2 * Copyright (c) 2009, 2010 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 const char *json_error_msg(const json_error_t *error)
65 return error ? error->msg : NULL;
68 int json_error_line(const json_error_t *error)
70 return error ? error->line : -1;
73 static void error_init(json_error_t **error)
79 static void error_set(json_error_t **error, const lex_t *lex,
83 char text[JSON_ERROR_MSG_LENGTH];
85 if(!error || *error) {
86 /* error not given or already set */
90 *error = malloc(sizeof(json_error_t));
95 vsnprintf(text, JSON_ERROR_MSG_LENGTH, msg, ap);
100 const char *saved_text = strbuffer_value(&lex->saved_text);
101 (*error)->line = lex->line;
102 if(saved_text && saved_text[0])
104 if(lex->saved_text.length <= 20) {
105 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH,
106 "%s near '%s'", text, saved_text);
109 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH, "%s", text);
113 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH,
114 "%s near end of file", text);
120 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH, "%s", text);
125 /*** lexical analyzer ***/
128 stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
133 stream->stream_pos = 0;
134 stream->buffer[0] = '\0';
135 stream->buffer_pos = 0;
138 static char stream_get(stream_t *stream, json_error_t **error)
142 if(!stream->buffer[stream->buffer_pos])
144 stream->buffer[0] = stream->get(stream->data);
145 stream->buffer_pos = 0;
147 c = stream->buffer[0];
149 if((unsigned char)c >= 0x80 && c != (char)EOF)
151 /* multi-byte UTF-8 sequence */
154 count = utf8_check_first(c);
160 for(i = 1; i < count; i++)
161 stream->buffer[i] = stream->get(stream->data);
163 if(!utf8_check_full(stream->buffer, count, NULL))
166 stream->stream_pos += count;
167 stream->buffer[count] = '\0';
170 stream->buffer[1] = '\0';
171 stream->stream_pos++;
175 return stream->buffer[stream->buffer_pos++];
178 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
179 (unsigned char)c, stream->stream_pos);
181 stream->buffer[0] = EOF;
182 stream->buffer[1] = '\0';
183 stream->buffer_pos = 1;
188 static void stream_unget(stream_t *stream, char c)
190 assert(stream->buffer_pos > 0);
191 stream->buffer_pos--;
192 assert(stream->buffer[stream->buffer_pos] == c);
196 static int lex_get(lex_t *lex, json_error_t **error)
198 return stream_get(&lex->stream, error);
201 static int lex_eof(lex_t *lex)
203 return lex->stream.eof(lex->stream.data);
206 static void lex_save(lex_t *lex, char c)
208 strbuffer_append_byte(&lex->saved_text, c);
211 static int lex_get_save(lex_t *lex, json_error_t **error)
213 char c = stream_get(&lex->stream, error);
218 static void lex_unget_unsave(lex_t *lex, char c)
221 stream_unget(&lex->stream, c);
222 d = strbuffer_pop(&lex->saved_text);
226 static void lex_save_cached(lex_t *lex)
228 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
230 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
231 lex->stream.buffer_pos++;
235 /* assumes that str points to 'u' plus at least 4 valid hex digits */
236 static int32_t decode_unicode_escape(const char *str)
241 assert(str[0] == 'u');
243 for(i = 1; i <= 4; i++) {
249 value += c - 'a' + 10;
251 value += c - 'A' + 10;
259 static void lex_scan_string(lex_t *lex, json_error_t **error)
266 lex->value.string = NULL;
267 lex->token = TOKEN_INVALID;
269 c = lex_get_save(lex, error);
273 lex_unget_unsave(lex, c);
275 error_set(error, lex, "premature end of input");
279 else if((unsigned char)c <= 0x1F) {
280 /* control character */
281 lex_unget_unsave(lex, c);
283 error_set(error, lex, "unexpected newline", c);
285 error_set(error, lex, "control character 0x%x", c);
290 c = lex_get_save(lex, error);
292 c = lex_get_save(lex, error);
293 for(i = 0; i < 4; i++) {
295 lex_unget_unsave(lex, c);
296 error_set(error, lex, "invalid escape");
299 c = lex_get_save(lex, error);
302 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
303 c == 'f' || c == 'n' || c == 'r' || c == 't')
304 c = lex_get_save(lex, error);
306 lex_unget_unsave(lex, c);
307 error_set(error, lex, "invalid escape");
312 c = lex_get_save(lex, error);
315 /* the actual value is at most of the same length as the source
317 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
318 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
319 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
320 are converted to 4 bytes
322 lex->value.string = malloc(lex->saved_text.length + 1);
323 if(!lex->value.string) {
324 /* this is not very nice, since TOKEN_INVALID is returned */
329 t = lex->value.string;
331 /* + 1 to skip the " */
332 p = strbuffer_value(&lex->saved_text) + 1;
342 value = decode_unicode_escape(p);
345 if(0xD800 <= value && value <= 0xDBFF) {
347 if(*p == '\\' && *(p + 1) == 'u') {
348 int32_t value2 = decode_unicode_escape(++p);
351 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
352 /* valid second surrogate */
354 ((value - 0xD800) << 10) +
359 /* invalid second surrogate */
360 error_set(error, lex,
361 "invalid Unicode '\\u%04X\\u%04X'",
367 /* no second surrogate */
368 error_set(error, lex, "invalid Unicode '\\u%04X'",
373 else if(0xDC00 <= value && value <= 0xDFFF) {
374 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
379 error_set(error, lex, "\\u0000 is not allowed");
383 if(utf8_encode(value, buffer, &length))
386 memcpy(t, buffer, length);
391 case '"': case '\\': case '/':
393 case 'b': *t = '\b'; break;
394 case 'f': *t = '\f'; break;
395 case 'n': *t = '\n'; break;
396 case 'r': *t = '\r'; break;
397 case 't': *t = '\t'; break;
408 lex->token = TOKEN_STRING;
412 free(lex->value.string);
415 #if JSON_INTEGER_IS_LONG_LONG
416 #define json_strtoint strtoll
418 #define json_strtoint strtol
421 static int lex_scan_number(lex_t *lex, char c, json_error_t **error)
423 const char *saved_text;
427 lex->token = TOKEN_INVALID;
430 c = lex_get_save(lex, error);
433 c = lex_get_save(lex, error);
435 lex_unget_unsave(lex, c);
439 else if(isdigit(c)) {
440 c = lex_get_save(lex, error);
442 c = lex_get_save(lex, error);
445 lex_unget_unsave(lex, c);
449 if(c != '.' && c != 'E' && c != 'e') {
452 lex_unget_unsave(lex, c);
454 saved_text = strbuffer_value(&lex->saved_text);
457 value = json_strtoint(saved_text, &end, 10);
458 if(errno == ERANGE) {
460 error_set(error, lex, "too big negative integer");
462 error_set(error, lex, "too big integer");
466 assert(end == saved_text + lex->saved_text.length);
468 lex->token = TOKEN_INTEGER;
469 lex->value.integer = value;
474 c = lex_get(lex, error);
479 c = lex_get_save(lex, error);
481 c = lex_get_save(lex, error);
484 if(c == 'E' || c == 'e') {
485 c = lex_get_save(lex, error);
486 if(c == '+' || c == '-')
487 c = lex_get_save(lex, error);
490 lex_unget_unsave(lex, c);
494 c = lex_get_save(lex, error);
496 c = lex_get_save(lex, error);
499 lex_unget_unsave(lex, c);
501 saved_text = strbuffer_value(&lex->saved_text);
502 value = strtod(saved_text, &end);
503 assert(end == saved_text + lex->saved_text.length);
505 if(errno == ERANGE && value != 0) {
506 error_set(error, lex, "real number overflow");
510 lex->token = TOKEN_REAL;
511 lex->value.real = value;
518 static int lex_scan(lex_t *lex, json_error_t **error)
522 strbuffer_clear(&lex->saved_text);
524 if(lex->token == TOKEN_STRING) {
525 free(lex->value.string);
526 lex->value.string = NULL;
529 c = lex_get(lex, error);
530 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
535 c = lex_get(lex, error);
540 lex->token = TOKEN_EOF;
542 lex->token = TOKEN_INVALID;
548 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
552 lex_scan_string(lex, error);
554 else if(isdigit(c) || c == '-') {
555 if(lex_scan_number(lex, c, error))
559 else if(isupper(c) || islower(c)) {
560 /* eat up the whole identifier for clearer error messages */
561 const char *saved_text;
563 c = lex_get_save(lex, error);
564 while(isupper(c) || islower(c))
565 c = lex_get_save(lex, error);
566 lex_unget_unsave(lex, c);
568 saved_text = strbuffer_value(&lex->saved_text);
570 if(strcmp(saved_text, "true") == 0)
571 lex->token = TOKEN_TRUE;
572 else if(strcmp(saved_text, "false") == 0)
573 lex->token = TOKEN_FALSE;
574 else if(strcmp(saved_text, "null") == 0)
575 lex->token = TOKEN_NULL;
577 lex->token = TOKEN_INVALID;
581 /* save the rest of the input UTF-8 sequence to get an error
582 message of valid UTF-8 */
583 lex_save_cached(lex);
584 lex->token = TOKEN_INVALID;
591 static char *lex_steal_string(lex_t *lex)
594 if(lex->token == TOKEN_STRING)
596 result = lex->value.string;
597 lex->value.string = NULL;
602 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
604 stream_init(&lex->stream, get, eof, data);
605 if(strbuffer_init(&lex->saved_text))
608 lex->token = TOKEN_INVALID;
614 static void lex_close(lex_t *lex)
616 if(lex->token == TOKEN_STRING)
617 free(lex->value.string);
618 strbuffer_close(&lex->saved_text);
624 static json_t *parse_value(lex_t *lex, json_error_t **error);
626 static json_t *parse_object(lex_t *lex, json_error_t **error)
628 json_t *object = json_object();
632 lex_scan(lex, error);
633 if(lex->token == '}')
640 if(lex->token != TOKEN_STRING) {
641 error_set(error, lex, "string or '}' expected");
645 key = lex_steal_string(lex);
649 lex_scan(lex, error);
650 if(lex->token != ':') {
652 error_set(error, lex, "':' expected");
656 lex_scan(lex, error);
657 value = parse_value(lex, error);
663 if(json_object_set_nocheck(object, key, value)) {
672 lex_scan(lex, error);
673 if(lex->token != ',')
676 lex_scan(lex, error);
679 if(lex->token != '}') {
680 error_set(error, lex, "'}' expected");
691 static json_t *parse_array(lex_t *lex, json_error_t **error)
693 json_t *array = json_array();
697 lex_scan(lex, error);
698 if(lex->token == ']')
702 json_t *elem = parse_value(lex, error);
706 if(json_array_append(array, elem)) {
712 lex_scan(lex, error);
713 if(lex->token != ',')
716 lex_scan(lex, error);
719 if(lex->token != ']') {
720 error_set(error, lex, "']' expected");
731 static json_t *parse_value(lex_t *lex, json_error_t **error)
737 json = json_string_nocheck(lex->value.string);
741 case TOKEN_INTEGER: {
742 json = json_integer(lex->value.integer);
747 json = json_real(lex->value.real);
764 json = parse_object(lex, error);
768 json = parse_array(lex, error);
772 error_set(error, lex, "invalid token");
776 error_set(error, lex, "unexpected token");
786 static json_t *parse_json(lex_t *lex, json_error_t **error)
790 lex_scan(lex, error);
791 if(lex->token != '[' && lex->token != '{') {
792 error_set(error, lex, "'[' or '{' expected");
796 return parse_value(lex, error);
805 static int string_get(void *data)
808 string_data_t *stream = (string_data_t *)data;
809 c = stream->data[stream->pos];
819 static int string_eof(void *data)
821 string_data_t *stream = (string_data_t *)data;
822 return (stream->data[stream->pos] == '\0');
825 json_t *json_loads(const char *string, size_t flags, json_error_t **error)
829 (void)flags; /* unused */
831 string_data_t stream_data = {string, 0};
833 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
836 result = parse_json(&lex, error);
840 lex_scan(&lex, error);
841 if(lex.token != TOKEN_EOF) {
842 error_set(error, &lex, "end of file expected");
852 json_t *json_loadf(FILE *input, size_t flags, json_error_t **error)
856 (void)flags; /* unused */
858 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
861 result = parse_json(&lex, error);
865 lex_scan(&lex, error);
866 if(lex.token != TOKEN_EOF) {
867 error_set(error, &lex, "end of file expected");
877 json_t *json_load_file(const char *path, size_t flags, json_error_t **error)
884 fp = fopen(path, "r");
887 error_set(error, NULL, "unable to open %s: %s",
888 path, strerror(errno));
892 result = json_loadf(fp, flags, error);