Andy Green | cd0c696 | 2016-03-28 10:12:37 +0800 | [diff] [blame] | 1 | /* |
| 2 | * Lightweight Embedded JSON Parser |
| 3 | * |
| 4 | * Copyright (C) 2013 Andy Green <andy@warmcat.com> |
| 5 | * This code is licensed under LGPL 2.1 |
| 6 | * http://www.gnu.org/licenses/lgpl-2.1.html |
| 7 | */ |
| 8 | |
| 9 | #include <string.h> |
| 10 | #include "lejp.h" |
| 11 | |
| 12 | #include <stdio.h> |
| 13 | |
| 14 | /** |
| 15 | * lejp_construct - prepare a struct lejp_ctx for use |
| 16 | * |
| 17 | * @ctx: pointer to your struct lejp_ctx |
| 18 | * @callback: your user callback which will received parsed tokens |
| 19 | * @user: optional user data pointer untouched by lejp |
| 20 | * @paths: your array of name elements you are interested in |
| 21 | * @count_paths: ARRAY_SIZE() of @paths |
| 22 | * |
| 23 | * Prepares your context struct for use with lejp |
| 24 | */ |
| 25 | |
| 26 | void |
| 27 | lejp_construct(struct lejp_ctx *ctx, |
| 28 | char (*callback)(struct lejp_ctx *ctx, char reason), void *user, |
| 29 | const char * const *paths, unsigned char count_paths) |
| 30 | { |
| 31 | ctx->st[0].s = 0; |
| 32 | ctx->st[0].p = 0; |
| 33 | ctx->st[0].i = 0; |
| 34 | ctx->st[0].b = 0; |
| 35 | ctx->sp = 0; |
| 36 | ctx->ipos = 0; |
| 37 | ctx->ppos = 0; |
| 38 | ctx->path_match = 0; |
| 39 | ctx->path[0] = '\0'; |
| 40 | ctx->callback = callback; |
| 41 | ctx->user = user; |
| 42 | ctx->paths = paths; |
| 43 | ctx->count_paths = count_paths; |
| 44 | ctx->line = 1; |
| 45 | ctx->callback(ctx, LEJPCB_CONSTRUCTED); |
| 46 | } |
| 47 | |
| 48 | /** |
| 49 | * lejp_destruct - retire a previously constructed struct lejp_ctx |
| 50 | * |
| 51 | * @ctx: pointer to your struct lejp_ctx |
| 52 | * |
| 53 | * lejp does not perform any allocations, but since your user code might, this |
| 54 | * provides a one-time LEJPCB_DESTRUCTED callback at destruction time where |
| 55 | * you can clean up in your callback. |
| 56 | */ |
| 57 | |
| 58 | void |
| 59 | lejp_destruct(struct lejp_ctx *ctx) |
| 60 | { |
| 61 | /* no allocations... just let callback know what it happening */ |
| 62 | ctx->callback(ctx, LEJPCB_DESTRUCTED); |
| 63 | } |
| 64 | |
| 65 | /** |
| 66 | * lejp_change_callback - switch to a different callback from now on |
| 67 | * |
| 68 | * @ctx: pointer to your struct lejp_ctx |
| 69 | * @callback: your user callback which will received parsed tokens |
| 70 | * |
| 71 | * This tells the old callback it was destroyed, in case you want to take any |
| 72 | * action because that callback "lost focus", then changes to the new |
| 73 | * callback and tells it first that it was constructed, and then started. |
| 74 | * |
| 75 | * Changing callback is a cheap and powerful trick to split out handlers |
| 76 | * according to information earlier in the parse. For example you may have |
| 77 | * a JSON pair "schema" whose value defines what can be expected for the rest |
| 78 | * of the JSON. Rather than having one huge callback for all cases, you can |
| 79 | * have an initial one looking for "schema" which then calls |
| 80 | * lejp_change_callback() to a handler specific for the schema. |
| 81 | * |
| 82 | * Notice that afterwards, you need to construct the context again anyway to |
| 83 | * parse another JSON object, and the callback is reset then to the main, |
| 84 | * schema-interpreting one. The construction action is very lightweight. |
| 85 | */ |
| 86 | |
| 87 | void |
| 88 | lejp_change_callback(struct lejp_ctx *ctx, |
| 89 | char (*callback)(struct lejp_ctx *ctx, char reason)) |
| 90 | { |
| 91 | ctx->callback(ctx, LEJPCB_DESTRUCTED); |
| 92 | ctx->callback = callback; |
| 93 | ctx->callback(ctx, LEJPCB_CONSTRUCTED); |
| 94 | ctx->callback(ctx, LEJPCB_START); |
| 95 | } |
| 96 | |
| 97 | static void |
| 98 | lejp_check_path_match(struct lejp_ctx *ctx) |
| 99 | { |
Andy Green | 37098ae | 2016-04-08 13:25:34 +0800 | [diff] [blame] | 100 | const char *p, *q; |
Andy Green | cd0c696 | 2016-03-28 10:12:37 +0800 | [diff] [blame] | 101 | int n; |
| 102 | |
| 103 | /* we only need to check if a match is not active */ |
| 104 | for (n = 0; !ctx->path_match && n < ctx->count_paths; n++) { |
Andy Green | 37098ae | 2016-04-08 13:25:34 +0800 | [diff] [blame] | 105 | ctx->wildcount = 0; |
| 106 | p = ctx->path; |
| 107 | q = ctx->paths[n]; |
| 108 | while (*p && *q) { |
| 109 | if (*q != '*') { |
| 110 | if (*p != *q) |
| 111 | break; |
| 112 | p++; |
| 113 | q++; |
| 114 | continue; |
| 115 | } |
| 116 | ctx->wild[ctx->wildcount++] = p - ctx->path; |
| 117 | q++; |
| 118 | while (*p && *p != '.') |
| 119 | p++; |
| 120 | } |
| 121 | if (*p || *q) |
Andy Green | cd0c696 | 2016-03-28 10:12:37 +0800 | [diff] [blame] | 122 | continue; |
Andy Green | 37098ae | 2016-04-08 13:25:34 +0800 | [diff] [blame] | 123 | |
Andy Green | cd0c696 | 2016-03-28 10:12:37 +0800 | [diff] [blame] | 124 | ctx->path_match = n + 1; |
| 125 | ctx->path_match_len = ctx->ppos; |
| 126 | return; |
| 127 | } |
Andy Green | 37098ae | 2016-04-08 13:25:34 +0800 | [diff] [blame] | 128 | |
| 129 | if (!ctx->path_match) |
| 130 | ctx->wildcount = 0; |
| 131 | } |
| 132 | |
| 133 | int |
| 134 | lejp_get_wildcard(struct lejp_ctx *ctx, int wildcard, char *dest, int len) |
| 135 | { |
| 136 | int n; |
| 137 | |
| 138 | if (wildcard >= ctx->wildcount || !len) |
| 139 | return 0; |
| 140 | |
| 141 | n = ctx->wild[wildcard]; |
| 142 | |
| 143 | while (--len && n < ctx->ppos && ctx->path[n] != '.') |
| 144 | *dest++ = ctx->path[n++]; |
| 145 | |
| 146 | *dest = '\0'; |
| 147 | n++; |
| 148 | |
| 149 | return n - ctx->wild[wildcard]; |
Andy Green | cd0c696 | 2016-03-28 10:12:37 +0800 | [diff] [blame] | 150 | } |
| 151 | |
| 152 | /** |
| 153 | * lejp_parse - interpret some more incoming data incrementally |
| 154 | * |
| 155 | * @ctx: previously constructed parsing context |
| 156 | * @json: char buffer with the new data to interpret |
| 157 | * @len: amount of data in the buffer |
| 158 | * |
| 159 | * Because lejp is a stream parser, it incrementally parses as new data |
| 160 | * becomes available, maintaining all state in the context struct. So an |
| 161 | * incomplete JSON is a normal situation, getting you a LEJP_CONTINUE |
| 162 | * return, signalling there's no error but to call again with more data when |
| 163 | * it comes to complete the parsing. Successful parsing completes with a |
| 164 | * 0 or positive integer indicating how much of the last input buffer was |
| 165 | * unused. |
| 166 | */ |
| 167 | |
| 168 | int |
| 169 | lejp_parse(struct lejp_ctx *ctx, const unsigned char *json, int len) |
| 170 | { |
| 171 | unsigned char c, n, s, ret = LEJP_REJECT_UNKNOWN; |
| 172 | static const char esc_char[] = "\"\\/bfnrt"; |
| 173 | static const char esc_tran[] = "\"\\/\b\f\n\r\t"; |
| 174 | static const char tokens[] = "rue alse ull "; |
| 175 | |
| 176 | if (!ctx->sp && !ctx->ppos) |
| 177 | ctx->callback(ctx, LEJPCB_START); |
| 178 | |
| 179 | while (len--) { |
| 180 | c = *json++; |
| 181 | |
| 182 | s = ctx->st[ctx->sp].s; |
| 183 | |
| 184 | /* skip whitespace unless we should care */ |
| 185 | if (c == ' ' || c == '\t' || c == '\n' || c == '\r' || c == '#') { |
| 186 | if (c == '\n') { |
| 187 | ctx->line++; |
| 188 | ctx->st[ctx->sp].s &= ~LEJP_FLAG_WS_COMMENTLINE; |
| 189 | } |
| 190 | if (!(s & LEJP_FLAG_WS_KEEP)) { |
| 191 | if (c == '#') |
| 192 | ctx->st[ctx->sp].s |= |
| 193 | LEJP_FLAG_WS_COMMENTLINE; |
| 194 | continue; |
| 195 | } |
| 196 | } |
| 197 | |
| 198 | if (ctx->st[ctx->sp].s & LEJP_FLAG_WS_COMMENTLINE) |
| 199 | continue; |
| 200 | |
| 201 | switch (s) { |
| 202 | case LEJP_IDLE: |
| 203 | if (c != '{') { |
| 204 | ret = LEJP_REJECT_IDLE_NO_BRACE; |
| 205 | goto reject; |
| 206 | } |
| 207 | ctx->callback(ctx, LEJPCB_OBJECT_START); |
| 208 | ctx->st[ctx->sp].s = LEJP_MEMBERS; |
| 209 | break; |
| 210 | case LEJP_MEMBERS: |
| 211 | if (c == '}') { |
| 212 | ctx->st[ctx->sp].s = LEJP_IDLE; |
| 213 | ret = LEJP_REJECT_MEMBERS_NO_CLOSE; |
| 214 | goto reject; |
| 215 | } |
| 216 | ctx->st[ctx->sp].s = LEJP_M_P; |
| 217 | goto redo_character; |
| 218 | case LEJP_M_P: |
| 219 | if (c != '\"') { |
| 220 | ret = LEJP_REJECT_MP_NO_OPEN_QUOTE; |
| 221 | goto reject; |
| 222 | } |
| 223 | /* push */ |
| 224 | ctx->st[ctx->sp].s = LEJP_MP_DELIM; |
| 225 | c = LEJP_MP_STRING; |
| 226 | goto add_stack_level; |
| 227 | |
| 228 | case LEJP_MP_STRING: |
| 229 | if (c == '\"') { |
| 230 | if (!ctx->sp) { |
| 231 | ret = LEJP_REJECT_MP_STRING_UNDERRUN; |
| 232 | goto reject; |
| 233 | } |
| 234 | if (ctx->st[ctx->sp - 1].s != LEJP_MP_DELIM) { |
| 235 | ctx->buf[ctx->npos] = '\0'; |
| 236 | if (ctx->callback(ctx, |
| 237 | LEJPCB_VAL_STR_END) < 0) { |
| 238 | ret = LEJP_REJECT_CALLBACK; |
| 239 | goto reject; |
| 240 | } |
| 241 | } |
| 242 | /* pop */ |
| 243 | ctx->sp--; |
| 244 | break; |
| 245 | } |
| 246 | if (c == '\\') { |
| 247 | ctx->st[ctx->sp].s = LEJP_MP_STRING_ESC; |
| 248 | break; |
| 249 | } |
| 250 | if (c < ' ') {/* "control characters" not allowed */ |
| 251 | ret = LEJP_REJECT_MP_ILLEGAL_CTRL; |
| 252 | goto reject; |
| 253 | } |
| 254 | goto emit_string_char; |
| 255 | |
| 256 | case LEJP_MP_STRING_ESC: |
| 257 | if (c == 'u') { |
| 258 | ctx->st[ctx->sp].s = LEJP_MP_STRING_ESC_U1; |
| 259 | ctx->uni = 0; |
| 260 | break; |
| 261 | } |
| 262 | for (n = 0; n < sizeof(esc_char); n++) { |
| 263 | if (c != esc_char[n]) |
| 264 | continue; |
| 265 | /* found it */ |
| 266 | c = esc_tran[n]; |
| 267 | ctx->st[ctx->sp].s = LEJP_MP_STRING; |
| 268 | goto emit_string_char; |
| 269 | } |
| 270 | ret = LEJP_REJECT_MP_STRING_ESC_ILLEGAL_ESC; |
| 271 | /* illegal escape char */ |
| 272 | goto reject; |
| 273 | |
| 274 | case LEJP_MP_STRING_ESC_U1: |
| 275 | case LEJP_MP_STRING_ESC_U2: |
| 276 | case LEJP_MP_STRING_ESC_U3: |
| 277 | case LEJP_MP_STRING_ESC_U4: |
| 278 | ctx->uni <<= 4; |
| 279 | if (c >= '0' && c <= '9') |
| 280 | ctx->uni |= c - '0'; |
| 281 | else |
| 282 | if (c >= 'a' && c <= 'f') |
| 283 | ctx->uni = c - 'a' + 10; |
| 284 | else |
| 285 | if (c >= 'A' && c <= 'F') |
| 286 | ctx->uni = c - 'A' + 10; |
| 287 | else { |
| 288 | ret = LEJP_REJECT_ILLEGAL_HEX; |
| 289 | goto reject; |
| 290 | } |
| 291 | ctx->st[ctx->sp].s++; |
| 292 | switch (s) { |
| 293 | case LEJP_MP_STRING_ESC_U2: |
| 294 | if (ctx->uni < 0x08) |
| 295 | break; |
| 296 | /* |
| 297 | * 0x08-0xff (0x0800 - 0xffff) |
| 298 | * emit 3-byte UTF-8 |
| 299 | */ |
| 300 | c = 0xe0 | ((ctx->uni >> 4) & 0xf); |
| 301 | goto emit_string_char; |
| 302 | |
| 303 | case LEJP_MP_STRING_ESC_U3: |
| 304 | if (ctx->uni >= 0x080) { |
| 305 | /* |
| 306 | * 0x080 - 0xfff (0x0800 - 0xffff) |
| 307 | * middle 3-byte seq |
| 308 | * send ....XXXXXX.. |
| 309 | */ |
| 310 | c = 0x80 | ((ctx->uni >> 2) & 0x3f); |
| 311 | goto emit_string_char; |
| 312 | } |
| 313 | if (ctx->uni < 0x008) |
| 314 | break; |
| 315 | /* |
| 316 | * 0x008 - 0x7f (0x0080 - 0x07ff) |
| 317 | * start 2-byte seq |
| 318 | */ |
| 319 | c = 0xc0 | (ctx->uni >> 2); |
| 320 | goto emit_string_char; |
| 321 | |
| 322 | case LEJP_MP_STRING_ESC_U4: |
| 323 | if (ctx->uni >= 0x0080) |
| 324 | /* end of 2 or 3-byte seq */ |
| 325 | c = 0x80 | (ctx->uni & 0x3f); |
| 326 | else |
| 327 | /* literal */ |
| 328 | c = (unsigned char)ctx->uni; |
| 329 | |
| 330 | ctx->st[ctx->sp].s = LEJP_MP_STRING; |
| 331 | goto emit_string_char; |
| 332 | default: |
| 333 | break; |
| 334 | } |
| 335 | break; |
| 336 | |
| 337 | case LEJP_MP_DELIM: |
| 338 | if (c != ':') { |
| 339 | ret = LEJP_REJECT_MP_DELIM_MISSING_COLON; |
| 340 | goto reject; |
| 341 | } |
| 342 | ctx->st[ctx->sp].s = LEJP_MP_VALUE; |
| 343 | ctx->path[ctx->ppos] = '\0'; |
| 344 | |
| 345 | lejp_check_path_match(ctx); |
| 346 | ctx->callback(ctx, LEJPCB_PAIR_NAME); |
| 347 | break; |
| 348 | |
| 349 | case LEJP_MP_VALUE: |
| 350 | if (c >= '0' && c <= '9') { |
| 351 | ctx->npos = 0; |
| 352 | ctx->dcount = 0; |
| 353 | ctx->f = 0; |
| 354 | ctx->st[ctx->sp].s = LEJP_MP_VALUE_NUM_INT; |
| 355 | goto redo_character; |
| 356 | } |
| 357 | switch (c) { |
| 358 | case'\"': |
| 359 | /* push */ |
| 360 | ctx->st[ctx->sp].s = LEJP_MP_COMMA_OR_END; |
| 361 | c = LEJP_MP_STRING; |
| 362 | ctx->npos = 0; |
| 363 | ctx->buf[0] = '\0'; |
| 364 | ctx->callback(ctx, LEJPCB_VAL_STR_START); |
| 365 | goto add_stack_level; |
| 366 | |
| 367 | case '{': |
| 368 | /* push */ |
| 369 | ctx->st[ctx->sp].s = LEJP_MP_COMMA_OR_END; |
| 370 | c = LEJP_MEMBERS; |
| 371 | lejp_check_path_match(ctx); |
| 372 | ctx->callback(ctx, LEJPCB_OBJECT_START); |
| 373 | ctx->path_match = 0; |
| 374 | goto add_stack_level; |
| 375 | |
| 376 | case '[': |
| 377 | /* push */ |
| 378 | ctx->st[ctx->sp].s = LEJP_MP_ARRAY_END; |
| 379 | c = LEJP_MP_VALUE; |
| 380 | ctx->path[ctx->ppos++] = '['; |
| 381 | ctx->path[ctx->ppos++] = ']'; |
| 382 | ctx->path[ctx->ppos] = '\0'; |
| 383 | ctx->callback(ctx, LEJPCB_ARRAY_START); |
| 384 | ctx->i[ctx->ipos++] = 0; |
| 385 | if (ctx->ipos > ARRAY_SIZE(ctx->i)) { |
| 386 | ret = LEJP_REJECT_MP_DELIM_ISTACK; |
| 387 | goto reject; |
| 388 | } |
| 389 | goto add_stack_level; |
| 390 | |
| 391 | case 't': /* true */ |
| 392 | ctx->uni = 0; |
| 393 | ctx->st[ctx->sp].s = LEJP_MP_VALUE_TOK; |
| 394 | break; |
| 395 | |
| 396 | case 'f': |
| 397 | ctx->uni = 4; |
| 398 | ctx->st[ctx->sp].s = LEJP_MP_VALUE_TOK; |
| 399 | break; |
| 400 | |
| 401 | case 'n': |
| 402 | ctx->uni = 4 + 5; |
| 403 | ctx->st[ctx->sp].s = LEJP_MP_VALUE_TOK; |
| 404 | break; |
| 405 | default: |
| 406 | ret = LEJP_REJECT_MP_DELIM_BAD_VALUE_START; |
| 407 | goto reject; |
| 408 | } |
| 409 | break; |
| 410 | |
| 411 | case LEJP_MP_VALUE_NUM_INT: |
| 412 | if (!ctx->npos && c == '-') { |
| 413 | ctx->f |= LEJP_SEEN_MINUS; |
| 414 | goto append_npos; |
| 415 | } |
| 416 | |
| 417 | if (ctx->dcount < 10 && c >= '0' && c <= '9') { |
| 418 | if (ctx->f & LEJP_SEEN_POINT) |
| 419 | ctx->f |= LEJP_SEEN_POST_POINT; |
| 420 | ctx->dcount++; |
| 421 | goto append_npos; |
| 422 | } |
| 423 | if (c == '.') { |
| 424 | if (ctx->dcount || (ctx->f & LEJP_SEEN_POINT)) { |
| 425 | ret = LEJP_REJECT_MP_VAL_NUM_FORMAT; |
| 426 | goto reject; |
| 427 | } |
| 428 | ctx->f |= LEJP_SEEN_POINT; |
| 429 | goto append_npos; |
| 430 | } |
| 431 | /* |
| 432 | * before exponent, if we had . we must have had at |
| 433 | * least one more digit |
| 434 | */ |
| 435 | if ((ctx->f & |
| 436 | (LEJP_SEEN_POINT | LEJP_SEEN_POST_POINT)) == |
| 437 | LEJP_SEEN_POINT) { |
| 438 | ret = LEJP_REJECT_MP_VAL_NUM_INT_NO_FRAC; |
| 439 | goto reject; |
| 440 | } |
| 441 | if (c == 'e' || c == 'E') { |
| 442 | if (ctx->f & LEJP_SEEN_EXP) { |
| 443 | ret = LEJP_REJECT_MP_VAL_NUM_FORMAT; |
| 444 | goto reject; |
| 445 | } |
| 446 | ctx->f |= LEJP_SEEN_EXP; |
| 447 | ctx->st[ctx->sp].s = LEJP_MP_VALUE_NUM_EXP; |
| 448 | goto append_npos; |
| 449 | } |
| 450 | /* if none of the above, did we even have a number? */ |
| 451 | if (!ctx->dcount) { |
| 452 | ret = LEJP_REJECT_MP_VAL_NUM_FORMAT; |
| 453 | goto reject; |
| 454 | } |
| 455 | |
| 456 | ctx->buf[ctx->npos] = '\0'; |
| 457 | if (ctx->f & LEJP_SEEN_POINT) |
| 458 | ctx->callback(ctx, LEJPCB_VAL_NUM_FLOAT); |
| 459 | else |
| 460 | ctx->callback(ctx, LEJPCB_VAL_NUM_INT); |
| 461 | |
| 462 | /* then this is the post-number character, loop */ |
| 463 | ctx->st[ctx->sp].s = LEJP_MP_COMMA_OR_END; |
| 464 | goto redo_character; |
| 465 | |
| 466 | case LEJP_MP_VALUE_NUM_EXP: |
| 467 | ctx->st[ctx->sp].s = LEJP_MP_VALUE_NUM_INT; |
| 468 | if (c >= '0' && c <= '9') |
| 469 | goto redo_character; |
| 470 | if (c == '+' || c == '-') |
| 471 | goto append_npos; |
| 472 | ret = LEJP_REJECT_MP_VAL_NUM_EXP_BAD_EXP; |
| 473 | goto reject; |
| 474 | |
| 475 | case LEJP_MP_VALUE_TOK: /* true, false, null */ |
| 476 | if (c != tokens[ctx->uni]) { |
| 477 | ret = LEJP_REJECT_MP_VAL_TOK_UNKNOWN; |
| 478 | goto reject; |
| 479 | } |
| 480 | ctx->uni++; |
| 481 | if (tokens[ctx->uni] != ' ') |
| 482 | break; |
| 483 | switch (ctx->uni) { |
| 484 | case 3: |
| 485 | ctx->buf[0] = '1'; |
| 486 | ctx->buf[1] = '\0'; |
| 487 | ctx->callback(ctx, LEJPCB_VAL_TRUE); |
| 488 | break; |
| 489 | case 8: |
| 490 | ctx->buf[0] = '0'; |
| 491 | ctx->buf[1] = '\0'; |
| 492 | ctx->callback(ctx, LEJPCB_VAL_FALSE); |
| 493 | break; |
| 494 | case 12: |
| 495 | ctx->buf[0] = '\0'; |
| 496 | ctx->callback(ctx, LEJPCB_VAL_NULL); |
| 497 | break; |
| 498 | } |
| 499 | ctx->st[ctx->sp].s = LEJP_MP_COMMA_OR_END; |
| 500 | break; |
| 501 | |
| 502 | case LEJP_MP_COMMA_OR_END: |
| 503 | ctx->path[ctx->ppos] = '\0'; |
| 504 | if (c == ',') { |
| 505 | /* increment this stack level's index */ |
| 506 | ctx->st[ctx->sp].s = LEJP_M_P; |
| 507 | if (!ctx->sp) { |
| 508 | ctx->ppos = 0; |
| 509 | /* |
| 510 | * since we came back to root level, |
| 511 | * no path can still match |
| 512 | */ |
| 513 | ctx->path_match = 0; |
| 514 | break; |
| 515 | } |
| 516 | ctx->ppos = ctx->st[ctx->sp - 1].p; |
| 517 | ctx->path[ctx->ppos] = '\0'; |
| 518 | if (ctx->path_match && |
| 519 | ctx->ppos <= ctx->path_match_len) |
| 520 | /* |
| 521 | * we shrank the path to be |
| 522 | * smaller than the matching point |
| 523 | */ |
| 524 | ctx->path_match = 0; |
| 525 | |
| 526 | if (ctx->st[ctx->sp - 1].s != LEJP_MP_ARRAY_END) |
| 527 | break; |
| 528 | /* top level is definitely an array... */ |
| 529 | if (ctx->ipos) |
| 530 | ctx->i[ctx->ipos - 1]++; |
| 531 | ctx->st[ctx->sp].s = LEJP_MP_VALUE; |
| 532 | break; |
| 533 | } |
| 534 | if (c == ']') { |
| 535 | if (!ctx->sp) { |
| 536 | ret = LEJP_REJECT_MP_C_OR_E_UNDERF; |
| 537 | goto reject; |
| 538 | } |
| 539 | /* pop */ |
| 540 | ctx->sp--; |
| 541 | if (ctx->st[ctx->sp].s != LEJP_MP_ARRAY_END) { |
| 542 | ret = LEJP_REJECT_MP_C_OR_E_NOTARRAY; |
| 543 | goto reject; |
| 544 | } |
| 545 | /* drop the path [n] bit */ |
| 546 | ctx->ppos = ctx->st[ctx->sp - 1].p; |
| 547 | ctx->ipos = ctx->st[ctx->sp - 1].i; |
| 548 | ctx->path[ctx->ppos] = '\0'; |
| 549 | if (ctx->path_match && |
| 550 | ctx->ppos <= ctx->path_match_len) |
| 551 | /* |
| 552 | * we shrank the path to be |
| 553 | * smaller than the matching point |
| 554 | */ |
| 555 | ctx->path_match = 0; |
| 556 | |
| 557 | /* do LEJP_MP_ARRAY_END processing */ |
| 558 | goto redo_character; |
| 559 | } |
| 560 | if (c == '}') { |
| 561 | if (ctx->sp == 0) { |
| 562 | lejp_check_path_match(ctx); |
| 563 | ctx->callback(ctx, LEJPCB_OBJECT_END); |
| 564 | ctx->callback(ctx, LEJPCB_COMPLETE); |
| 565 | /* done, return unused amount */ |
| 566 | return len; |
| 567 | } |
| 568 | /* pop */ |
| 569 | ctx->sp--; |
| 570 | ctx->ppos = ctx->st[ctx->sp - 1].p; |
| 571 | ctx->ipos = ctx->st[ctx->sp - 1].i; |
| 572 | ctx->path[ctx->ppos] = '\0'; |
| 573 | if (ctx->path_match && |
| 574 | ctx->ppos <= ctx->path_match_len) |
| 575 | /* |
| 576 | * we shrank the path to be |
| 577 | * smaller than the matching point |
| 578 | */ |
| 579 | ctx->path_match = 0; |
| 580 | lejp_check_path_match(ctx); |
| 581 | ctx->callback(ctx, LEJPCB_OBJECT_END); |
| 582 | break; |
| 583 | } |
| 584 | |
| 585 | ret = LEJP_REJECT_MP_C_OR_E_NEITHER; |
| 586 | goto reject; |
| 587 | |
| 588 | case LEJP_MP_ARRAY_END: |
| 589 | ctx->path[ctx->ppos] = '\0'; |
| 590 | if (c == ',') { |
| 591 | /* increment this stack level's index */ |
| 592 | if (ctx->ipos) |
| 593 | ctx->i[ctx->ipos - 1]++; |
| 594 | ctx->st[ctx->sp].s = LEJP_MP_VALUE; |
| 595 | if (ctx->sp) |
| 596 | ctx->ppos = ctx->st[ctx->sp - 1].p; |
| 597 | ctx->path[ctx->ppos] = '\0'; |
| 598 | break; |
| 599 | } |
| 600 | if (c != ']') { |
| 601 | ret = LEJP_REJECT_MP_ARRAY_END_MISSING; |
| 602 | goto reject; |
| 603 | } |
| 604 | |
| 605 | ctx->st[ctx->sp].s = LEJP_MP_COMMA_OR_END; |
| 606 | ctx->callback(ctx, LEJPCB_ARRAY_END); |
| 607 | break; |
| 608 | } |
| 609 | |
| 610 | continue; |
| 611 | |
| 612 | emit_string_char: |
| 613 | if (!ctx->sp || ctx->st[ctx->sp - 1].s != LEJP_MP_DELIM) { |
| 614 | /* assemble the string value into chunks */ |
| 615 | ctx->buf[ctx->npos++] = c; |
| 616 | if (ctx->npos == sizeof(ctx->buf) - 1) { |
| 617 | ctx->callback(ctx, LEJPCB_VAL_STR_CHUNK); |
| 618 | ctx->npos = 0; |
| 619 | } |
| 620 | continue; |
| 621 | } |
| 622 | /* name part of name:value pair */ |
| 623 | ctx->path[ctx->ppos++] = c; |
| 624 | continue; |
| 625 | |
| 626 | add_stack_level: |
| 627 | /* push on to the object stack */ |
| 628 | if (ctx->ppos && ctx->st[ctx->sp].s != LEJP_MP_COMMA_OR_END && |
| 629 | ctx->st[ctx->sp].s != LEJP_MP_ARRAY_END) |
| 630 | ctx->path[ctx->ppos++] = '.'; |
| 631 | |
| 632 | ctx->st[ctx->sp].p = ctx->ppos; |
| 633 | ctx->st[ctx->sp].i = ctx->ipos; |
| 634 | if (++ctx->sp == ARRAY_SIZE(ctx->st)) { |
| 635 | ret = LEJP_REJECT_STACK_OVERFLOW; |
| 636 | goto reject; |
| 637 | } |
| 638 | ctx->path[ctx->ppos] = '\0'; |
| 639 | ctx->st[ctx->sp].s = c; |
| 640 | ctx->st[ctx->sp].b = 0; |
| 641 | continue; |
| 642 | |
| 643 | append_npos: |
| 644 | if (ctx->npos >= sizeof(ctx->buf)) { |
| 645 | ret = LEJP_REJECT_NUM_TOO_LONG; |
| 646 | goto reject; |
| 647 | } |
| 648 | ctx->buf[ctx->npos++] = c; |
| 649 | continue; |
| 650 | |
| 651 | redo_character: |
| 652 | json--; |
| 653 | len++; |
| 654 | } |
| 655 | |
| 656 | return LEJP_CONTINUE; |
| 657 | |
| 658 | reject: |
| 659 | ctx->callback(ctx, LEJPCB_FAILED); |
| 660 | return ret; |
| 661 | } |