12 #include "jansson_private.h"
13 #include "strbuffer.h"
16 #define TOKEN_INVALID -1
18 #define TOKEN_STRING 256
19 #define TOKEN_INTEGER 257
20 #define TOKEN_REAL 258
21 #define TOKEN_TRUE 259
22 #define TOKEN_FALSE 260
23 #define TOKEN_NULL 261
25 /* read one byte from stream, return EOF on end of file */
26 typedef int (*get_func)(void *data);
28 /* return non-zero if end of file has been reached */
29 typedef int (*eof_func)(void *data);
43 strbuffer_t saved_text;
54 /*** error reporting ***/
56 static void error_init(json_error_t *error)
60 error->text[0] = '\0';
65 static void error_set(json_error_t *error, const lex_t *lex,
69 char text[JSON_ERROR_TEXT_LENGTH];
71 if(!error || error->text[0] != '\0') {
72 /* error already set */
77 vsnprintf(text, JSON_ERROR_TEXT_LENGTH, msg, ap);
82 const char *saved_text = strbuffer_value(&lex->saved_text);
83 error->line = lex->line;
84 if(saved_text && saved_text[0])
86 if(lex->saved_text.length <= 20) {
87 snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
88 "%s near '%s'", text, saved_text);
91 snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
95 snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
96 "%s near end of file", text);
102 snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
107 /*** lexical analyzer ***/
109 void stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
114 stream->stream_pos = 0;
115 stream->buffer[0] = '\0';
116 stream->buffer_pos = 0;
119 static char stream_get(stream_t *stream, json_error_t *error)
123 if(!stream->buffer[stream->buffer_pos])
125 stream->buffer[0] = stream->get(stream->data);
126 stream->buffer_pos = 0;
128 c = stream->buffer[0];
130 if(c == EOF && stream->eof(stream->data))
135 /* multi-byte UTF-8 sequence */
138 count = utf8_check_first(c);
144 for(i = 1; i < count; i++)
145 stream->buffer[i] = stream->get(stream->data);
147 if(!utf8_check_full(stream->buffer, count))
150 stream->stream_pos += count;
151 stream->buffer[count] = '\0';
154 stream->buffer[1] = '\0';
155 stream->stream_pos++;
159 return stream->buffer[stream->buffer_pos++];
162 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
163 (unsigned char)c, stream->stream_pos);
165 stream->buffer[0] = EOF;
166 stream->buffer[1] = '\0';
167 stream->buffer_pos = 1;
172 static void stream_unget(stream_t *stream, char c)
174 assert(stream->buffer_pos > 0);
175 stream->buffer_pos--;
176 assert(stream->buffer[stream->buffer_pos] == c);
180 static int lex_get(lex_t *lex, json_error_t *error)
182 return stream_get(&lex->stream, error);
185 static int lex_eof(lex_t *lex)
187 return lex->stream.eof(lex->stream.data);
190 static void lex_save(lex_t *lex, char c)
192 strbuffer_append_byte(&lex->saved_text, c);
195 static int lex_get_save(lex_t *lex, json_error_t *error)
197 char c = stream_get(&lex->stream, error);
202 static void lex_unget_unsave(lex_t *lex, char c)
205 stream_unget(&lex->stream, c);
206 d = strbuffer_pop(&lex->saved_text);
210 static void lex_save_cached(lex_t *lex)
212 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
214 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
215 lex->stream.buffer_pos++;
219 /* assumes that str points to 'u' plus at least 4 valid hex digits */
220 static int decode_unicode_escape(const char *str)
225 assert(str[0] == 'u');
227 for(i = 1; i <= 4; i++) {
233 value += c - 'a' + 10;
235 value += c - 'A' + 10;
243 static void lex_scan_string(lex_t *lex, json_error_t *error)
250 lex->value.string = NULL;
251 lex->token = TOKEN_INVALID;
254 c = lex_get_save(lex, error);
259 error_set(error, lex, "premature end of input");
263 else if(0 <= c && c <= 0x1F) {
264 /* control character */
265 lex_unget_unsave(lex, c);
267 error_set(error, lex, "unexpected newline", c);
269 error_set(error, lex, "control character 0x%x", c);
274 c = lex_get_save(lex, error);
276 c = lex_get_save(lex, error);
277 for(i = 0; i < 4; i++) {
279 lex_unget_unsave(lex, c);
280 error_set(error, lex, "invalid escape");
283 c = lex_get_save(lex, error);
286 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
287 c == 'f' || c == 'n' || c == 'r' || c == 't')
288 c = lex_get_save(lex, error);
290 lex_unget_unsave(lex, c);
291 error_set(error, lex, "invalid escape");
296 c = lex_get_save(lex, error);
299 /* the actual value is at most of the same length as the source
301 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
302 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
303 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
304 are converted to 4 bytes
306 lex->value.string = malloc(lex->saved_text.length + 1);
307 if(!lex->value.string) {
308 /* this is not very nice, since TOKEN_INVALID is returned */
313 t = lex->value.string;
315 /* + 1 to skip the " */
316 p = strbuffer_value(&lex->saved_text) + 1;
326 value = decode_unicode_escape(p);
329 if(0xD800 <= value && value <= 0xDBFF) {
331 if(*p == '\\' && *(p + 1) == 'u') {
332 int value2 = decode_unicode_escape(++p);
335 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
336 /* valid second surrogate */
337 value = ((value - 0xD800) << 10) +
342 /* invalid second surrogate */
343 error_set(error, lex,
344 "invalid Unicode '\\u%04X\\u%04X'",
350 /* no second surrogate */
351 error_set(error, lex, "invalid Unicode '\\u%04X'",
356 else if(0xDC00 <= value && value <= 0xDFFF) {
357 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
362 error_set(error, lex, "\\u0000 is not allowed");
366 if(utf8_encode(value, buffer, &length))
369 memcpy(t, buffer, length);
374 case '"': case '\\': case '/':
376 case 'b': *t = '\b'; break;
377 case 'f': *t = '\f'; break;
378 case 'n': *t = '\n'; break;
379 case 'r': *t = '\r'; break;
380 case 't': *t = '\t'; break;
391 lex->token = TOKEN_STRING;
395 free(lex->value.string);
398 static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
400 const char *saved_text;
403 lex->token = TOKEN_INVALID;
406 c = lex_get_save(lex, error);
409 c = lex_get_save(lex, error);
411 lex_unget_unsave(lex, c);
415 else /* c != '0' */ {
416 c = lex_get_save(lex, error);
418 c = lex_get_save(lex, error);
421 if(c != '.' && c != 'E' && c != 'e') {
422 lex_unget_unsave(lex, c);
423 lex->token = TOKEN_INTEGER;
425 saved_text = strbuffer_value(&lex->saved_text);
426 lex->value.integer = strtol(saved_text, &end, 10);
427 assert(end == saved_text + lex->saved_text.length);
433 c = lex_get(lex, error);
438 c = lex_get_save(lex, error);
440 c = lex_get_save(lex, error);
443 if(c == 'E' || c == 'e') {
444 c = lex_get_save(lex, error);
445 if(c == '+' || c == '-')
446 c = lex_get_save(lex, error);
449 lex_unget_unsave(lex, c);
453 c = lex_get_save(lex, error);
455 c = lex_get_save(lex, error);
458 lex_unget_unsave(lex, c);
459 lex->token = TOKEN_REAL;
461 saved_text = strbuffer_value(&lex->saved_text);
462 lex->value.real = strtod(saved_text, &end);
463 assert(end == saved_text + lex->saved_text.length);
469 static int lex_scan(lex_t *lex, json_error_t *error)
473 strbuffer_clear(&lex->saved_text);
475 if(lex->token == TOKEN_STRING) {
476 free(lex->value.string);
477 lex->value.string = NULL;
480 c = lex_get(lex, error);
481 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
486 c = lex_get(lex, error);
491 lex->token = TOKEN_EOF;
493 lex->token = TOKEN_INVALID;
499 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
503 lex_scan_string(lex, error);
505 else if(isdigit(c) || c == '-')
506 lex_scan_number(lex, c, error);
508 else if(isupper(c) || islower(c)) {
509 /* eat up the whole identifier for clearer error messages */
510 const char *saved_text;
512 c = lex_get_save(lex, error);
513 while(isupper(c) || islower(c))
514 c = lex_get_save(lex, error);
515 lex_unget_unsave(lex, c);
517 saved_text = strbuffer_value(&lex->saved_text);
519 if(strcmp(saved_text, "true") == 0)
520 lex->token = TOKEN_TRUE;
521 else if(strcmp(saved_text, "false") == 0)
522 lex->token = TOKEN_FALSE;
523 else if(strcmp(saved_text, "null") == 0)
524 lex->token = TOKEN_NULL;
526 lex->token = TOKEN_INVALID;
530 /* save the rest of the input UTF-8 sequence to get an error
531 message of valid UTF-8 */
532 lex_save_cached(lex);
533 lex->token = TOKEN_INVALID;
540 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
542 stream_init(&lex->stream, get, eof, data);
543 if(strbuffer_init(&lex->saved_text))
546 lex->token = TOKEN_INVALID;
552 static void lex_close(lex_t *lex)
554 if(lex->token == TOKEN_STRING)
555 free(lex->value.string);
556 strbuffer_close(&lex->saved_text);
562 static json_t *parse_value(lex_t *lex, json_error_t *error);
564 static json_t *parse_object(lex_t *lex, json_error_t *error)
566 json_t *object = json_object();
570 lex_scan(lex, error);
571 if(lex->token == '}')
578 if(lex->token != TOKEN_STRING) {
579 error_set(error, lex, "string or '}' expected");
583 key = strdup(lex->value.string);
587 lex_scan(lex, error);
588 if(lex->token != ':') {
590 error_set(error, lex, "':' expected");
594 lex_scan(lex, error);
595 value = parse_value(lex, error);
601 if(json_object_set_nocheck(object, key, value)) {
610 lex_scan(lex, error);
611 if(lex->token != ',')
614 lex_scan(lex, error);
617 if(lex->token != '}') {
618 error_set(error, lex, "'}' expected");
629 static json_t *parse_array(lex_t *lex, json_error_t *error)
631 json_t *array = json_array();
635 lex_scan(lex, error);
636 if(lex->token == ']')
640 json_t *elem = parse_value(lex, error);
644 if(json_array_append(array, elem)) {
650 lex_scan(lex, error);
651 if(lex->token != ',')
654 lex_scan(lex, error);
657 if(lex->token != ']') {
658 error_set(error, lex, "']' expected");
669 static json_t *parse_value(lex_t *lex, json_error_t *error)
675 json = json_string_nocheck(lex->value.string);
679 case TOKEN_INTEGER: {
680 json = json_integer(lex->value.integer);
685 json = json_real(lex->value.real);
702 json = parse_object(lex, error);
706 json = parse_array(lex, error);
710 error_set(error, lex, "invalid token");
714 error_set(error, lex, "unexpected token");
724 json_t *parse_json(lex_t *lex, json_error_t *error)
728 lex_scan(lex, error);
729 if(lex->token != '[' && lex->token != '{') {
730 error_set(error, lex, "'[' or '{' expected");
734 return parse_value(lex, error);
737 json_t *json_load(const char *path, json_error_t *error)
742 fp = fopen(path, "r");
745 error_set(error, NULL, "unable to open %s: %s",
746 path, strerror(errno));
750 result = json_loadf(fp, error);
762 static int string_get(void *data)
765 string_data_t *stream = (string_data_t *)data;
766 c = stream->data[stream->pos];
776 static int string_eof(void *data)
778 string_data_t *stream = (string_data_t *)data;
779 return (stream->data[stream->pos] == '\0');
782 json_t *json_loads(const char *string, json_error_t *error)
787 string_data_t stream_data = {
792 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
795 result = parse_json(&lex, error);
799 lex_scan(&lex, error);
800 if(lex.token != TOKEN_EOF) {
801 error_set(error, &lex, "end of file expected");
811 json_t *json_loadf(FILE *input, json_error_t *error)
816 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
819 result = parse_json(&lex, error);