2 * Copyright (c) 2009 Petri Lehtinen <petri@digip.org>
4 * Jansson is free software; you can redistribute it and/or modify
5 * it under the terms of the MIT license. See LICENSE for details.
19 #include "jansson_private.h"
20 #include "strbuffer.h"
23 #define TOKEN_INVALID -1
25 #define TOKEN_STRING 256
26 #define TOKEN_INTEGER 257
27 #define TOKEN_REAL 258
28 #define TOKEN_TRUE 259
29 #define TOKEN_FALSE 260
30 #define TOKEN_NULL 261
32 /* read one byte from stream, return EOF on end of file */
33 typedef int (*get_func)(void *data);
35 /* return non-zero if end of file has been reached */
36 typedef int (*eof_func)(void *data);
50 strbuffer_t saved_text;
61 /*** error reporting ***/
63 static void error_init(json_error_t *error)
67 error->text[0] = '\0';
72 static void error_set(json_error_t *error, const lex_t *lex,
76 char text[JSON_ERROR_TEXT_LENGTH];
78 if(!error || error->text[0] != '\0') {
79 /* error already set */
84 vsnprintf(text, JSON_ERROR_TEXT_LENGTH, msg, ap);
89 const char *saved_text = strbuffer_value(&lex->saved_text);
90 error->line = lex->line;
91 if(saved_text && saved_text[0])
93 if(lex->saved_text.length <= 20) {
94 snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
95 "%s near '%s'", text, saved_text);
98 snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
102 snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
103 "%s near end of file", text);
109 snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
114 /*** lexical analyzer ***/
116 void stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
121 stream->stream_pos = 0;
122 stream->buffer[0] = '\0';
123 stream->buffer_pos = 0;
126 static char stream_get(stream_t *stream, json_error_t *error)
130 if(!stream->buffer[stream->buffer_pos])
132 stream->buffer[0] = stream->get(stream->data);
133 stream->buffer_pos = 0;
135 c = stream->buffer[0];
137 if((unsigned char)c >= 0x80 && c != (char)EOF)
139 /* multi-byte UTF-8 sequence */
142 count = utf8_check_first(c);
148 for(i = 1; i < count; i++)
149 stream->buffer[i] = stream->get(stream->data);
151 if(!utf8_check_full(stream->buffer, count))
154 stream->stream_pos += count;
155 stream->buffer[count] = '\0';
158 stream->buffer[1] = '\0';
159 stream->stream_pos++;
163 return stream->buffer[stream->buffer_pos++];
166 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
167 (unsigned char)c, stream->stream_pos);
169 stream->buffer[0] = EOF;
170 stream->buffer[1] = '\0';
171 stream->buffer_pos = 1;
176 static void stream_unget(stream_t *stream, char c)
178 assert(stream->buffer_pos > 0);
179 stream->buffer_pos--;
180 assert(stream->buffer[stream->buffer_pos] == c);
184 static int lex_get(lex_t *lex, json_error_t *error)
186 return stream_get(&lex->stream, error);
189 static int lex_eof(lex_t *lex)
191 return lex->stream.eof(lex->stream.data);
194 static void lex_save(lex_t *lex, char c)
196 strbuffer_append_byte(&lex->saved_text, c);
199 static int lex_get_save(lex_t *lex, json_error_t *error)
201 char c = stream_get(&lex->stream, error);
206 static void lex_unget_unsave(lex_t *lex, char c)
209 stream_unget(&lex->stream, c);
210 d = strbuffer_pop(&lex->saved_text);
214 static void lex_save_cached(lex_t *lex)
216 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
218 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
219 lex->stream.buffer_pos++;
223 /* assumes that str points to 'u' plus at least 4 valid hex digits */
224 static int decode_unicode_escape(const char *str)
229 assert(str[0] == 'u');
231 for(i = 1; i <= 4; i++) {
237 value += c - 'a' + 10;
239 value += c - 'A' + 10;
247 static void lex_scan_string(lex_t *lex, json_error_t *error)
254 lex->value.string = NULL;
255 lex->token = TOKEN_INVALID;
257 c = lex_get_save(lex, error);
261 lex_unget_unsave(lex, c);
263 error_set(error, lex, "premature end of input");
267 else if((unsigned char)c <= 0x1F) {
268 /* control character */
269 lex_unget_unsave(lex, c);
271 error_set(error, lex, "unexpected newline", c);
273 error_set(error, lex, "control character 0x%x", c);
278 c = lex_get_save(lex, error);
280 c = lex_get_save(lex, error);
281 for(i = 0; i < 4; i++) {
283 lex_unget_unsave(lex, c);
284 error_set(error, lex, "invalid escape");
287 c = lex_get_save(lex, error);
290 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
291 c == 'f' || c == 'n' || c == 'r' || c == 't')
292 c = lex_get_save(lex, error);
294 lex_unget_unsave(lex, c);
295 error_set(error, lex, "invalid escape");
300 c = lex_get_save(lex, error);
303 /* the actual value is at most of the same length as the source
305 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
306 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
307 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
308 are converted to 4 bytes
310 lex->value.string = malloc(lex->saved_text.length + 1);
311 if(!lex->value.string) {
312 /* this is not very nice, since TOKEN_INVALID is returned */
317 t = lex->value.string;
319 /* + 1 to skip the " */
320 p = strbuffer_value(&lex->saved_text) + 1;
330 value = decode_unicode_escape(p);
333 if(0xD800 <= value && value <= 0xDBFF) {
335 if(*p == '\\' && *(p + 1) == 'u') {
336 int value2 = decode_unicode_escape(++p);
339 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
340 /* valid second surrogate */
341 value = ((value - 0xD800) << 10) +
346 /* invalid second surrogate */
347 error_set(error, lex,
348 "invalid Unicode '\\u%04X\\u%04X'",
354 /* no second surrogate */
355 error_set(error, lex, "invalid Unicode '\\u%04X'",
360 else if(0xDC00 <= value && value <= 0xDFFF) {
361 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
366 error_set(error, lex, "\\u0000 is not allowed");
370 if(utf8_encode(value, buffer, &length))
373 memcpy(t, buffer, length);
378 case '"': case '\\': case '/':
380 case 'b': *t = '\b'; break;
381 case 'f': *t = '\f'; break;
382 case 'n': *t = '\n'; break;
383 case 'r': *t = '\r'; break;
384 case 't': *t = '\t'; break;
395 lex->token = TOKEN_STRING;
399 free(lex->value.string);
402 static int lex_scan_number(lex_t *lex, char c, json_error_t *error)
404 const char *saved_text;
408 lex->token = TOKEN_INVALID;
411 c = lex_get_save(lex, error);
414 c = lex_get_save(lex, error);
416 lex_unget_unsave(lex, c);
420 else /* c != '0' */ {
421 c = lex_get_save(lex, error);
423 c = lex_get_save(lex, error);
426 if(c != '.' && c != 'E' && c != 'e') {
429 lex_unget_unsave(lex, c);
431 saved_text = strbuffer_value(&lex->saved_text);
432 value = strtol(saved_text, &end, 10);
433 assert(end == saved_text + lex->saved_text.length);
435 if((value == LONG_MAX && errno == ERANGE) || value > INT_MAX) {
436 error_set(error, lex, "too big integer");
439 else if((value == LONG_MIN && errno == ERANGE) || value < INT_MIN) {
440 error_set(error, lex, "too big negative integer");
444 lex->token = TOKEN_INTEGER;
445 lex->value.integer = (int)value;
450 c = lex_get(lex, error);
455 c = lex_get_save(lex, error);
457 c = lex_get_save(lex, error);
460 if(c == 'E' || c == 'e') {
461 c = lex_get_save(lex, error);
462 if(c == '+' || c == '-')
463 c = lex_get_save(lex, error);
466 lex_unget_unsave(lex, c);
470 c = lex_get_save(lex, error);
472 c = lex_get_save(lex, error);
475 lex_unget_unsave(lex, c);
477 saved_text = strbuffer_value(&lex->saved_text);
478 value = strtod(saved_text, &end);
479 assert(end == saved_text + lex->saved_text.length);
481 if(value == 0 && errno == ERANGE) {
482 error_set(error, lex, "real number underflow");
486 /* Cannot test for +/-HUGE_VAL because the HUGE_VAL constant is
487 only defined in C99 mode. So let's trust in sole errno. */
488 else if(errno == ERANGE) {
489 error_set(error, lex, "real number overflow");
493 lex->token = TOKEN_REAL;
494 lex->value.real = value;
501 static int lex_scan(lex_t *lex, json_error_t *error)
505 strbuffer_clear(&lex->saved_text);
507 if(lex->token == TOKEN_STRING) {
508 free(lex->value.string);
509 lex->value.string = NULL;
512 c = lex_get(lex, error);
513 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
518 c = lex_get(lex, error);
523 lex->token = TOKEN_EOF;
525 lex->token = TOKEN_INVALID;
531 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
535 lex_scan_string(lex, error);
537 else if(isdigit(c) || c == '-') {
538 if(lex_scan_number(lex, c, error))
542 else if(isupper(c) || islower(c)) {
543 /* eat up the whole identifier for clearer error messages */
544 const char *saved_text;
546 c = lex_get_save(lex, error);
547 while(isupper(c) || islower(c))
548 c = lex_get_save(lex, error);
549 lex_unget_unsave(lex, c);
551 saved_text = strbuffer_value(&lex->saved_text);
553 if(strcmp(saved_text, "true") == 0)
554 lex->token = TOKEN_TRUE;
555 else if(strcmp(saved_text, "false") == 0)
556 lex->token = TOKEN_FALSE;
557 else if(strcmp(saved_text, "null") == 0)
558 lex->token = TOKEN_NULL;
560 lex->token = TOKEN_INVALID;
564 /* save the rest of the input UTF-8 sequence to get an error
565 message of valid UTF-8 */
566 lex_save_cached(lex);
567 lex->token = TOKEN_INVALID;
574 static char *lex_steal_string(lex_t *lex)
577 if(lex->token == TOKEN_STRING)
579 result = lex->value.string;
580 lex->value.string = NULL;
585 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
587 stream_init(&lex->stream, get, eof, data);
588 if(strbuffer_init(&lex->saved_text))
591 lex->token = TOKEN_INVALID;
597 static void lex_close(lex_t *lex)
599 if(lex->token == TOKEN_STRING)
600 free(lex->value.string);
601 strbuffer_close(&lex->saved_text);
607 static json_t *parse_value(lex_t *lex, json_error_t *error);
609 static json_t *parse_object(lex_t *lex, json_error_t *error)
611 json_t *object = json_object();
615 lex_scan(lex, error);
616 if(lex->token == '}')
623 if(lex->token != TOKEN_STRING) {
624 error_set(error, lex, "string or '}' expected");
628 key = lex_steal_string(lex);
632 lex_scan(lex, error);
633 if(lex->token != ':') {
635 error_set(error, lex, "':' expected");
639 lex_scan(lex, error);
640 value = parse_value(lex, error);
646 if(json_object_set_nocheck(object, key, value)) {
655 lex_scan(lex, error);
656 if(lex->token != ',')
659 lex_scan(lex, error);
662 if(lex->token != '}') {
663 error_set(error, lex, "'}' expected");
674 static json_t *parse_array(lex_t *lex, json_error_t *error)
676 json_t *array = json_array();
680 lex_scan(lex, error);
681 if(lex->token == ']')
685 json_t *elem = parse_value(lex, error);
689 if(json_array_append(array, elem)) {
695 lex_scan(lex, error);
696 if(lex->token != ',')
699 lex_scan(lex, error);
702 if(lex->token != ']') {
703 error_set(error, lex, "']' expected");
714 static json_t *parse_value(lex_t *lex, json_error_t *error)
720 json = json_string_nocheck(lex->value.string);
724 case TOKEN_INTEGER: {
725 json = json_integer(lex->value.integer);
730 json = json_real(lex->value.real);
747 json = parse_object(lex, error);
751 json = parse_array(lex, error);
755 error_set(error, lex, "invalid token");
759 error_set(error, lex, "unexpected token");
769 json_t *parse_json(lex_t *lex, json_error_t *error)
773 lex_scan(lex, error);
774 if(lex->token != '[' && lex->token != '{') {
775 error_set(error, lex, "'[' or '{' expected");
779 return parse_value(lex, error);
788 static int string_get(void *data)
791 string_data_t *stream = (string_data_t *)data;
792 c = stream->data[stream->pos];
802 static int string_eof(void *data)
804 string_data_t *stream = (string_data_t *)data;
805 return (stream->data[stream->pos] == '\0');
808 json_t *json_loads(const char *string, json_error_t *error)
813 string_data_t stream_data = {
818 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
821 result = parse_json(&lex, error);
825 lex_scan(&lex, error);
826 if(lex.token != TOKEN_EOF) {
827 error_set(error, &lex, "end of file expected");
837 json_t *json_loadf(FILE *input, json_error_t *error)
842 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
845 result = parse_json(&lex, error);
849 lex_scan(&lex, error);
850 if(lex.token != TOKEN_EOF) {
851 error_set(error, &lex, "end of file expected");
861 json_t *json_load_file(const char *path, json_error_t *error)
866 fp = fopen(path, "r");
869 error_set(error, NULL, "unable to open %s: %s",
870 path, strerror(errno));
874 result = json_loadf(fp, error);