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 #define JSON_ERROR_MSG_LENGTH 160
66 char msg[JSON_ERROR_MSG_LENGTH];
70 const char *json_error_msg(const json_error_t *error)
72 return error ? error->msg : NULL;
75 int json_error_line(const json_error_t *error)
77 return error ? error->line : -1;
80 static void error_init(json_error_t **error)
86 static void error_set(json_error_t **error, const lex_t *lex,
90 char text[JSON_ERROR_MSG_LENGTH];
92 if(!error || *error) {
93 /* error not given or already set */
97 *error = malloc(sizeof(json_error_t));
102 vsnprintf(text, JSON_ERROR_MSG_LENGTH, msg, ap);
107 const char *saved_text = strbuffer_value(&lex->saved_text);
108 (*error)->line = lex->line;
109 if(saved_text && saved_text[0])
111 if(lex->saved_text.length <= 20) {
112 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH,
113 "%s near '%s'", text, saved_text);
116 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH, "%s", text);
120 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH,
121 "%s near end of file", text);
127 snprintf((*error)->msg, JSON_ERROR_MSG_LENGTH, "%s", text);
132 /*** lexical analyzer ***/
135 stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
140 stream->stream_pos = 0;
141 stream->buffer[0] = '\0';
142 stream->buffer_pos = 0;
145 static char stream_get(stream_t *stream, json_error_t **error)
149 if(!stream->buffer[stream->buffer_pos])
151 stream->buffer[0] = stream->get(stream->data);
152 stream->buffer_pos = 0;
154 c = stream->buffer[0];
156 if((unsigned char)c >= 0x80 && c != (char)EOF)
158 /* multi-byte UTF-8 sequence */
161 count = utf8_check_first(c);
167 for(i = 1; i < count; i++)
168 stream->buffer[i] = stream->get(stream->data);
170 if(!utf8_check_full(stream->buffer, count, NULL))
173 stream->stream_pos += count;
174 stream->buffer[count] = '\0';
177 stream->buffer[1] = '\0';
178 stream->stream_pos++;
182 return stream->buffer[stream->buffer_pos++];
185 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
186 (unsigned char)c, stream->stream_pos);
188 stream->buffer[0] = EOF;
189 stream->buffer[1] = '\0';
190 stream->buffer_pos = 1;
195 static void stream_unget(stream_t *stream, char c)
197 assert(stream->buffer_pos > 0);
198 stream->buffer_pos--;
199 assert(stream->buffer[stream->buffer_pos] == c);
203 static int lex_get(lex_t *lex, json_error_t **error)
205 return stream_get(&lex->stream, error);
208 static int lex_eof(lex_t *lex)
210 return lex->stream.eof(lex->stream.data);
213 static void lex_save(lex_t *lex, char c)
215 strbuffer_append_byte(&lex->saved_text, c);
218 static int lex_get_save(lex_t *lex, json_error_t **error)
220 char c = stream_get(&lex->stream, error);
225 static void lex_unget_unsave(lex_t *lex, char c)
228 stream_unget(&lex->stream, c);
229 d = strbuffer_pop(&lex->saved_text);
233 static void lex_save_cached(lex_t *lex)
235 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
237 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
238 lex->stream.buffer_pos++;
242 /* assumes that str points to 'u' plus at least 4 valid hex digits */
243 static int32_t decode_unicode_escape(const char *str)
248 assert(str[0] == 'u');
250 for(i = 1; i <= 4; i++) {
256 value += c - 'a' + 10;
258 value += c - 'A' + 10;
266 static void lex_scan_string(lex_t *lex, json_error_t **error)
273 lex->value.string = NULL;
274 lex->token = TOKEN_INVALID;
276 c = lex_get_save(lex, error);
280 lex_unget_unsave(lex, c);
282 error_set(error, lex, "premature end of input");
286 else if((unsigned char)c <= 0x1F) {
287 /* control character */
288 lex_unget_unsave(lex, c);
290 error_set(error, lex, "unexpected newline", c);
292 error_set(error, lex, "control character 0x%x", c);
297 c = lex_get_save(lex, error);
299 c = lex_get_save(lex, error);
300 for(i = 0; i < 4; i++) {
302 lex_unget_unsave(lex, c);
303 error_set(error, lex, "invalid escape");
306 c = lex_get_save(lex, error);
309 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
310 c == 'f' || c == 'n' || c == 'r' || c == 't')
311 c = lex_get_save(lex, error);
313 lex_unget_unsave(lex, c);
314 error_set(error, lex, "invalid escape");
319 c = lex_get_save(lex, error);
322 /* the actual value is at most of the same length as the source
324 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
325 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
326 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
327 are converted to 4 bytes
329 lex->value.string = malloc(lex->saved_text.length + 1);
330 if(!lex->value.string) {
331 /* this is not very nice, since TOKEN_INVALID is returned */
336 t = lex->value.string;
338 /* + 1 to skip the " */
339 p = strbuffer_value(&lex->saved_text) + 1;
349 value = decode_unicode_escape(p);
352 if(0xD800 <= value && value <= 0xDBFF) {
354 if(*p == '\\' && *(p + 1) == 'u') {
355 int32_t value2 = decode_unicode_escape(++p);
358 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
359 /* valid second surrogate */
361 ((value - 0xD800) << 10) +
366 /* invalid second surrogate */
367 error_set(error, lex,
368 "invalid Unicode '\\u%04X\\u%04X'",
374 /* no second surrogate */
375 error_set(error, lex, "invalid Unicode '\\u%04X'",
380 else if(0xDC00 <= value && value <= 0xDFFF) {
381 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
386 error_set(error, lex, "\\u0000 is not allowed");
390 if(utf8_encode(value, buffer, &length))
393 memcpy(t, buffer, length);
398 case '"': case '\\': case '/':
400 case 'b': *t = '\b'; break;
401 case 'f': *t = '\f'; break;
402 case 'n': *t = '\n'; break;
403 case 'r': *t = '\r'; break;
404 case 't': *t = '\t'; break;
415 lex->token = TOKEN_STRING;
419 free(lex->value.string);
422 #if JSON_INTEGER_IS_LONG_LONG
423 #define json_strtoint strtoll
425 #define json_strtoint strtol
428 static int lex_scan_number(lex_t *lex, char c, json_error_t **error)
430 const char *saved_text;
434 lex->token = TOKEN_INVALID;
437 c = lex_get_save(lex, error);
440 c = lex_get_save(lex, error);
442 lex_unget_unsave(lex, c);
446 else if(isdigit(c)) {
447 c = lex_get_save(lex, error);
449 c = lex_get_save(lex, error);
452 lex_unget_unsave(lex, c);
456 if(c != '.' && c != 'E' && c != 'e') {
459 lex_unget_unsave(lex, c);
461 saved_text = strbuffer_value(&lex->saved_text);
464 value = json_strtoint(saved_text, &end, 10);
465 if(errno == ERANGE) {
467 error_set(error, lex, "too big negative integer");
469 error_set(error, lex, "too big integer");
473 assert(end == saved_text + lex->saved_text.length);
475 lex->token = TOKEN_INTEGER;
476 lex->value.integer = value;
481 c = lex_get(lex, error);
486 c = lex_get_save(lex, error);
488 c = lex_get_save(lex, error);
491 if(c == 'E' || c == 'e') {
492 c = lex_get_save(lex, error);
493 if(c == '+' || c == '-')
494 c = lex_get_save(lex, error);
497 lex_unget_unsave(lex, c);
501 c = lex_get_save(lex, error);
503 c = lex_get_save(lex, error);
506 lex_unget_unsave(lex, c);
508 saved_text = strbuffer_value(&lex->saved_text);
509 value = strtod(saved_text, &end);
510 assert(end == saved_text + lex->saved_text.length);
512 if(errno == ERANGE && value != 0) {
513 error_set(error, lex, "real number overflow");
517 lex->token = TOKEN_REAL;
518 lex->value.real = value;
525 static int lex_scan(lex_t *lex, json_error_t **error)
529 strbuffer_clear(&lex->saved_text);
531 if(lex->token == TOKEN_STRING) {
532 free(lex->value.string);
533 lex->value.string = NULL;
536 c = lex_get(lex, error);
537 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
542 c = lex_get(lex, error);
547 lex->token = TOKEN_EOF;
549 lex->token = TOKEN_INVALID;
555 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
559 lex_scan_string(lex, error);
561 else if(isdigit(c) || c == '-') {
562 if(lex_scan_number(lex, c, error))
566 else if(isupper(c) || islower(c)) {
567 /* eat up the whole identifier for clearer error messages */
568 const char *saved_text;
570 c = lex_get_save(lex, error);
571 while(isupper(c) || islower(c))
572 c = lex_get_save(lex, error);
573 lex_unget_unsave(lex, c);
575 saved_text = strbuffer_value(&lex->saved_text);
577 if(strcmp(saved_text, "true") == 0)
578 lex->token = TOKEN_TRUE;
579 else if(strcmp(saved_text, "false") == 0)
580 lex->token = TOKEN_FALSE;
581 else if(strcmp(saved_text, "null") == 0)
582 lex->token = TOKEN_NULL;
584 lex->token = TOKEN_INVALID;
588 /* save the rest of the input UTF-8 sequence to get an error
589 message of valid UTF-8 */
590 lex_save_cached(lex);
591 lex->token = TOKEN_INVALID;
598 static char *lex_steal_string(lex_t *lex)
601 if(lex->token == TOKEN_STRING)
603 result = lex->value.string;
604 lex->value.string = NULL;
609 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
611 stream_init(&lex->stream, get, eof, data);
612 if(strbuffer_init(&lex->saved_text))
615 lex->token = TOKEN_INVALID;
621 static void lex_close(lex_t *lex)
623 if(lex->token == TOKEN_STRING)
624 free(lex->value.string);
625 strbuffer_close(&lex->saved_text);
631 static json_t *parse_value(lex_t *lex, json_error_t **error);
633 static json_t *parse_object(lex_t *lex, json_error_t **error)
635 json_t *object = json_object();
639 lex_scan(lex, error);
640 if(lex->token == '}')
647 if(lex->token != TOKEN_STRING) {
648 error_set(error, lex, "string or '}' expected");
652 key = lex_steal_string(lex);
656 lex_scan(lex, error);
657 if(lex->token != ':') {
659 error_set(error, lex, "':' expected");
663 lex_scan(lex, error);
664 value = parse_value(lex, error);
670 if(json_object_set_nocheck(object, key, value)) {
679 lex_scan(lex, error);
680 if(lex->token != ',')
683 lex_scan(lex, error);
686 if(lex->token != '}') {
687 error_set(error, lex, "'}' expected");
698 static json_t *parse_array(lex_t *lex, json_error_t **error)
700 json_t *array = json_array();
704 lex_scan(lex, error);
705 if(lex->token == ']')
709 json_t *elem = parse_value(lex, error);
713 if(json_array_append(array, elem)) {
719 lex_scan(lex, error);
720 if(lex->token != ',')
723 lex_scan(lex, error);
726 if(lex->token != ']') {
727 error_set(error, lex, "']' expected");
738 static json_t *parse_value(lex_t *lex, json_error_t **error)
744 json = json_string_nocheck(lex->value.string);
748 case TOKEN_INTEGER: {
749 json = json_integer(lex->value.integer);
754 json = json_real(lex->value.real);
771 json = parse_object(lex, error);
775 json = parse_array(lex, error);
779 error_set(error, lex, "invalid token");
783 error_set(error, lex, "unexpected token");
793 static json_t *parse_json(lex_t *lex, json_error_t **error)
797 lex_scan(lex, error);
798 if(lex->token != '[' && lex->token != '{') {
799 error_set(error, lex, "'[' or '{' expected");
803 return parse_value(lex, error);
812 static int string_get(void *data)
815 string_data_t *stream = (string_data_t *)data;
816 c = stream->data[stream->pos];
826 static int string_eof(void *data)
828 string_data_t *stream = (string_data_t *)data;
829 return (stream->data[stream->pos] == '\0');
832 json_t *json_loads(const char *string, size_t flags, json_error_t **error)
836 (void)flags; /* unused */
838 string_data_t stream_data = {string, 0};
840 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
843 result = parse_json(&lex, error);
847 lex_scan(&lex, error);
848 if(lex.token != TOKEN_EOF) {
849 error_set(error, &lex, "end of file expected");
859 json_t *json_loadf(FILE *input, size_t flags, json_error_t **error)
863 (void)flags; /* unused */
865 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
868 result = parse_json(&lex, error);
872 lex_scan(&lex, error);
873 if(lex.token != TOKEN_EOF) {
874 error_set(error, &lex, "end of file expected");
884 json_t *json_load_file(const char *path, size_t flags, json_error_t **error)
891 fp = fopen(path, "r");
894 error_set(error, NULL, "unable to open %s: %s",
895 path, strerror(errno));
899 result = json_loadf(fp, flags, error);