2 * Copyright (c) 2009-2011 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.
17 #include "jansson_private.h"
18 #include "strbuffer.h"
21 #define STREAM_STATE_OK 0
22 #define STREAM_STATE_EOF -1
23 #define STREAM_STATE_ERROR -2
25 #define TOKEN_INVALID -1
27 #define TOKEN_STRING 256
28 #define TOKEN_INTEGER 257
29 #define TOKEN_REAL 258
30 #define TOKEN_TRUE 259
31 #define TOKEN_FALSE 260
32 #define TOKEN_NULL 261
34 /* Locale independent versions of isxxx() functions */
35 #define l_isupper(c) ('A' <= c && c <= 'Z')
36 #define l_islower(c) ('a' <= c && c <= 'z')
37 #define l_isalpha(c) (l_isupper(c) || l_islower(c))
38 #define l_isdigit(c) ('0' <= c && c <= '9')
39 #define l_isxdigit(c) \
40 (l_isdigit(c) || 'A' <= c || c <= 'F' || 'a' <= c || c <= 'f')
42 /* Read one byte from stream, convert to unsigned char, then int, and
43 return. return EOF on end of file. This corresponds to the
44 behaviour of fgetc(). */
45 typedef int (*get_func)(void *data);
54 int column, last_column;
60 strbuffer_t saved_text;
69 #define stream_to_lex(stream) container_of(stream, lex_t, stream)
72 /*** error reporting ***/
74 static void error_set(json_error_t *error, const lex_t *lex,
78 char msg_text[JSON_ERROR_TEXT_LENGTH];
79 char msg_with_context[JSON_ERROR_TEXT_LENGTH];
81 int line = -1, col = -1;
83 const char *result = msg_text;
89 vsnprintf(msg_text, JSON_ERROR_TEXT_LENGTH, msg, ap);
94 const char *saved_text = strbuffer_value(&lex->saved_text);
96 line = lex->stream.line;
97 col = lex->stream.column;
98 pos = lex->stream.position;
100 if(saved_text && saved_text[0])
102 if(lex->saved_text.length <= 20) {
103 snprintf(msg_with_context, JSON_ERROR_TEXT_LENGTH,
104 "%s near '%s'", msg_text, saved_text);
105 result = msg_with_context;
110 if(lex->stream.state == STREAM_STATE_ERROR) {
111 /* No context for UTF-8 decoding errors */
115 snprintf(msg_with_context, JSON_ERROR_TEXT_LENGTH,
116 "%s near end of file", msg_text);
117 result = msg_with_context;
122 jsonp_error_set(error, line, col, pos, "%s", result);
126 /*** lexical analyzer ***/
129 stream_init(stream_t *stream, get_func get, void *data)
133 stream->buffer[0] = '\0';
134 stream->buffer_pos = 0;
136 stream->state = STREAM_STATE_OK;
139 stream->position = 0;
142 static int stream_get(stream_t *stream, json_error_t *error)
146 if(stream->state != STREAM_STATE_OK)
147 return stream->state;
149 if(!stream->buffer[stream->buffer_pos])
151 c = stream->get(stream->data);
153 stream->state = STREAM_STATE_EOF;
154 return STREAM_STATE_EOF;
157 stream->buffer[0] = c;
158 stream->buffer_pos = 0;
160 if(0x80 <= c && c <= 0xFF)
162 /* multi-byte UTF-8 sequence */
165 count = utf8_check_first(c);
171 for(i = 1; i < count; i++)
172 stream->buffer[i] = stream->get(stream->data);
174 if(!utf8_check_full(stream->buffer, count, NULL))
177 stream->buffer[count] = '\0';
180 stream->buffer[1] = '\0';
183 c = stream->buffer[stream->buffer_pos++];
188 stream->last_column = stream->column;
191 else if(utf8_check_first(c)) {
192 /* track the Unicode character column, so increment only if
193 this is the first character of a UTF-8 sequence */
200 stream->state = STREAM_STATE_ERROR;
201 error_set(error, stream_to_lex(stream), "unable to decode byte 0x%x", c);
202 return STREAM_STATE_ERROR;
205 static void stream_unget(stream_t *stream, int c)
207 if(c == STREAM_STATE_EOF || c == STREAM_STATE_ERROR)
213 stream->column = stream->last_column;
215 else if(utf8_check_first(c))
218 assert(stream->buffer_pos > 0);
219 stream->buffer_pos--;
220 assert(stream->buffer[stream->buffer_pos] == c);
224 static int lex_get(lex_t *lex, json_error_t *error)
226 return stream_get(&lex->stream, error);
229 static void lex_save(lex_t *lex, int c)
231 strbuffer_append_byte(&lex->saved_text, c);
234 static int lex_get_save(lex_t *lex, json_error_t *error)
236 int c = stream_get(&lex->stream, error);
237 if(c != STREAM_STATE_EOF && c != STREAM_STATE_ERROR)
242 static void lex_unget(lex_t *lex, int c)
244 stream_unget(&lex->stream, c);
247 static void lex_unget_unsave(lex_t *lex, int c)
249 if(c != STREAM_STATE_EOF && c != STREAM_STATE_ERROR) {
251 stream_unget(&lex->stream, c);
252 d = strbuffer_pop(&lex->saved_text);
257 static void lex_save_cached(lex_t *lex)
259 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
261 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
262 lex->stream.buffer_pos++;
263 lex->stream.position++;
267 /* assumes that str points to 'u' plus at least 4 valid hex digits */
268 static int32_t decode_unicode_escape(const char *str)
273 assert(str[0] == 'u');
275 for(i = 1; i <= 4; i++) {
280 else if(l_islower(c))
281 value += c - 'a' + 10;
282 else if(l_isupper(c))
283 value += c - 'A' + 10;
291 static void lex_scan_string(lex_t *lex, json_error_t *error)
298 lex->value.string = NULL;
299 lex->token = TOKEN_INVALID;
301 c = lex_get_save(lex, error);
304 if(c == STREAM_STATE_ERROR)
307 else if(c == STREAM_STATE_EOF) {
308 error_set(error, lex, "premature end of input");
312 else if(0 <= c && c <= 0x1F) {
313 /* control character */
314 lex_unget_unsave(lex, c);
316 error_set(error, lex, "unexpected newline", c);
318 error_set(error, lex, "control character 0x%x", c);
323 c = lex_get_save(lex, error);
325 c = lex_get_save(lex, error);
326 for(i = 0; i < 4; i++) {
328 error_set(error, lex, "invalid escape");
331 c = lex_get_save(lex, error);
334 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
335 c == 'f' || c == 'n' || c == 'r' || c == 't')
336 c = lex_get_save(lex, error);
338 error_set(error, lex, "invalid escape");
343 c = lex_get_save(lex, error);
346 /* the actual value is at most of the same length as the source
348 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
349 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
350 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
351 are converted to 4 bytes
353 lex->value.string = jsonp_malloc(lex->saved_text.length + 1);
354 if(!lex->value.string) {
355 /* this is not very nice, since TOKEN_INVALID is returned */
360 t = lex->value.string;
362 /* + 1 to skip the " */
363 p = strbuffer_value(&lex->saved_text) + 1;
373 value = decode_unicode_escape(p);
376 if(0xD800 <= value && value <= 0xDBFF) {
378 if(*p == '\\' && *(p + 1) == 'u') {
379 int32_t value2 = decode_unicode_escape(++p);
382 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
383 /* valid second surrogate */
385 ((value - 0xD800) << 10) +
390 /* invalid second surrogate */
391 error_set(error, lex,
392 "invalid Unicode '\\u%04X\\u%04X'",
398 /* no second surrogate */
399 error_set(error, lex, "invalid Unicode '\\u%04X'",
404 else if(0xDC00 <= value && value <= 0xDFFF) {
405 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
410 error_set(error, lex, "\\u0000 is not allowed");
414 if(utf8_encode(value, buffer, &length))
417 memcpy(t, buffer, length);
422 case '"': case '\\': case '/':
424 case 'b': *t = '\b'; break;
425 case 'f': *t = '\f'; break;
426 case 'n': *t = '\n'; break;
427 case 'r': *t = '\r'; break;
428 case 't': *t = '\t'; break;
439 lex->token = TOKEN_STRING;
443 jsonp_free(lex->value.string);
446 #if JSON_INTEGER_IS_LONG_LONG
447 #define json_strtoint strtoll
449 #define json_strtoint strtol
452 static int lex_scan_number(lex_t *lex, int c, json_error_t *error)
454 const char *saved_text;
458 lex->token = TOKEN_INVALID;
461 c = lex_get_save(lex, error);
464 c = lex_get_save(lex, error);
466 lex_unget_unsave(lex, c);
470 else if(l_isdigit(c)) {
471 c = lex_get_save(lex, error);
473 c = lex_get_save(lex, error);
476 lex_unget_unsave(lex, c);
480 if(c != '.' && c != 'E' && c != 'e') {
483 lex_unget_unsave(lex, c);
485 saved_text = strbuffer_value(&lex->saved_text);
488 value = json_strtoint(saved_text, &end, 10);
489 if(errno == ERANGE) {
491 error_set(error, lex, "too big negative integer");
493 error_set(error, lex, "too big integer");
497 assert(end == saved_text + lex->saved_text.length);
499 lex->token = TOKEN_INTEGER;
500 lex->value.integer = value;
505 c = lex_get(lex, error);
512 c = lex_get_save(lex, error);
514 c = lex_get_save(lex, error);
517 if(c == 'E' || c == 'e') {
518 c = lex_get_save(lex, error);
519 if(c == '+' || c == '-')
520 c = lex_get_save(lex, error);
523 lex_unget_unsave(lex, c);
527 c = lex_get_save(lex, error);
529 c = lex_get_save(lex, error);
532 lex_unget_unsave(lex, c);
534 if(jsonp_strtod(&lex->saved_text, &value)) {
535 error_set(error, lex, "real number overflow");
539 lex->token = TOKEN_REAL;
540 lex->value.real = value;
547 static int lex_scan(lex_t *lex, json_error_t *error)
551 strbuffer_clear(&lex->saved_text);
553 if(lex->token == TOKEN_STRING) {
554 jsonp_free(lex->value.string);
555 lex->value.string = NULL;
558 c = lex_get(lex, error);
559 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
560 c = lex_get(lex, error);
562 if(c == STREAM_STATE_EOF) {
563 lex->token = TOKEN_EOF;
567 if(c == STREAM_STATE_ERROR) {
568 lex->token = TOKEN_INVALID;
574 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
578 lex_scan_string(lex, error);
580 else if(l_isdigit(c) || c == '-') {
581 if(lex_scan_number(lex, c, error))
585 else if(l_isalpha(c)) {
586 /* eat up the whole identifier for clearer error messages */
587 const char *saved_text;
589 c = lex_get_save(lex, error);
591 c = lex_get_save(lex, error);
592 lex_unget_unsave(lex, c);
594 saved_text = strbuffer_value(&lex->saved_text);
596 if(strcmp(saved_text, "true") == 0)
597 lex->token = TOKEN_TRUE;
598 else if(strcmp(saved_text, "false") == 0)
599 lex->token = TOKEN_FALSE;
600 else if(strcmp(saved_text, "null") == 0)
601 lex->token = TOKEN_NULL;
603 lex->token = TOKEN_INVALID;
607 /* save the rest of the input UTF-8 sequence to get an error
608 message of valid UTF-8 */
609 lex_save_cached(lex);
610 lex->token = TOKEN_INVALID;
617 static char *lex_steal_string(lex_t *lex)
620 if(lex->token == TOKEN_STRING)
622 result = lex->value.string;
623 lex->value.string = NULL;
628 static int lex_init(lex_t *lex, get_func get, void *data)
630 stream_init(&lex->stream, get, data);
631 if(strbuffer_init(&lex->saved_text))
634 lex->token = TOKEN_INVALID;
638 static void lex_close(lex_t *lex)
640 if(lex->token == TOKEN_STRING)
641 jsonp_free(lex->value.string);
642 strbuffer_close(&lex->saved_text);
648 static json_t *parse_value(lex_t *lex, size_t flags, json_error_t *error);
650 static json_t *parse_object(lex_t *lex, size_t flags, json_error_t *error)
652 json_t *object = json_object();
656 lex_scan(lex, error);
657 if(lex->token == '}')
664 if(lex->token != TOKEN_STRING) {
665 error_set(error, lex, "string or '}' expected");
669 key = lex_steal_string(lex);
673 if(flags & JSON_REJECT_DUPLICATES) {
674 if(json_object_get(object, key)) {
676 error_set(error, lex, "duplicate object key");
681 lex_scan(lex, error);
682 if(lex->token != ':') {
684 error_set(error, lex, "':' expected");
688 lex_scan(lex, error);
689 value = parse_value(lex, flags, error);
695 if(json_object_set_nocheck(object, key, value)) {
704 lex_scan(lex, error);
705 if(lex->token != ',')
708 lex_scan(lex, error);
711 if(lex->token != '}') {
712 error_set(error, lex, "'}' expected");
723 static json_t *parse_array(lex_t *lex, size_t flags, json_error_t *error)
725 json_t *array = json_array();
729 lex_scan(lex, error);
730 if(lex->token == ']')
734 json_t *elem = parse_value(lex, flags, error);
738 if(json_array_append(array, elem)) {
744 lex_scan(lex, error);
745 if(lex->token != ',')
748 lex_scan(lex, error);
751 if(lex->token != ']') {
752 error_set(error, lex, "']' expected");
763 static json_t *parse_value(lex_t *lex, size_t flags, json_error_t *error)
769 json = json_string_nocheck(lex->value.string);
773 case TOKEN_INTEGER: {
774 json = json_integer(lex->value.integer);
779 json = json_real(lex->value.real);
796 json = parse_object(lex, flags, error);
800 json = parse_array(lex, flags, error);
804 error_set(error, lex, "invalid token");
808 error_set(error, lex, "unexpected token");
818 static json_t *parse_json(lex_t *lex, size_t flags, json_error_t *error)
822 lex_scan(lex, error);
823 if(!(flags & JSON_DECODE_ANY)) {
824 if(lex->token != '[' && lex->token != '{') {
825 error_set(error, lex, "'[' or '{' expected");
830 result = parse_value(lex, flags, error);
834 if(!(flags & JSON_DISABLE_EOF_CHECK)) {
835 lex_scan(lex, error);
836 if(lex->token != TOKEN_EOF) {
837 error_set(error, lex, "end of file expected");
844 /* Save the position even though there was no error */
845 error->position = lex->stream.position;
857 static int string_get(void *data)
860 string_data_t *stream = (string_data_t *)data;
861 c = stream->data[stream->pos];
867 return (unsigned char)c;
871 json_t *json_loads(const char *string, size_t flags, json_error_t *error)
875 string_data_t stream_data;
877 jsonp_error_init(error, "<string>");
879 if (string == NULL) {
880 error_set(error, NULL, "wrong arguments");
884 stream_data.data = string;
887 if(lex_init(&lex, string_get, (void *)&stream_data))
890 result = parse_json(&lex, flags, error);
903 static int buffer_get(void *data)
906 buffer_data_t *stream = data;
907 if(stream->pos >= stream->len)
910 c = stream->data[stream->pos];
912 return (unsigned char)c;
915 json_t *json_loadb(const char *buffer, size_t buflen, size_t flags, json_error_t *error)
919 buffer_data_t stream_data;
921 jsonp_error_init(error, "<buffer>");
923 if (buffer == NULL) {
924 error_set(error, NULL, "wrong arguments");
928 stream_data.data = buffer;
930 stream_data.len = buflen;
932 if(lex_init(&lex, buffer_get, (void *)&stream_data))
935 result = parse_json(&lex, flags, error);
941 json_t *json_loadf(FILE *input, size_t flags, json_error_t *error)
952 jsonp_error_init(error, source);
955 error_set(error, NULL, "wrong arguments");
959 if(lex_init(&lex, (get_func)fgetc, input))
962 result = parse_json(&lex, flags, error);
968 json_t *json_load_file(const char *path, size_t flags, json_error_t *error)
973 jsonp_error_init(error, path);
976 error_set(error, NULL, "wrong arguments");
980 fp = fopen(path, "rb");
983 error_set(error, NULL, "unable to open %s: %s",
984 path, strerror(errno));
988 result = json_loadf(fp, flags, error);