| 1 | /* |
| 2 | * JSON Parser |
| 3 | * |
| 4 | * Copyright IBM, Corp. 2009 |
| 5 | * |
| 6 | * Authors: |
| 7 | * Anthony Liguori <aliguori@us.ibm.com> |
| 8 | * |
| 9 | * This work is licensed under the terms of the GNU LGPL, version 2.1 or later. |
| 10 | * See the COPYING.LIB file in the top-level directory. |
| 11 | * |
| 12 | */ |
| 13 | |
| 14 | #include "qemu/osdep.h" |
| 15 | #include "qemu/ctype.h" |
| 16 | #include "qemu/cutils.h" |
| 17 | #include "qemu/unicode.h" |
| 18 | #include "qapi/error.h" |
| 19 | #include "qobject/qbool.h" |
| 20 | #include "qobject/qdict.h" |
| 21 | #include "qobject/qlist.h" |
| 22 | #include "qobject/qnull.h" |
| 23 | #include "qobject/qnum.h" |
| 24 | #include "qobject/qstring.h" |
| 25 | #include "json-parser-int.h" |
| 26 | |
| 27 | /* |
| 28 | * The JSON parser is a push parser, returning a completed top-level |
| 29 | * object, an error, or NULL (if the object is incomplete and no error |
| 30 | * happened) after every token. Therefore it has an explicit |
| 31 | * representation of its parser stack; each stack entry consists of a |
| 32 | * parser state and a QObject: |
| 33 | * - a QList, for an array that is being added to |
| 34 | * - a QDict, for a dictionary that is being added to |
| 35 | * - a QString, for the key of the next pair that will be added to a QDict |
| 36 | * |
| 37 | * The stack represents an arbitrary nesting of arrays and dictionaries |
| 38 | * (whose next key has been parsed); it can also have a dictionary whose |
| 39 | * next key has not been parsed, but that can only happen at the top level. |
| 40 | * Because of this, the stack contents are always of the form |
| 41 | * "(QList | QDict QString)* QDict?". |
| 42 | * |
| 43 | * An empty stack represents the beginning of the parsing process, with |
| 44 | * start state BEFORE_VALUE. |
| 45 | */ |
| 46 | |
| 47 | typedef enum JSONParserState { |
| 48 | AFTER_LCURLY, |
| 49 | AFTER_LSQUARE, |
| 50 | BEFORE_KEY, |
| 51 | BEFORE_VALUE, |
| 52 | END_OF_KEY, |
| 53 | END_OF_VALUE, |
| 54 | } JSONParserState; |
| 55 | |
| 56 | typedef struct JSONParserStackEntry { |
| 57 | /* |
| 58 | * State when the container is completed or, for the top of the stack, |
| 59 | * entry state for the next token. |
| 60 | */ |
| 61 | JSONParserState state; |
| 62 | |
| 63 | /* |
| 64 | * A QString with the last parsed key, or a QList/QDict for the current |
| 65 | * container. |
| 66 | */ |
| 67 | QObject *partial; |
| 68 | } JSONParserStackEntry; |
| 69 | |
| 70 | /* |
| 71 | * This is the JSON grammar that's parsed, with the state transition and |
| 72 | * action at each point of the grammar. While this is not a formal |
| 73 | * description, "-> action" represents the pseudocode of the action |
| 74 | * and "-> STATE" sets the top stack entry's state to STATE. |
| 75 | * |
| 76 | * The state alone is enough to tell you what to parse; the state plus |
| 77 | * the type of the top of stack tells you which action to take. |
| 78 | * |
| 79 | * // The initial state is BEFORE_VALUE. |
| 80 | * input := value -> END_OF_VALUE -> return parsed value |
| 81 | * (input | END_OF_INPUT) |
| 82 | * |
| 83 | * // entered on BEFORE_VALUE; after any of these rules are processed, the |
| 84 | * // parser has completed a QObject and is in the END_OF_VALUE state. |
| 85 | * // |
| 86 | * // When the parser reaches the END_OF_VALUE state, it examines the |
| 87 | * // top of the stack to see if it's coming from "input" (stack empty), |
| 88 | * // "array_items" (TOS is a QList) or "dict_pairs" (TOS is a QString; the |
| 89 | * // item below will be a QDict). It then proceeds with the corresponding |
| 90 | * // actions, which will be one of: |
| 91 | * // - return parsed value |
| 92 | * // - add value to QList |
| 93 | * // - pop QString with the key, add key/value to the QDict |
| 94 | * value := literal -> END_OF_VALUE |
| 95 | * | '[' -> push empty QList -> AFTER_LSQUARE |
| 96 | * after_lsquare -> END_OF_VALUE |
| 97 | * | '{' -> push empty QDict -> AFTER_LCURLY |
| 98 | * after_lcurly -> END_OF_VALUE |
| 99 | * |
| 100 | * // non-recursive values, entered on BEFORE_VALUE |
| 101 | * literal := INTEGER -> END_OF_VALUE |
| 102 | * | FLOAT -> END_OF_VALUE |
| 103 | * | KEYWORD -> END_OF_VALUE |
| 104 | * | STRING -> END_OF_VALUE |
| 105 | * | INTERP -> END_OF_VALUE |
| 106 | * |
| 107 | * // entered on AFTER_LSQUARE |
| 108 | * after_lsquare := ']' -> pop completed QList -> END_OF_VALUE |
| 109 | * | ϵ -> BEFORE_VALUE |
| 110 | * array_items -> END_OF_VALUE |
| 111 | * |
| 112 | * // entered on BEFORE_VALUE, with TOS being a QList |
| 113 | * array_items := value -> add value to QList -> END_OF_VALUE |
| 114 | * (']' -> pop completed QList -> END_OF_VALUE |
| 115 | * | ',' -> BEFORE_VALUE |
| 116 | * array_items) -> END_OF_VALUE |
| 117 | * |
| 118 | * // entered on AFTER_LCURLY |
| 119 | * after_lcurly := '}' -> pop completed QDict -> END_OF_VALUE |
| 120 | * | ϵ -> BEFORE_KEY |
| 121 | * dict_pairs -> END_OF_VALUE |
| 122 | * |
| 123 | * // entered on BEFORE_KEY, with TOS being a QDict |
| 124 | * dict_pairs := (STRING | INTERP) -> push QString -> END_OF_KEY |
| 125 | * ':' -> BEFORE_VALUE |
| 126 | * value -> pop QString + add pair to QDict -> END_OF_VALUE |
| 127 | * ('}' -> pop completed QDict -> END_OF_VALUE |
| 128 | * | ',' -> BEFORE_KEY |
| 129 | * dict_pairs) -> END_OF_VALUE |
| 130 | * |
| 131 | * Parse errors ignore the token. json_parser_reset() can be |
| 132 | * called to restart parsing from scratch, with an empty stack. |
| 133 | */ |
| 134 | |
| 135 | #define BUG_ON(cond) assert(!(cond)) |
| 136 | |
| 137 | static inline JSONParserStackEntry *current_entry(JSONParserContext *ctxt) |
| 138 | { |
| 139 | return g_queue_peek_tail(ctxt->stack); |
| 140 | } |
| 141 | |
| 142 | static void push_entry(JSONParserContext *ctxt, QObject *partial, |
| 143 | JSONParserState state) |
| 144 | { |
| 145 | JSONParserStackEntry *entry = g_new(JSONParserStackEntry, 1); |
| 146 | entry->partial = partial; |
| 147 | entry->state = state; |
| 148 | g_queue_push_tail(ctxt->stack, entry); |
| 149 | } |
| 150 | |
| 151 | /* Drop the top entry and return the new top entry. */ |
| 152 | static JSONParserStackEntry *pop_entry(JSONParserContext *ctxt) |
| 153 | { |
| 154 | JSONParserStackEntry *entry = g_queue_pop_tail(ctxt->stack); |
| 155 | g_free(entry); |
| 156 | return current_entry(ctxt); |
| 157 | } |
| 158 | |
| 159 | /** |
| 160 | * Error handler |
| 161 | */ |
| 162 | static void G_GNUC_PRINTF(3, 4) parse_error(JSONParserContext *ctxt, |
| 163 | const JSONToken *token, |
| 164 | const char *msg, ...) |
| 165 | { |
| 166 | va_list ap; |
| 167 | char message[1024]; |
| 168 | |
| 169 | if (ctxt->err) { |
| 170 | return; |
| 171 | } |
| 172 | va_start(ap, msg); |
| 173 | vsnprintf(message, sizeof(message), msg, ap); |
| 174 | va_end(ap); |
| 175 | error_setg(&ctxt->err, "%d:%d: JSON parse error, %s", |
| 176 | token->y, token->x, message); |
| 177 | } |
| 178 | |
| 179 | static int cvt4hex(const char *s) |
| 180 | { |
| 181 | int cp, i; |
| 182 | |
| 183 | cp = 0; |
| 184 | for (i = 0; i < 4; i++) { |
| 185 | if (!qemu_isxdigit(s[i])) { |
| 186 | return -1; |
| 187 | } |
| 188 | cp <<= 4; |
| 189 | if (s[i] >= '0' && s[i] <= '9') { |
| 190 | cp |= s[i] - '0'; |
| 191 | } else if (s[i] >= 'a' && s[i] <= 'f') { |
| 192 | cp |= 10 + s[i] - 'a'; |
| 193 | } else if (s[i] >= 'A' && s[i] <= 'F') { |
| 194 | cp |= 10 + s[i] - 'A'; |
| 195 | } else { |
| 196 | return -1; |
| 197 | } |
| 198 | } |
| 199 | return cp; |
| 200 | } |
| 201 | |
| 202 | /** |
| 203 | * parse_string(): Parse a JSON string |
| 204 | * |
| 205 | * From RFC 8259 "The JavaScript Object Notation (JSON) Data |
| 206 | * Interchange Format": |
| 207 | * |
| 208 | * char = unescaped / |
| 209 | * escape ( |
| 210 | * %x22 / ; " quotation mark U+0022 |
| 211 | * %x5C / ; \ reverse solidus U+005C |
| 212 | * %x2F / ; / solidus U+002F |
| 213 | * %x62 / ; b backspace U+0008 |
| 214 | * %x66 / ; f form feed U+000C |
| 215 | * %x6E / ; n line feed U+000A |
| 216 | * %x72 / ; r carriage return U+000D |
| 217 | * %x74 / ; t tab U+0009 |
| 218 | * %x75 4HEXDIG ) ; uXXXX U+XXXX |
| 219 | * escape = %x5C ; \ |
| 220 | * quotation-mark = %x22 ; " |
| 221 | * unescaped = %x20-21 / %x23-5B / %x5D-10FFFF |
| 222 | * |
| 223 | * Extensions over RFC 8259: |
| 224 | * - Extra escape sequence in strings: |
| 225 | * 0x27 (apostrophe) is recognized after escape, too |
| 226 | * - Single-quoted strings: |
| 227 | * Like double-quoted strings, except they're delimited by %x27 |
| 228 | * (apostrophe) instead of %x22 (quotation mark), and can't contain |
| 229 | * unescaped apostrophe, but can contain unescaped quotation mark. |
| 230 | * |
| 231 | * Note: |
| 232 | * - Encoding is modified UTF-8. |
| 233 | * - Invalid Unicode characters are rejected. |
| 234 | * - Control characters \x00..\x1F are rejected by the lexer. |
| 235 | */ |
| 236 | static QString *parse_string(JSONParserContext *ctxt, const JSONToken *token) |
| 237 | { |
| 238 | const char *ptr = token->str; |
| 239 | GString *str; |
| 240 | char quote; |
| 241 | const char *beg; |
| 242 | int cp, trailing; |
| 243 | char *end; |
| 244 | ssize_t len; |
| 245 | char utf8_buf[5]; |
| 246 | |
| 247 | assert(*ptr == '"' || *ptr == '\''); |
| 248 | quote = *ptr++; |
| 249 | str = g_string_new(NULL); |
| 250 | |
| 251 | while (*ptr != quote) { |
| 252 | assert(*ptr); |
| 253 | switch (*ptr) { |
| 254 | case '\\': |
| 255 | beg = ptr++; |
| 256 | switch (*ptr++) { |
| 257 | case '"': |
| 258 | g_string_append_c(str, '"'); |
| 259 | break; |
| 260 | case '\'': |
| 261 | g_string_append_c(str, '\''); |
| 262 | break; |
| 263 | case '\\': |
| 264 | g_string_append_c(str, '\\'); |
| 265 | break; |
| 266 | case '/': |
| 267 | g_string_append_c(str, '/'); |
| 268 | break; |
| 269 | case 'b': |
| 270 | g_string_append_c(str, '\b'); |
| 271 | break; |
| 272 | case 'f': |
| 273 | g_string_append_c(str, '\f'); |
| 274 | break; |
| 275 | case 'n': |
| 276 | g_string_append_c(str, '\n'); |
| 277 | break; |
| 278 | case 'r': |
| 279 | g_string_append_c(str, '\r'); |
| 280 | break; |
| 281 | case 't': |
| 282 | g_string_append_c(str, '\t'); |
| 283 | break; |
| 284 | case 'u': |
| 285 | cp = cvt4hex(ptr); |
| 286 | ptr += 4; |
| 287 | |
| 288 | /* handle surrogate pairs */ |
| 289 | if (cp >= 0xD800 && cp <= 0xDBFF |
| 290 | && ptr[0] == '\\' && ptr[1] == 'u') { |
| 291 | /* leading surrogate followed by \u */ |
| 292 | cp = 0x10000 + ((cp & 0x3FF) << 10); |
| 293 | trailing = cvt4hex(ptr + 2); |
| 294 | if (trailing >= 0xDC00 && trailing <= 0xDFFF) { |
| 295 | /* followed by trailing surrogate */ |
| 296 | cp |= trailing & 0x3FF; |
| 297 | ptr += 6; |
| 298 | } else { |
| 299 | cp = -1; /* invalid */ |
| 300 | } |
| 301 | } |
| 302 | |
| 303 | if (mod_utf8_encode(utf8_buf, sizeof(utf8_buf), cp) < 0) { |
| 304 | parse_error(ctxt, token, |
| 305 | "%.*s is not a valid Unicode character", |
| 306 | (int)(ptr - beg), beg); |
| 307 | goto out; |
| 308 | } |
| 309 | g_string_append(str, utf8_buf); |
| 310 | break; |
| 311 | default: |
| 312 | parse_error(ctxt, token, "invalid escape sequence in string"); |
| 313 | goto out; |
| 314 | } |
| 315 | break; |
| 316 | case '%': |
| 317 | if (ctxt->ap) { |
| 318 | if (ptr[1] != '%') { |
| 319 | parse_error(ctxt, token, "can't interpolate into string"); |
| 320 | goto out; |
| 321 | } |
| 322 | ptr++; |
| 323 | } |
| 324 | /* fall through */ |
| 325 | default: |
| 326 | cp = mod_utf8_codepoint(ptr, 6, &end); |
| 327 | if (cp < 0) { |
| 328 | parse_error(ctxt, token, "invalid UTF-8 sequence in string"); |
| 329 | goto out; |
| 330 | } |
| 331 | ptr = end; |
| 332 | len = mod_utf8_encode(utf8_buf, sizeof(utf8_buf), cp); |
| 333 | assert(len >= 0); |
| 334 | g_string_append(str, utf8_buf); |
| 335 | } |
| 336 | } |
| 337 | |
| 338 | return qstring_from_gstring(str); |
| 339 | |
| 340 | out: |
| 341 | g_string_free(str, true); |
| 342 | return NULL; |
| 343 | } |
| 344 | |
| 345 | /* Terminals */ |
| 346 | |
| 347 | static QObject *parse_keyword(JSONParserContext *ctxt, const JSONToken *token) |
| 348 | { |
| 349 | assert(token && token->type == JSON_KEYWORD); |
| 350 | |
| 351 | if (!strcmp(token->str, "true")) { |
| 352 | return QOBJECT(qbool_from_bool(true)); |
| 353 | } else if (!strcmp(token->str, "false")) { |
| 354 | return QOBJECT(qbool_from_bool(false)); |
| 355 | } else if (!strcmp(token->str, "null")) { |
| 356 | return QOBJECT(qnull()); |
| 357 | } |
| 358 | parse_error(ctxt, token, "invalid keyword '%s'", token->str); |
| 359 | return NULL; |
| 360 | } |
| 361 | |
| 362 | static QObject *parse_interpolation(JSONParserContext *ctxt, |
| 363 | const JSONToken *token) |
| 364 | { |
| 365 | assert(token && token->type == JSON_INTERP); |
| 366 | |
| 367 | if (!strcmp(token->str, "%p")) { |
| 368 | return va_arg(*ctxt->ap, QObject *); |
| 369 | } else if (!strcmp(token->str, "%i")) { |
| 370 | return QOBJECT(qbool_from_bool(va_arg(*ctxt->ap, int))); |
| 371 | } else if (!strcmp(token->str, "%d")) { |
| 372 | return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, int))); |
| 373 | } else if (!strcmp(token->str, "%ld")) { |
| 374 | return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, long))); |
| 375 | } else if (!strcmp(token->str, "%lld")) { |
| 376 | return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, long long))); |
| 377 | } else if (!strcmp(token->str, "%" PRId64)) { |
| 378 | return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, int64_t))); |
| 379 | } else if (!strcmp(token->str, "%u")) { |
| 380 | return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, unsigned int))); |
| 381 | } else if (!strcmp(token->str, "%lu")) { |
| 382 | return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, unsigned long))); |
| 383 | } else if (!strcmp(token->str, "%llu")) { |
| 384 | return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, unsigned long long))); |
| 385 | } else if (!strcmp(token->str, "%" PRIu64)) { |
| 386 | return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, uint64_t))); |
| 387 | } else if (!strcmp(token->str, "%s")) { |
| 388 | return QOBJECT(qstring_from_str(va_arg(*ctxt->ap, const char *))); |
| 389 | } else if (!strcmp(token->str, "%f")) { |
| 390 | return QOBJECT(qnum_from_double(va_arg(*ctxt->ap, double))); |
| 391 | } |
| 392 | parse_error(ctxt, token, "invalid interpolation '%s'", token->str); |
| 393 | return NULL; |
| 394 | } |
| 395 | |
| 396 | static QObject *parse_literal(JSONParserContext *ctxt, const JSONToken *token) |
| 397 | { |
| 398 | assert(token); |
| 399 | |
| 400 | switch (token->type) { |
| 401 | case JSON_STRING: |
| 402 | return QOBJECT(parse_string(ctxt, token)); |
| 403 | case JSON_INTEGER: { |
| 404 | /* |
| 405 | * Represent JSON_INTEGER as QNUM_I64 if possible, else as |
| 406 | * QNUM_U64, else as QNUM_DOUBLE. Note that qemu_strtoi64() |
| 407 | * and qemu_strtou64() fail with ERANGE when it's not |
| 408 | * possible. |
| 409 | * |
| 410 | * qnum_get_int() will then work for any signed 64-bit |
| 411 | * JSON_INTEGER, qnum_get_uint() for any unsigned 64-bit |
| 412 | * integer, and qnum_get_double() both for any JSON_INTEGER |
| 413 | * and any JSON_FLOAT (with precision loss for integers beyond |
| 414 | * 53 bits) |
| 415 | */ |
| 416 | int ret; |
| 417 | int64_t value; |
| 418 | uint64_t uvalue; |
| 419 | |
| 420 | ret = qemu_strtoi64(token->str, NULL, 10, &value); |
| 421 | if (!ret) { |
| 422 | return QOBJECT(qnum_from_int(value)); |
| 423 | } |
| 424 | assert(ret == -ERANGE); |
| 425 | |
| 426 | if (token->str[0] != '-') { |
| 427 | ret = qemu_strtou64(token->str, NULL, 10, &uvalue); |
| 428 | if (!ret) { |
| 429 | return QOBJECT(qnum_from_uint(uvalue)); |
| 430 | } |
| 431 | assert(ret == -ERANGE); |
| 432 | } |
| 433 | } |
| 434 | /* fall through to JSON_FLOAT */ |
| 435 | case JSON_FLOAT: |
| 436 | /* FIXME dependent on locale; a pervasive issue in QEMU */ |
| 437 | /* FIXME our lexer matches RFC 8259 in forbidding Inf or NaN, |
| 438 | * but those might be useful extensions beyond JSON */ |
| 439 | return QOBJECT(qnum_from_double(strtod(token->str, NULL))); |
| 440 | default: |
| 441 | abort(); |
| 442 | } |
| 443 | } |
| 444 | |
| 445 | /* Parsing state machine */ |
| 446 | |
| 447 | static QObject *parse_begin_value(JSONParserContext *ctxt, |
| 448 | const JSONToken *token) |
| 449 | { |
| 450 | switch (token->type) { |
| 451 | case JSON_LCURLY: |
| 452 | push_entry(ctxt, QOBJECT(qdict_new()), AFTER_LCURLY); |
| 453 | return NULL; |
| 454 | case JSON_LSQUARE: |
| 455 | push_entry(ctxt, QOBJECT(qlist_new()), AFTER_LSQUARE); |
| 456 | return NULL; |
| 457 | case JSON_INTERP: |
| 458 | return parse_interpolation(ctxt, token); |
| 459 | case JSON_INTEGER: |
| 460 | case JSON_FLOAT: |
| 461 | case JSON_STRING: |
| 462 | return parse_literal(ctxt, token); |
| 463 | case JSON_KEYWORD: |
| 464 | return parse_keyword(ctxt, token); |
| 465 | default: |
| 466 | parse_error(ctxt, token, "expecting value"); |
| 467 | return NULL; |
| 468 | } |
| 469 | } |
| 470 | |
| 471 | static QObject *parse_token(JSONParserContext *ctxt, const JSONToken *token) |
| 472 | { |
| 473 | JSONParserStackEntry *entry; |
| 474 | JSONParserState state; |
| 475 | QString *key; |
| 476 | QObject *key_obj = NULL, *value = NULL; |
| 477 | |
| 478 | entry = current_entry(ctxt); |
| 479 | state = entry ? entry->state : BEFORE_VALUE; |
| 480 | switch (state) { |
| 481 | case AFTER_LCURLY: |
| 482 | /* Grab '}' for empty object or fall through to BEFORE_KEY */ |
| 483 | assert(qobject_type(entry->partial) == QTYPE_QDICT); |
| 484 | if (token->type == JSON_RCURLY) { |
| 485 | value = entry->partial; |
| 486 | entry = pop_entry(ctxt); |
| 487 | break; |
| 488 | } |
| 489 | entry->state = BEFORE_KEY; |
| 490 | /* fall through */ |
| 491 | |
| 492 | case BEFORE_KEY: |
| 493 | /* Expecting object key */ |
| 494 | assert(qobject_type(entry->partial) == QTYPE_QDICT); |
| 495 | if (token->type != JSON_STRING && token->type != JSON_INTERP) { |
| 496 | parse_error(ctxt, token, "expecting key"); |
| 497 | return NULL; |
| 498 | } |
| 499 | |
| 500 | key_obj = parse_begin_value(ctxt, token); |
| 501 | if (!key_obj) { |
| 502 | /* Parse error already reported */ |
| 503 | } else if (qobject_type(key_obj) != QTYPE_QSTRING) { |
| 504 | /* An interpolation was valid syntactically but not %s */ |
| 505 | parse_error(ctxt, token, "key is not a string in object"); |
| 506 | } else { |
| 507 | /* Store key in a special entry on the stack */ |
| 508 | push_entry(ctxt, key_obj, END_OF_KEY); |
| 509 | } |
| 510 | return NULL; |
| 511 | |
| 512 | case END_OF_KEY: |
| 513 | /* Expecting ':' after key */ |
| 514 | assert(qobject_type(entry->partial) == QTYPE_QSTRING); |
| 515 | if (token->type == JSON_COLON) { |
| 516 | entry->state = BEFORE_VALUE; |
| 517 | } else { |
| 518 | parse_error(ctxt, token, "expecting ':'"); |
| 519 | } |
| 520 | return NULL; |
| 521 | |
| 522 | case AFTER_LSQUARE: |
| 523 | /* Grab ']' for empty array or fall through to BEFORE_VALUE */ |
| 524 | assert(qobject_type(entry->partial) == QTYPE_QLIST); |
| 525 | if (token->type == JSON_RSQUARE) { |
| 526 | value = entry->partial; |
| 527 | entry = pop_entry(ctxt); |
| 528 | break; |
| 529 | } |
| 530 | entry->state = BEFORE_VALUE; |
| 531 | /* fall through */ |
| 532 | |
| 533 | case BEFORE_VALUE: |
| 534 | /* Expecting value */ |
| 535 | assert(!entry || qobject_type(entry->partial) != QTYPE_QDICT); |
| 536 | value = parse_begin_value(ctxt, token); |
| 537 | if (!value) { |
| 538 | /* Error or '['/'{' */ |
| 539 | return NULL; |
| 540 | } |
| 541 | /* Return value or insert it into a container */ |
| 542 | break; |
| 543 | |
| 544 | case END_OF_VALUE: |
| 545 | /* Grab ',' or ']' for array; ',' or '}' for object */ |
| 546 | if (qobject_to(QList, entry->partial)) { |
| 547 | /* Array */ |
| 548 | if (token->type != JSON_RSQUARE) { |
| 549 | if (token->type == JSON_COMMA) { |
| 550 | entry->state = BEFORE_VALUE; |
| 551 | } else { |
| 552 | parse_error(ctxt, token, "expected ',' or ']'"); |
| 553 | } |
| 554 | return NULL; |
| 555 | } |
| 556 | } else if (qobject_to(QDict, entry->partial)) { |
| 557 | /* Object */ |
| 558 | if (token->type != JSON_RCURLY) { |
| 559 | if (token->type == JSON_COMMA) { |
| 560 | entry->state = BEFORE_KEY; |
| 561 | } else { |
| 562 | parse_error(ctxt, token, "expected ',' or '}'"); |
| 563 | } |
| 564 | return NULL; |
| 565 | } |
| 566 | } else { |
| 567 | g_assert_not_reached(); |
| 568 | } |
| 569 | |
| 570 | /* Got ']' or '}'; return full value or insert into parent container */ |
| 571 | value = entry->partial; |
| 572 | entry = pop_entry(ctxt); |
| 573 | break; |
| 574 | } |
| 575 | |
| 576 | assert(value); |
| 577 | if (entry == NULL) { |
| 578 | /* Parse stack now empty, the top-level value is complete. */ |
| 579 | return value; |
| 580 | } |
| 581 | |
| 582 | /* |
| 583 | * Parse stack is not empty and entry->partial is the top of stack. |
| 584 | * It's a QString with the key (and a QDict is below it) if we're |
| 585 | * parsing an object, or a QList if we're parsing an array. |
| 586 | */ |
| 587 | key = qobject_to(QString, entry->partial); |
| 588 | if (key) { |
| 589 | const char *key_str; |
| 590 | QDict *dict; |
| 591 | |
| 592 | /* Pop off key, and store (key, value) in QDict. */ |
| 593 | entry = pop_entry(ctxt); |
| 594 | dict = qobject_to(QDict, entry->partial); |
| 595 | assert(dict); |
| 596 | key_str = qstring_get_str(key); |
| 597 | if (qdict_haskey(dict, key_str)) { |
| 598 | parse_error(ctxt, token, "duplicate key"); |
| 599 | qobject_unref(value); |
| 600 | return NULL; |
| 601 | } |
| 602 | qdict_put_obj(dict, key_str, value); |
| 603 | qobject_unref(key); |
| 604 | } else { |
| 605 | /* Array, just store value in the QList. */ |
| 606 | qlist_append_obj(qobject_to(QList, entry->partial), value); |
| 607 | } |
| 608 | |
| 609 | entry->state = END_OF_VALUE; |
| 610 | return NULL; |
| 611 | } |
| 612 | |
| 613 | |
| 614 | void json_parser_reset(JSONParserContext *ctxt) |
| 615 | { |
| 616 | JSONParserStackEntry *entry; |
| 617 | |
| 618 | ctxt->err = NULL; |
| 619 | while ((entry = g_queue_pop_tail(ctxt->stack)) != NULL) { |
| 620 | qobject_unref(entry->partial); |
| 621 | g_free(entry); |
| 622 | } |
| 623 | } |
| 624 | |
| 625 | void json_parser_init(JSONParserContext *ctxt, va_list *ap) |
| 626 | { |
| 627 | ctxt->stack = g_queue_new(); |
| 628 | ctxt->ap = ap; |
| 629 | json_parser_reset(ctxt); |
| 630 | } |
| 631 | |
| 632 | void json_parser_destroy(JSONParserContext *ctxt) |
| 633 | { |
| 634 | json_parser_reset(ctxt); |
| 635 | g_queue_free(ctxt->stack); |
| 636 | ctxt->stack = NULL; |
| 637 | } |
| 638 | |
| 639 | /* |
| 640 | * Advance the parser based on the token that is passed. |
| 641 | * Return the finished top-level value if the token completes it, else |
| 642 | * NULL. |
| 643 | * Once an error is returned, the function must not be called again |
| 644 | * without first resetting the parser. |
| 645 | */ |
| 646 | QObject *json_parser_feed(JSONParserContext *ctxt, const JSONToken *token, |
| 647 | Error **errp) |
| 648 | { |
| 649 | QObject *result = NULL; |
| 650 | |
| 651 | assert(!ctxt->err); |
| 652 | switch (token->type) { |
| 653 | case JSON_ERROR: |
| 654 | parse_error(ctxt, token, "stray '%s'", token->str); |
| 655 | break; |
| 656 | |
| 657 | case JSON_END_OF_INPUT: |
| 658 | /* Check for premature end of input */ |
| 659 | if (!g_queue_is_empty(ctxt->stack)) { |
| 660 | parse_error(ctxt, token, "premature end of input"); |
| 661 | } |
| 662 | break; |
| 663 | |
| 664 | default: |
| 665 | result = parse_token(ctxt, token); |
| 666 | break; |
| 667 | } |
| 668 | |
| 669 | error_propagate(errp, ctxt->err); |
| 670 | return result; |
| 671 | } |