| 1 | /* |
| 2 | * hz.c - HZ textual encoding of ASCII and GB2312, as defined in RFC 1843. |
| 3 | */ |
| 4 | |
| 5 | #ifndef ENUM_CHARSETS |
| 6 | |
| 7 | #include <assert.h> |
| 8 | |
| 9 | #include "charset.h" |
| 10 | #include "internal.h" |
| 11 | |
| 12 | static void read_hz(charset_spec const *charset, long int input_chr, |
| 13 | charset_state *state, |
| 14 | void (*emit)(void *ctx, long int output), void *emitctx) |
| 15 | { |
| 16 | /* |
| 17 | * When reading, our state variables are: |
| 18 | * |
| 19 | * - s0 is 0 in ASCII mode, 1 in GB2312 mode. |
| 20 | * |
| 21 | * - s1 stores a character we have just seen but not fully |
| 22 | * processed. So in ASCII mode, this can only ever be zero |
| 23 | * (no character) or 0x7E (~); in GB2312 mode it can be |
| 24 | * anything from 0x21-0x7E. |
| 25 | */ |
| 26 | |
| 27 | UNUSEDARG(charset); |
| 28 | |
| 29 | if (state->s0 == 0) { |
| 30 | /* |
| 31 | * ASCII mode. |
| 32 | */ |
| 33 | |
| 34 | if (state->s1) { |
| 35 | assert(state->s1 == '~'); |
| 36 | state->s1 = 0; |
| 37 | /* Process the character after a tilde. */ |
| 38 | switch (input_chr) { |
| 39 | case '~': |
| 40 | emit(emitctx, input_chr); |
| 41 | return; |
| 42 | case '\n': |
| 43 | return; /* ~\n is ignored */ |
| 44 | case '{': |
| 45 | state->s0 = 1; /* switch to GB2312 mode */ |
| 46 | return; |
| 47 | } |
| 48 | } else if (input_chr == '~') { |
| 49 | state->s1 = '~'; |
| 50 | return; |
| 51 | } else { |
| 52 | /* In ASCII mode, any non-tildes go straight */ |
| 53 | emit(emitctx, input_chr); |
| 54 | return; |
| 55 | } |
| 56 | } else { |
| 57 | /* |
| 58 | * GB2312 mode. As I understand it, we expect never to see |
| 59 | * anything in this mode that isn't 0x21-0x7E. So if we do, |
| 60 | * we'll simply throw an error and return to ASCII mode. |
| 61 | */ |
| 62 | if (input_chr < 0x21 || input_chr > 0x7E) { |
| 63 | emit(emitctx, ERROR); |
| 64 | state->s0 = state->s1 = 0; |
| 65 | return; |
| 66 | } |
| 67 | |
| 68 | /* |
| 69 | * So if we don't have a character stored already, store |
| 70 | * this one... |
| 71 | */ |
| 72 | if (!state->s1) { |
| 73 | state->s1 = input_chr; |
| 74 | return; |
| 75 | } |
| 76 | |
| 77 | /* |
| 78 | * ... otherwise, combine the stored char with this one. |
| 79 | * This will give either `~}', the escape sequence to |
| 80 | * return to ASCII mode, or something which we translate |
| 81 | * through GB2312. |
| 82 | */ |
| 83 | if (state->s1 == '~' && input_chr == '}') { |
| 84 | state->s1 = state->s0 = 0; |
| 85 | return; |
| 86 | } |
| 87 | |
| 88 | emit(emitctx, gb2312_to_unicode(state->s1 - 0x21, input_chr - 0x21)); |
| 89 | state->s1 = 0; |
| 90 | } |
| 91 | } |
| 92 | |
| 93 | static int write_hz(charset_spec const *charset, long int input_chr, |
| 94 | charset_state *state, |
| 95 | void (*emit)(void *ctx, long int output), void *emitctx) |
| 96 | { |
| 97 | int desired_state, r, c; |
| 98 | |
| 99 | UNUSEDARG(charset); |
| 100 | |
| 101 | /* |
| 102 | * Analyse the input char. |
| 103 | */ |
| 104 | if (input_chr < 0x80) { |
| 105 | desired_state = 0; |
| 106 | c = input_chr; |
| 107 | } else if (unicode_to_gb2312(input_chr, &r, &c)) { |
| 108 | desired_state = 1; |
| 109 | } else { |
| 110 | return FALSE; |
| 111 | } |
| 112 | |
| 113 | if (state->s0 != (unsigned)desired_state) { |
| 114 | emit(emitctx, '~'); |
| 115 | emit(emitctx, desired_state ? '{' : '}'); |
| 116 | state->s0 = desired_state; |
| 117 | } |
| 118 | |
| 119 | if (input_chr < 0) |
| 120 | return TRUE; /* special case: just reset state */ |
| 121 | |
| 122 | if (state->s0) { |
| 123 | /* |
| 124 | * GB mode. |
| 125 | */ |
| 126 | emit(emitctx, 0x21 + r); |
| 127 | emit(emitctx, 0x21 + c); |
| 128 | } else { |
| 129 | emit(emitctx, c); |
| 130 | } |
| 131 | return TRUE; |
| 132 | } |
| 133 | |
| 134 | const charset_spec charset_CS_HZ = { |
| 135 | CS_HZ, read_hz, write_hz, NULL |
| 136 | }; |
| 137 | |
| 138 | #else /* ENUM_CHARSETS */ |
| 139 | |
| 140 | ENUM_CHARSET(CS_HZ) |
| 141 | |
| 142 | #endif /* ENUM_CHARSETS */ |