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 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 ***/
117 stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
122 stream->stream_pos = 0;
123 stream->buffer[0] = '\0';
124 stream->buffer_pos = 0;
127 static char stream_get(stream_t *stream, json_error_t *error)
131 if(!stream->buffer[stream->buffer_pos])
133 stream->buffer[0] = stream->get(stream->data);
134 stream->buffer_pos = 0;
136 c = stream->buffer[0];
138 if((unsigned char)c >= 0x80 && c != (char)EOF)
140 /* multi-byte UTF-8 sequence */
143 count = utf8_check_first(c);
149 for(i = 1; i < count; i++)
150 stream->buffer[i] = stream->get(stream->data);
152 if(!utf8_check_full(stream->buffer, count, NULL))
155 stream->stream_pos += count;
156 stream->buffer[count] = '\0';
159 stream->buffer[1] = '\0';
160 stream->stream_pos++;
164 return stream->buffer[stream->buffer_pos++];
167 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
168 (unsigned char)c, stream->stream_pos);
170 stream->buffer[0] = EOF;
171 stream->buffer[1] = '\0';
172 stream->buffer_pos = 1;
177 static void stream_unget(stream_t *stream, char c)
179 assert(stream->buffer_pos > 0);
180 stream->buffer_pos--;
181 assert(stream->buffer[stream->buffer_pos] == c);
185 static int lex_get(lex_t *lex, json_error_t *error)
187 return stream_get(&lex->stream, error);
190 static int lex_eof(lex_t *lex)
192 return lex->stream.eof(lex->stream.data);
195 static void lex_save(lex_t *lex, char c)
197 strbuffer_append_byte(&lex->saved_text, c);
200 static int lex_get_save(lex_t *lex, json_error_t *error)
202 char c = stream_get(&lex->stream, error);
207 static void lex_unget_unsave(lex_t *lex, char c)
210 stream_unget(&lex->stream, c);
211 d = strbuffer_pop(&lex->saved_text);
215 static void lex_save_cached(lex_t *lex)
217 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
219 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
220 lex->stream.buffer_pos++;
224 /* assumes that str points to 'u' plus at least 4 valid hex digits */
225 static int32_t decode_unicode_escape(const char *str)
230 assert(str[0] == 'u');
232 for(i = 1; i <= 4; i++) {
238 value += c - 'a' + 10;
240 value += c - 'A' + 10;
248 static void lex_scan_string(lex_t *lex, json_error_t *error)
255 lex->value.string = NULL;
256 lex->token = TOKEN_INVALID;
258 c = lex_get_save(lex, error);
262 lex_unget_unsave(lex, c);
264 error_set(error, lex, "premature end of input");
268 else if((unsigned char)c <= 0x1F) {
269 /* control character */
270 lex_unget_unsave(lex, c);
272 error_set(error, lex, "unexpected newline", c);
274 error_set(error, lex, "control character 0x%x", c);
279 c = lex_get_save(lex, error);
281 c = lex_get_save(lex, error);
282 for(i = 0; i < 4; i++) {
284 lex_unget_unsave(lex, c);
285 error_set(error, lex, "invalid escape");
288 c = lex_get_save(lex, error);
291 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
292 c == 'f' || c == 'n' || c == 'r' || c == 't')
293 c = lex_get_save(lex, error);
295 lex_unget_unsave(lex, c);
296 error_set(error, lex, "invalid escape");
301 c = lex_get_save(lex, error);
304 /* the actual value is at most of the same length as the source
306 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
307 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
308 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
309 are converted to 4 bytes
311 lex->value.string = malloc(lex->saved_text.length + 1);
312 if(!lex->value.string) {
313 /* this is not very nice, since TOKEN_INVALID is returned */
318 t = lex->value.string;
320 /* + 1 to skip the " */
321 p = strbuffer_value(&lex->saved_text) + 1;
331 value = decode_unicode_escape(p);
334 if(0xD800 <= value && value <= 0xDBFF) {
336 if(*p == '\\' && *(p + 1) == 'u') {
337 int32_t value2 = decode_unicode_escape(++p);
340 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
341 /* valid second surrogate */
343 ((value - 0xD800) << 10) +
348 /* invalid second surrogate */
349 error_set(error, lex,
350 "invalid Unicode '\\u%04X\\u%04X'",
356 /* no second surrogate */
357 error_set(error, lex, "invalid Unicode '\\u%04X'",
362 else if(0xDC00 <= value && value <= 0xDFFF) {
363 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
368 error_set(error, lex, "\\u0000 is not allowed");
372 if(utf8_encode(value, buffer, &length))
375 memcpy(t, buffer, length);
380 case '"': case '\\': case '/':
382 case 'b': *t = '\b'; break;
383 case 'f': *t = '\f'; break;
384 case 'n': *t = '\n'; break;
385 case 'r': *t = '\r'; break;
386 case 't': *t = '\t'; break;
397 lex->token = TOKEN_STRING;
401 free(lex->value.string);
404 static int lex_scan_number(lex_t *lex, char c, json_error_t *error)
406 const char *saved_text;
410 lex->token = TOKEN_INVALID;
413 c = lex_get_save(lex, error);
416 c = lex_get_save(lex, error);
418 lex_unget_unsave(lex, c);
422 else if(isdigit(c)) {
423 c = lex_get_save(lex, error);
425 c = lex_get_save(lex, error);
428 lex_unget_unsave(lex, c);
432 if(c != '.' && c != 'E' && c != 'e') {
435 lex_unget_unsave(lex, c);
437 saved_text = strbuffer_value(&lex->saved_text);
438 value = strtol(saved_text, &end, 10);
439 assert(end == saved_text + lex->saved_text.length);
441 if(value == LONG_MAX && errno == ERANGE) {
442 error_set(error, lex, "too big integer");
445 else if(value == LONG_MIN && errno == ERANGE) {
446 error_set(error, lex, "too big negative integer");
450 lex->token = TOKEN_INTEGER;
451 lex->value.integer = value;
456 c = lex_get(lex, error);
461 c = lex_get_save(lex, error);
463 c = lex_get_save(lex, error);
466 if(c == 'E' || c == 'e') {
467 c = lex_get_save(lex, error);
468 if(c == '+' || c == '-')
469 c = lex_get_save(lex, error);
472 lex_unget_unsave(lex, c);
476 c = lex_get_save(lex, error);
478 c = lex_get_save(lex, error);
481 lex_unget_unsave(lex, c);
483 saved_text = strbuffer_value(&lex->saved_text);
484 value = strtod(saved_text, &end);
485 assert(end == saved_text + lex->saved_text.length);
487 if(errno == ERANGE && value != 0) {
488 error_set(error, lex, "real number overflow");
492 lex->token = TOKEN_REAL;
493 lex->value.real = value;
500 static int lex_scan(lex_t *lex, json_error_t *error)
504 strbuffer_clear(&lex->saved_text);
506 if(lex->token == TOKEN_STRING) {
507 free(lex->value.string);
508 lex->value.string = NULL;
511 c = lex_get(lex, error);
512 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
517 c = lex_get(lex, error);
522 lex->token = TOKEN_EOF;
524 lex->token = TOKEN_INVALID;
530 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
534 lex_scan_string(lex, error);
536 else if(isdigit(c) || c == '-') {
537 if(lex_scan_number(lex, c, error))
541 else if(isupper(c) || islower(c)) {
542 /* eat up the whole identifier for clearer error messages */
543 const char *saved_text;
545 c = lex_get_save(lex, error);
546 while(isupper(c) || islower(c))
547 c = lex_get_save(lex, error);
548 lex_unget_unsave(lex, c);
550 saved_text = strbuffer_value(&lex->saved_text);
552 if(strcmp(saved_text, "true") == 0)
553 lex->token = TOKEN_TRUE;
554 else if(strcmp(saved_text, "false") == 0)
555 lex->token = TOKEN_FALSE;
556 else if(strcmp(saved_text, "null") == 0)
557 lex->token = TOKEN_NULL;
559 lex->token = TOKEN_INVALID;
563 /* save the rest of the input UTF-8 sequence to get an error
564 message of valid UTF-8 */
565 lex_save_cached(lex);
566 lex->token = TOKEN_INVALID;
573 static char *lex_steal_string(lex_t *lex)
576 if(lex->token == TOKEN_STRING)
578 result = lex->value.string;
579 lex->value.string = NULL;
584 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
586 stream_init(&lex->stream, get, eof, data);
587 if(strbuffer_init(&lex->saved_text))
590 lex->token = TOKEN_INVALID;
596 static void lex_close(lex_t *lex)
598 if(lex->token == TOKEN_STRING)
599 free(lex->value.string);
600 strbuffer_close(&lex->saved_text);
606 static json_t *parse_value(lex_t *lex, json_error_t *error);
608 static json_t *parse_object(lex_t *lex, json_error_t *error)
610 json_t *object = json_object();
614 lex_scan(lex, error);
615 if(lex->token == '}')
622 if(lex->token != TOKEN_STRING) {
623 error_set(error, lex, "string or '}' expected");
627 key = lex_steal_string(lex);
631 lex_scan(lex, error);
632 if(lex->token != ':') {
634 error_set(error, lex, "':' expected");
638 lex_scan(lex, error);
639 value = parse_value(lex, error);
645 if(json_object_set_nocheck(object, key, value)) {
654 lex_scan(lex, error);
655 if(lex->token != ',')
658 lex_scan(lex, error);
661 if(lex->token != '}') {
662 error_set(error, lex, "'}' expected");
673 static json_t *parse_array(lex_t *lex, json_error_t *error)
675 json_t *array = json_array();
679 lex_scan(lex, error);
680 if(lex->token == ']')
684 json_t *elem = parse_value(lex, error);
688 if(json_array_append(array, elem)) {
694 lex_scan(lex, error);
695 if(lex->token != ',')
698 lex_scan(lex, error);
701 if(lex->token != ']') {
702 error_set(error, lex, "']' expected");
713 static json_t *parse_value(lex_t *lex, json_error_t *error)
719 json = json_string_nocheck(lex->value.string);
723 case TOKEN_INTEGER: {
724 json = json_integer(lex->value.integer);
729 json = json_real(lex->value.real);
746 json = parse_object(lex, error);
750 json = parse_array(lex, error);
754 error_set(error, lex, "invalid token");
758 error_set(error, lex, "unexpected token");
768 static json_t *parse_json(lex_t *lex, json_error_t *error)
772 lex_scan(lex, error);
773 if(lex->token != '[' && lex->token != '{') {
774 error_set(error, lex, "'[' or '{' expected");
778 return parse_value(lex, error);
787 static int string_get(void *data)
790 string_data_t *stream = (string_data_t *)data;
791 c = stream->data[stream->pos];
801 static int string_eof(void *data)
803 string_data_t *stream = (string_data_t *)data;
804 return (stream->data[stream->pos] == '\0');
807 json_t *json_loads(const char *string, json_error_t *error)
812 string_data_t stream_data = {
817 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
820 result = parse_json(&lex, error);
824 lex_scan(&lex, error);
825 if(lex.token != TOKEN_EOF) {
826 error_set(error, &lex, "end of file expected");
836 json_t *json_loadf(FILE *input, json_error_t *error)
841 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
844 result = parse_json(&lex, error);
848 lex_scan(&lex, error);
849 if(lex.token != TOKEN_EOF) {
850 error_set(error, &lex, "end of file expected");
860 json_t *json_load_file(const char *path, json_error_t *error)
867 fp = fopen(path, "r");
870 error_set(error, NULL, "unable to open %s: %s",
871 path, strerror(errno));
875 result = json_loadf(fp, error);