4 * Copyright IBM, Corp. 2009
7 * Anthony Liguori <aliguori@us.ibm.com>
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.
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 "qapi/qmp/qbool.h"
20 #include "qapi/qmp/qdict.h"
21 #include "qapi/qmp/qlist.h"
22 #include "qapi/qmp/qnull.h"
23 #include "qapi/qmp/qnum.h"
24 #include "qapi/qmp/qstring.h"
25 #include "json-parser-int.h"
34 typedef struct JSONParserContext {
41 #define BUG_ON(cond) assert(!(cond))
46 * 0) make errors meaningful again
47 * 1) add geometry information to tokens
48 * 3) should we return a parsed size?
49 * 4) deal with premature EOI
52 static QObject *parse_value(JSONParserContext *ctxt);
57 static void G_GNUC_PRINTF(3, 4) parse_error(JSONParserContext *ctxt,
58 JSONToken *token, const char *msg, ...)
67 vsnprintf(message, sizeof(message), msg, ap);
69 error_setg(&ctxt->err, "JSON parse error, %s", message);
72 static int cvt4hex(const char *s)
77 for (i = 0; i < 4; i++) {
78 if (!qemu_isxdigit(s[i])) {
82 if (s[i] >= '0' && s[i] <= '9') {
84 } else if (s[i] >= 'a' && s[i] <= 'f') {
85 cp |= 10 + s[i] - 'a';
86 } else if (s[i] >= 'A' && s[i] <= 'F') {
87 cp |= 10 + s[i] - 'A';
96 * parse_string(): Parse a JSON string
98 * From RFC 8259 "The JavaScript Object Notation (JSON) Data
99 * Interchange Format":
103 * %x22 / ; " quotation mark U+0022
104 * %x5C / ; \ reverse solidus U+005C
105 * %x2F / ; / solidus U+002F
106 * %x62 / ; b backspace U+0008
107 * %x66 / ; f form feed U+000C
108 * %x6E / ; n line feed U+000A
109 * %x72 / ; r carriage return U+000D
110 * %x74 / ; t tab U+0009
111 * %x75 4HEXDIG ) ; uXXXX U+XXXX
113 * quotation-mark = %x22 ; "
114 * unescaped = %x20-21 / %x23-5B / %x5D-10FFFF
116 * Extensions over RFC 8259:
117 * - Extra escape sequence in strings:
118 * 0x27 (apostrophe) is recognized after escape, too
119 * - Single-quoted strings:
120 * Like double-quoted strings, except they're delimited by %x27
121 * (apostrophe) instead of %x22 (quotation mark), and can't contain
122 * unescaped apostrophe, but can contain unescaped quotation mark.
125 * - Encoding is modified UTF-8.
126 * - Invalid Unicode characters are rejected.
127 * - Control characters \x00..\x1F are rejected by the lexer.
129 static QString *parse_string(JSONParserContext *ctxt, JSONToken *token)
131 const char *ptr = token->str;
140 assert(*ptr == '"' || *ptr == '\'');
142 str = g_string_new(NULL);
144 while (*ptr != quote) {
151 g_string_append_c(str, '"');
154 g_string_append_c(str, '\'');
157 g_string_append_c(str, '\\');
160 g_string_append_c(str, '/');
163 g_string_append_c(str, '\b');
166 g_string_append_c(str, '\f');
169 g_string_append_c(str, '\n');
172 g_string_append_c(str, '\r');
175 g_string_append_c(str, '\t');
181 /* handle surrogate pairs */
182 if (cp >= 0xD800 && cp <= 0xDBFF
183 && ptr[0] == '\\' && ptr[1] == 'u') {
184 /* leading surrogate followed by \u */
185 cp = 0x10000 + ((cp & 0x3FF) << 10);
186 trailing = cvt4hex(ptr + 2);
187 if (trailing >= 0xDC00 && trailing <= 0xDFFF) {
188 /* followed by trailing surrogate */
189 cp |= trailing & 0x3FF;
192 cp = -1; /* invalid */
196 if (mod_utf8_encode(utf8_buf, sizeof(utf8_buf), cp) < 0) {
197 parse_error(ctxt, token,
198 "%.*s is not a valid Unicode character",
199 (int)(ptr - beg), beg);
202 g_string_append(str, utf8_buf);
205 parse_error(ctxt, token, "invalid escape sequence in string");
212 parse_error(ctxt, token, "can't interpolate into string");
219 cp = mod_utf8_codepoint(ptr, 6, &end);
221 parse_error(ctxt, token, "invalid UTF-8 sequence in string");
225 len = mod_utf8_encode(utf8_buf, sizeof(utf8_buf), cp);
227 g_string_append(str, utf8_buf);
231 return qstring_from_gstring(str);
234 g_string_free(str, true);
238 /* Note: the token object returned by parser_context_peek_token or
239 * parser_context_pop_token is deleted as soon as parser_context_pop_token
242 static JSONToken *parser_context_pop_token(JSONParserContext *ctxt)
244 g_free(ctxt->current);
245 ctxt->current = g_queue_pop_head(ctxt->buf);
246 return ctxt->current;
249 static JSONToken *parser_context_peek_token(JSONParserContext *ctxt)
251 return g_queue_peek_head(ctxt->buf);
257 static int parse_pair(JSONParserContext *ctxt, QDict *dict)
259 QObject *key_obj = NULL;
262 JSONToken *peek, *token;
264 peek = parser_context_peek_token(ctxt);
266 parse_error(ctxt, NULL, "premature EOI");
270 key_obj = parse_value(ctxt);
271 key = qobject_to(QString, key_obj);
273 parse_error(ctxt, peek, "key is not a string in object");
277 token = parser_context_pop_token(ctxt);
279 parse_error(ctxt, NULL, "premature EOI");
283 if (token->type != JSON_COLON) {
284 parse_error(ctxt, token, "missing : in object pair");
288 value = parse_value(ctxt);
290 parse_error(ctxt, token, "Missing value in dict");
294 if (qdict_haskey(dict, qstring_get_str(key))) {
295 parse_error(ctxt, token, "duplicate key");
299 qdict_put_obj(dict, qstring_get_str(key), value);
301 qobject_unref(key_obj);
305 qobject_unref(key_obj);
309 static QObject *parse_object(JSONParserContext *ctxt)
312 JSONToken *token, *peek;
314 token = parser_context_pop_token(ctxt);
315 assert(token && token->type == JSON_LCURLY);
319 peek = parser_context_peek_token(ctxt);
321 parse_error(ctxt, NULL, "premature EOI");
325 if (peek->type != JSON_RCURLY) {
326 if (parse_pair(ctxt, dict) == -1) {
330 token = parser_context_pop_token(ctxt);
332 parse_error(ctxt, NULL, "premature EOI");
336 while (token->type != JSON_RCURLY) {
337 if (token->type != JSON_COMMA) {
338 parse_error(ctxt, token, "expected separator in dict");
342 if (parse_pair(ctxt, dict) == -1) {
346 token = parser_context_pop_token(ctxt);
348 parse_error(ctxt, NULL, "premature EOI");
353 (void)parser_context_pop_token(ctxt);
356 return QOBJECT(dict);
363 static QObject *parse_array(JSONParserContext *ctxt)
366 JSONToken *token, *peek;
368 token = parser_context_pop_token(ctxt);
369 assert(token && token->type == JSON_LSQUARE);
373 peek = parser_context_peek_token(ctxt);
375 parse_error(ctxt, NULL, "premature EOI");
379 if (peek->type != JSON_RSQUARE) {
382 obj = parse_value(ctxt);
384 parse_error(ctxt, token, "expecting value");
388 qlist_append_obj(list, obj);
390 token = parser_context_pop_token(ctxt);
392 parse_error(ctxt, NULL, "premature EOI");
396 while (token->type != JSON_RSQUARE) {
397 if (token->type != JSON_COMMA) {
398 parse_error(ctxt, token, "expected separator in list");
402 obj = parse_value(ctxt);
404 parse_error(ctxt, token, "expecting value");
408 qlist_append_obj(list, obj);
410 token = parser_context_pop_token(ctxt);
412 parse_error(ctxt, NULL, "premature EOI");
417 (void)parser_context_pop_token(ctxt);
420 return QOBJECT(list);
427 static QObject *parse_keyword(JSONParserContext *ctxt)
431 token = parser_context_pop_token(ctxt);
432 assert(token && token->type == JSON_KEYWORD);
434 if (!strcmp(token->str, "true")) {
435 return QOBJECT(qbool_from_bool(true));
436 } else if (!strcmp(token->str, "false")) {
437 return QOBJECT(qbool_from_bool(false));
438 } else if (!strcmp(token->str, "null")) {
439 return QOBJECT(qnull());
441 parse_error(ctxt, token, "invalid keyword '%s'", token->str);
445 static QObject *parse_interpolation(JSONParserContext *ctxt)
449 token = parser_context_pop_token(ctxt);
450 assert(token && token->type == JSON_INTERP);
452 if (!strcmp(token->str, "%p")) {
453 return va_arg(*ctxt->ap, QObject *);
454 } else if (!strcmp(token->str, "%i")) {
455 return QOBJECT(qbool_from_bool(va_arg(*ctxt->ap, int)));
456 } else if (!strcmp(token->str, "%d")) {
457 return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, int)));
458 } else if (!strcmp(token->str, "%ld")) {
459 return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, long)));
460 } else if (!strcmp(token->str, "%lld")) {
461 return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, long long)));
462 } else if (!strcmp(token->str, "%" PRId64)) {
463 return QOBJECT(qnum_from_int(va_arg(*ctxt->ap, int64_t)));
464 } else if (!strcmp(token->str, "%u")) {
465 return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, unsigned int)));
466 } else if (!strcmp(token->str, "%lu")) {
467 return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, unsigned long)));
468 } else if (!strcmp(token->str, "%llu")) {
469 return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, unsigned long long)));
470 } else if (!strcmp(token->str, "%" PRIu64)) {
471 return QOBJECT(qnum_from_uint(va_arg(*ctxt->ap, uint64_t)));
472 } else if (!strcmp(token->str, "%s")) {
473 return QOBJECT(qstring_from_str(va_arg(*ctxt->ap, const char *)));
474 } else if (!strcmp(token->str, "%f")) {
475 return QOBJECT(qnum_from_double(va_arg(*ctxt->ap, double)));
477 parse_error(ctxt, token, "invalid interpolation '%s'", token->str);
481 static QObject *parse_literal(JSONParserContext *ctxt)
485 token = parser_context_pop_token(ctxt);
488 switch (token->type) {
490 return QOBJECT(parse_string(ctxt, token));
493 * Represent JSON_INTEGER as QNUM_I64 if possible, else as
494 * QNUM_U64, else as QNUM_DOUBLE. Note that qemu_strtoi64()
495 * and qemu_strtou64() fail with ERANGE when it's not
498 * qnum_get_int() will then work for any signed 64-bit
499 * JSON_INTEGER, qnum_get_uint() for any unsigned 64-bit
500 * integer, and qnum_get_double() both for any JSON_INTEGER
501 * and any JSON_FLOAT (with precision loss for integers beyond
508 ret = qemu_strtoi64(token->str, NULL, 10, &value);
510 return QOBJECT(qnum_from_int(value));
512 assert(ret == -ERANGE);
514 if (token->str[0] != '-') {
515 ret = qemu_strtou64(token->str, NULL, 10, &uvalue);
517 return QOBJECT(qnum_from_uint(uvalue));
519 assert(ret == -ERANGE);
522 /* fall through to JSON_FLOAT */
524 /* FIXME dependent on locale; a pervasive issue in QEMU */
525 /* FIXME our lexer matches RFC 8259 in forbidding Inf or NaN,
526 * but those might be useful extensions beyond JSON */
527 return QOBJECT(qnum_from_double(strtod(token->str, NULL)));
533 static QObject *parse_value(JSONParserContext *ctxt)
537 token = parser_context_peek_token(ctxt);
539 parse_error(ctxt, NULL, "premature EOI");
543 switch (token->type) {
545 return parse_object(ctxt);
547 return parse_array(ctxt);
549 return parse_interpolation(ctxt);
553 return parse_literal(ctxt);
555 return parse_keyword(ctxt);
557 parse_error(ctxt, token, "expecting value");
562 JSONToken *json_token(JSONTokenType type, int x, int y, GString *tokstr)
564 JSONToken *token = g_malloc(sizeof(JSONToken) + tokstr->len + 1);
567 memcpy(token->str, tokstr->str, tokstr->len);
568 token->str[tokstr->len] = 0;
574 QObject *json_parser_parse(GQueue *tokens, va_list *ap, Error **errp)
576 JSONParserContext ctxt = { .buf = tokens, .ap = ap };
579 result = parse_value(&ctxt);
580 assert(ctxt.err || g_queue_is_empty(ctxt.buf));
582 error_propagate(errp, ctxt.err);
584 while (!g_queue_is_empty(ctxt.buf)) {
585 parser_context_pop_token(&ctxt);
587 g_free(ctxt.current);