UTF-8 generates these composed sequence + we get a better chance of a successful MARC-8 -> ISO-8859-1 + conversion */ unsigned char *outp = (unsigned char *) *outbuf; - if (x > 255 || x < 1) + + if (cd->compose_char) { - cd->my_errno = YAZ_ICONV_EILSEQ; - return (size_t) -1; + int i; + for (i = 0; latin1_comb[i].x1; i++) + if (cd->compose_char == latin1_comb[i].x1 && x == latin1_comb[i].x2) + { + x = latin1_comb[i].y; + break; + } + if (*outbytesleft < 1) + { /* no room. Retain compose_char and bail out */ + cd->my_errno = YAZ_ICONV_E2BIG; + return (size_t)(-1); + } + if (!latin1_comb[i].x1) + { /* not found. Just write compose_char */ + *outp++ = (unsigned char) cd->compose_char; + (*outbytesleft)--; + *outbuf = (char *) outp; + } + /* compose_char used so reset it. x now holds current char */ + cd->compose_char = 0; } - else if (*outbytesleft >= 1) + + if (x > 32 && x < 127 && cd->compose_char == 0) { - *outp++ = (unsigned char) x; - (*outbytesleft)--; + cd->compose_char = x; + return 0; } - else + else if (x > 255 || x < 1) + { + cd->my_errno = YAZ_ICONV_EILSEQ; + return (size_t) -1; + } + else if (*outbytesleft < 1) { cd->my_errno = YAZ_ICONV_E2BIG; return (size_t)(-1); } + *outp++ = (unsigned char) x; + (*outbytesleft)--; *outbuf = (char *) outp; return 0; } +static size_t yaz_flush_ISO8859_1(yaz_iconv_t cd, + char **outbuf, size_t *outbytesleft) +{ + if (cd->compose_char) + { + unsigned char *outp = (unsigned char *) *outbuf; + if (*outbytesleft < 1) + { + cd->my_errno = YAZ_ICONV_E2BIG; + return (size_t)(-1); + } + *outp++ = (unsigned char) cd->compose_char; + (*outbytesleft)--; + *outbuf = (char *) outp; + cd->compose_char = 0; + } + return 0; +} static size_t yaz_write_UCS4 (yaz_iconv_t cd, unsigned long x, char **outbuf, size_t *outbytesleft) @@ -343,9 +1046,261 @@ static size_t yaz_write_UCS4LE (yaz_iconv_t cd, unsigned long x, return 0; } +static unsigned long lookup_marc8(yaz_iconv_t cd, + unsigned long x, int *comb, + const char **page_chr) +{ + char utf8_buf[7]; + char *utf8_outbuf = utf8_buf; + size_t utf8_outbytesleft = sizeof(utf8_buf)-1, r; + + r = yaz_write_UTF8(cd, x, &utf8_outbuf, &utf8_outbytesleft); + if (r == (size_t)(-1)) + { + cd->my_errno = YAZ_ICONV_EILSEQ; + return 0; + } + else + { + unsigned char *inp; + size_t inbytesleft, no_read_sub = 0; + unsigned long x; + + *utf8_outbuf = '\0'; + inp = (unsigned char *) utf8_buf; + inbytesleft = strlen(utf8_buf); + + x = yaz_marc8r_1_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033(B"; + return x; + } + x = yaz_marc8r_2_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033g"; + return x; + } + x = yaz_marc8r_3_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033b"; + return x; + } + x = yaz_marc8r_4_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033p"; + return x; + } + x = yaz_marc8r_5_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033(2"; + return x; + } + x = yaz_marc8r_6_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033(N"; + return x; + } + x = yaz_marc8r_7_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033(3"; + return x; + } + x = yaz_marc8r_8_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033(S"; + return x; + } + x = yaz_marc8r_9_conv(inp, inbytesleft, &no_read_sub, comb); + if (x) + { + *page_chr = "\033$1"; + return x; + } + cd->my_errno = YAZ_ICONV_EILSEQ; + return x; + } +} + +static size_t flush_combos(yaz_iconv_t cd, + char **outbuf, size_t *outbytesleft) +{ + unsigned long y = cd->write_marc8_last; + unsigned char byte; + char out_buf[10]; + size_t i, out_no = 0; + + if (!y) + return 0; + + byte = (unsigned char )((y>>16) & 0xff); + if (byte) + out_buf[out_no++] = byte; + byte = (unsigned char)((y>>8) & 0xff); + if (byte) + out_buf[out_no++] = byte; + byte = (unsigned char )(y & 0xff); + if (byte) + out_buf[out_no++] = byte; + + if (out_no + cd->write_marc8_comb_no + 1 > *outbytesleft) + { + cd->my_errno = YAZ_ICONV_E2BIG; + return (size_t) (-1); + } + + for (i = 0; i < cd->write_marc8_comb_no; i++) + { + /* all MARC-8 combined characters are simple bytes */ + byte = (unsigned char )(cd->write_marc8_comb_ch[i]); + *(*outbuf)++ = byte; + (*outbytesleft)--; + } + memcpy(*outbuf, out_buf, out_no); + *outbuf += out_no; + (*outbytesleft) -= out_no; + if (cd->write_marc8_second_half_char) + { + *(*outbuf)++ = cd->write_marc8_second_half_char; + (*outbytesleft)--; + } + + cd->write_marc8_last = 0; + cd->write_marc8_comb_no = 0; + cd->write_marc8_second_half_char = 0; + return 0; +} + +static size_t yaz_write_marc8_page_chr(yaz_iconv_t cd, + char **outbuf, size_t *outbytesleft, + const char *page_chr) +{ + const char *old_page_chr = cd->write_marc8_page_chr; + if (strcmp(page_chr, old_page_chr)) + { + size_t plen = 0; + const char *page_out = page_chr; + + if (*outbytesleft < 8) + { + cd->my_errno = YAZ_ICONV_E2BIG; + + return (size_t) (-1); + } + cd->write_marc8_page_chr = page_chr; + + if (!strcmp(old_page_chr, "\033p") + || !strcmp(old_page_chr, "\033g") + || !strcmp(old_page_chr, "\033b")) + { + /* Technique 1 leave */ + page_out = "\033s"; + if (strcmp(page_chr, "\033(B")) /* Not going ASCII page? */ + { + /* Must leave script + enter new page */ + plen = strlen(page_out); + memcpy(*outbuf, page_out, plen); + (*outbuf) += plen; + (*outbytesleft) -= plen; + page_out = page_chr; + } + } + plen = strlen(page_out); + memcpy(*outbuf, page_out, plen); + (*outbuf) += plen; + (*outbytesleft) -= plen; + } + return 0; +} + + +static size_t yaz_write_marc8_2(yaz_iconv_t cd, unsigned long x, + char **outbuf, size_t *outbytesleft) +{ + int comb = 0; + const char *page_chr = 0; + unsigned long y = lookup_marc8(cd, x, &comb, &page_chr); + + if (!y) + return (size_t) (-1); + + if (comb) + { + if (x == 0x0361) + cd->write_marc8_second_half_char = 0xEC; + else if (x == 0x0360) + cd->write_marc8_second_half_char = 0xFB; + + if (cd->write_marc8_comb_no < 6) + cd->write_marc8_comb_ch[cd->write_marc8_comb_no++] = y; + } + else + { + size_t r = flush_combos(cd, outbuf, outbytesleft); + if (r) + return r; + + r = yaz_write_marc8_page_chr(cd, outbuf, outbytesleft, page_chr); + if (r) + return r; + cd->write_marc8_last = y; + } + return 0; +} + +static size_t yaz_flush_marc8(yaz_iconv_t cd, + char **outbuf, size_t *outbytesleft) +{ + size_t r = flush_combos(cd, outbuf, outbytesleft); + if (r) + return r; + return yaz_write_marc8_page_chr(cd, outbuf, outbytesleft, "\033(B"); +} + +static size_t yaz_write_marc8(yaz_iconv_t cd, unsigned long x, + char **outbuf, size_t *outbytesleft) +{ + int i; + for (i = 0; latin1_comb[i].x1; i++) + { + if (x == latin1_comb[i].y) + { + size_t r ; + /* save the output pointers .. */ + char *outbuf0 = *outbuf; + size_t outbytesleft0 = *outbytesleft; + int last_ch = cd->write_marc8_last; + + r = yaz_write_marc8_2(cd, latin1_comb[i].x1, + outbuf, outbytesleft); + if (r) + return r; + r = yaz_write_marc8_2(cd, latin1_comb[i].x2, + outbuf, outbytesleft); + if (r && cd->my_errno == YAZ_ICONV_E2BIG) + { + /* not enough room. reset output to original values */ + *outbuf = outbuf0; + *outbytesleft = outbytesleft0; + cd->write_marc8_last = last_ch; + } + return r; + } + } + return yaz_write_marc8_2(cd, x, outbuf, outbytesleft); +} + + #if HAVE_WCHAR_H -static size_t yaz_write_wchar_t (yaz_iconv_t cd, unsigned long x, - char **outbuf, size_t *outbytesleft) +static size_t yaz_write_wchar_t(yaz_iconv_t cd, unsigned long x, + char **outbuf, size_t *outbytesleft) { unsigned char *outp = (unsigned char *) *outbuf; @@ -378,6 +1333,7 @@ yaz_iconv_t yaz_iconv_open (const char *tocode, const char *fromcode) cd->write_handle = 0; cd->read_handle = 0; cd->init_handle = 0; + cd->flush_handle = 0; cd->my_errno = YAZ_ICONV_UNKNOWN; /* a useful hack: if fromcode has leading @, @@ -399,6 +1355,10 @@ yaz_iconv_t yaz_iconv_open (const char *tocode, const char *fromcode) cd->read_handle = yaz_read_UCS4LE; else if (!yaz_matchstr(fromcode, "MARC8")) cd->read_handle = yaz_read_marc8; + else if (!yaz_matchstr(fromcode, "MARC8s")) + cd->read_handle = yaz_read_marc8s; + else if (!yaz_matchstr(fromcode, "advancegreek")) + cd->read_handle = yaz_read_advancegreek; #if HAVE_WCHAR_H else if (!yaz_matchstr(fromcode, "WCHAR_T")) cd->read_handle = yaz_read_wchar_t; @@ -407,11 +1367,28 @@ yaz_iconv_t yaz_iconv_open (const char *tocode, const char *fromcode) if (!yaz_matchstr(tocode, "UTF8")) cd->write_handle = yaz_write_UTF8; else if (!yaz_matchstr(tocode, "ISO88591")) + { cd->write_handle = yaz_write_ISO8859_1; + cd->flush_handle = yaz_flush_ISO8859_1; + } else if (!yaz_matchstr (tocode, "UCS4")) cd->write_handle = yaz_write_UCS4; else if (!yaz_matchstr(tocode, "UCS4LE")) cd->write_handle = yaz_write_UCS4LE; + else if (!yaz_matchstr(tocode, "MARC8")) + { + cd->write_handle = yaz_write_marc8; + cd->flush_handle = yaz_flush_marc8; + } + else if (!yaz_matchstr(tocode, "MARC8s")) + { + cd->write_handle = yaz_write_marc8; + cd->flush_handle = yaz_flush_marc8; + } + else if (!yaz_matchstr(tocode, "advancegreek")) + { + cd->write_handle = yaz_write_advancegreek; + } #if HAVE_WCHAR_H else if (!yaz_matchstr(tocode, "WCHAR_T")) cd->write_handle = yaz_write_wchar_t; @@ -439,11 +1416,12 @@ yaz_iconv_t yaz_iconv_open (const char *tocode, const char *fromcode) return cd; } -size_t yaz_iconv (yaz_iconv_t cd, char **inbuf, size_t *inbytesleft, - char **outbuf, size_t *outbytesleft) +size_t yaz_iconv(yaz_iconv_t cd, char **inbuf, size_t *inbytesleft, + char **outbuf, size_t *outbytesleft) { - char *inbuf0; + char *inbuf0 = 0; size_t r = 0; + #if HAVE_ICONV_H if (cd->iconv_cd) { @@ -469,19 +1447,32 @@ size_t yaz_iconv (yaz_iconv_t cd, char **inbuf, size_t *inbytesleft, return r; } #endif - if (inbuf == 0 || *inbuf == 0) + + if (inbuf) + inbuf0 = *inbuf; + + if (cd->init_flag) { - cd->init_flag = 1; cd->my_errno = YAZ_ICONV_UNKNOWN; - return 0; + cd->marc8_esc_mode = 'B'; + + cd->comb_offset = cd->comb_size = 0; + cd->compose_char = 0; + + cd->write_marc8_comb_no = 0; + cd->write_marc8_second_half_char = 0; + cd->write_marc8_last = 0; + cd->write_marc8_page_chr = "\033(B"; + + cd->unget_x = 0; + cd->no_read_x = 0; } - inbuf0 = *inbuf; if (cd->init_flag) { - if (cd->init_handle) + if (cd->init_handle && inbuf && *inbuf) { - size_t no_read; + size_t no_read = 0; size_t r = (cd->init_handle)(cd, (unsigned char *) *inbuf, *inbytesleft, &no_read); if (r) @@ -494,29 +1485,64 @@ size_t yaz_iconv (yaz_iconv_t cd, char **inbuf, size_t *inbytesleft, *inbytesleft -= no_read; *inbuf += no_read; } - cd->init_flag = 0; + } + cd->init_flag = 0; + + if (!inbuf || !*inbuf) + { + if (outbuf && *outbuf) + { + if (cd->unget_x) + r = (*cd->write_handle)(cd, cd->unget_x, outbuf, outbytesleft); + if (cd->flush_handle) + r = (*cd->flush_handle)(cd, outbuf, outbytesleft); + } + if (r == 0) + cd->init_flag = 1; + cd->unget_x = 0; + return r; } while (1) { unsigned long x; size_t no_read; - if (*inbytesleft == 0) + if (cd->unget_x) { - r = *inbuf - inbuf0; - break; + x = cd->unget_x; + no_read = cd->no_read_x; } - - x = (cd->read_handle)(cd, (unsigned char *) *inbuf, *inbytesleft, - &no_read); - if (no_read == 0) + else { - r = (size_t)(-1); - break; + if (*inbytesleft == 0) + { + r = *inbuf - inbuf0; + break; + } + x = (*cd->read_handle)(cd, (unsigned char *) *inbuf, *inbytesleft, + &no_read); + if (no_read == 0) + { + r = (size_t)(-1); + break; + } + } + if (x) + { + r = (*cd->write_handle)(cd, x, outbuf, outbytesleft); + if (r) + { + /* unable to write it. save it because read_handle cannot + rewind .. */ + if (cd->my_errno == YAZ_ICONV_E2BIG) + { + cd->unget_x = x; + cd->no_read_x = no_read; + break; + } + } + cd->unget_x = 0; } - r = (cd->write_handle)(cd, x, outbuf, outbytesleft); - if (r) - break; *inbytesleft -= no_read; (*inbuf) += no_read; } @@ -538,4 +1564,11 @@ int yaz_iconv_close (yaz_iconv_t cd) return 0; } - +/* + * Local variables: + * c-basic-offset: 4 + * indent-tabs-mode: nil + * End: + * vim: shiftwidth=4 tabstop=8 expandtab + */ +