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.
18 #include "jansson_private.h"
19 #include "strbuffer.h"
22 #define TOKEN_INVALID -1
24 #define TOKEN_STRING 256
25 #define TOKEN_INTEGER 257
26 #define TOKEN_REAL 258
27 #define TOKEN_TRUE 259
28 #define TOKEN_FALSE 260
29 #define TOKEN_NULL 261
31 /* read one byte from stream, return EOF on end of file */
32 typedef int (*get_func)(void *data);
34 /* return non-zero if end of file has been reached */
35 typedef int (*eof_func)(void *data);
49 strbuffer_t saved_text;
60 /*** error reporting ***/
62 static void error_init(json_error_t *error)
66 error->text[0] = '\0';
71 static void error_set(json_error_t *error, const lex_t *lex,
75 char text[JSON_ERROR_TEXT_LENGTH];
77 if(!error || error->text[0] != '\0') {
78 /* error already set */
83 vsnprintf(text, JSON_ERROR_TEXT_LENGTH, msg, ap);
88 const char *saved_text = strbuffer_value(&lex->saved_text);
89 error->line = lex->line;
90 if(saved_text && saved_text[0])
92 if(lex->saved_text.length <= 20) {
93 snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
94 "%s near '%s'", text, saved_text);
97 snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
101 snprintf(error->text, JSON_ERROR_TEXT_LENGTH,
102 "%s near end of file", text);
108 snprintf(error->text, JSON_ERROR_TEXT_LENGTH, "%s", text);
113 /*** lexical analyzer ***/
115 void stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
120 stream->stream_pos = 0;
121 stream->buffer[0] = '\0';
122 stream->buffer_pos = 0;
125 static char stream_get(stream_t *stream, json_error_t *error)
129 if(!stream->buffer[stream->buffer_pos])
131 stream->buffer[0] = stream->get(stream->data);
132 stream->buffer_pos = 0;
134 c = stream->buffer[0];
136 if(c < 0 && c != EOF)
138 /* multi-byte UTF-8 sequence */
141 count = utf8_check_first(c);
147 for(i = 1; i < count; i++)
148 stream->buffer[i] = stream->get(stream->data);
150 if(!utf8_check_full(stream->buffer, count))
153 stream->stream_pos += count;
154 stream->buffer[count] = '\0';
157 stream->buffer[1] = '\0';
158 stream->stream_pos++;
162 return stream->buffer[stream->buffer_pos++];
165 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
166 (unsigned char)c, stream->stream_pos);
168 stream->buffer[0] = EOF;
169 stream->buffer[1] = '\0';
170 stream->buffer_pos = 1;
175 static void stream_unget(stream_t *stream, char c)
177 assert(stream->buffer_pos > 0);
178 stream->buffer_pos--;
179 assert(stream->buffer[stream->buffer_pos] == c);
183 static int lex_get(lex_t *lex, json_error_t *error)
185 return stream_get(&lex->stream, error);
188 static int lex_eof(lex_t *lex)
190 return lex->stream.eof(lex->stream.data);
193 static void lex_save(lex_t *lex, char c)
195 strbuffer_append_byte(&lex->saved_text, c);
198 static int lex_get_save(lex_t *lex, json_error_t *error)
200 char c = stream_get(&lex->stream, error);
205 static void lex_unget_unsave(lex_t *lex, char c)
208 stream_unget(&lex->stream, c);
209 d = strbuffer_pop(&lex->saved_text);
213 static void lex_save_cached(lex_t *lex)
215 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
217 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
218 lex->stream.buffer_pos++;
222 /* assumes that str points to 'u' plus at least 4 valid hex digits */
223 static int decode_unicode_escape(const char *str)
228 assert(str[0] == 'u');
230 for(i = 1; i <= 4; i++) {
236 value += c - 'a' + 10;
238 value += c - 'A' + 10;
246 static void lex_scan_string(lex_t *lex, json_error_t *error)
253 lex->value.string = NULL;
254 lex->token = TOKEN_INVALID;
256 c = lex_get_save(lex, error);
260 lex_unget_unsave(lex, c);
262 error_set(error, lex, "premature end of input");
266 else if(0 <= c && c <= 0x1F) {
267 /* control character */
268 lex_unget_unsave(lex, c);
270 error_set(error, lex, "unexpected newline", c);
272 error_set(error, lex, "control character 0x%x", c);
277 c = lex_get_save(lex, error);
279 c = lex_get_save(lex, error);
280 for(i = 0; i < 4; i++) {
282 lex_unget_unsave(lex, c);
283 error_set(error, lex, "invalid escape");
286 c = lex_get_save(lex, error);
289 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
290 c == 'f' || c == 'n' || c == 'r' || c == 't')
291 c = lex_get_save(lex, error);
293 lex_unget_unsave(lex, c);
294 error_set(error, lex, "invalid escape");
299 c = lex_get_save(lex, error);
302 /* the actual value is at most of the same length as the source
304 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
305 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
306 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
307 are converted to 4 bytes
309 lex->value.string = malloc(lex->saved_text.length + 1);
310 if(!lex->value.string) {
311 /* this is not very nice, since TOKEN_INVALID is returned */
316 t = lex->value.string;
318 /* + 1 to skip the " */
319 p = strbuffer_value(&lex->saved_text) + 1;
329 value = decode_unicode_escape(p);
332 if(0xD800 <= value && value <= 0xDBFF) {
334 if(*p == '\\' && *(p + 1) == 'u') {
335 int value2 = decode_unicode_escape(++p);
338 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
339 /* valid second surrogate */
340 value = ((value - 0xD800) << 10) +
345 /* invalid second surrogate */
346 error_set(error, lex,
347 "invalid Unicode '\\u%04X\\u%04X'",
353 /* no second surrogate */
354 error_set(error, lex, "invalid Unicode '\\u%04X'",
359 else if(0xDC00 <= value && value <= 0xDFFF) {
360 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
365 error_set(error, lex, "\\u0000 is not allowed");
369 if(utf8_encode(value, buffer, &length))
372 memcpy(t, buffer, length);
377 case '"': case '\\': case '/':
379 case 'b': *t = '\b'; break;
380 case 'f': *t = '\f'; break;
381 case 'n': *t = '\n'; break;
382 case 'r': *t = '\r'; break;
383 case 't': *t = '\t'; break;
394 lex->token = TOKEN_STRING;
398 free(lex->value.string);
401 static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
403 const char *saved_text;
406 lex->token = TOKEN_INVALID;
409 c = lex_get_save(lex, error);
412 c = lex_get_save(lex, error);
414 lex_unget_unsave(lex, c);
418 else /* c != '0' */ {
419 c = lex_get_save(lex, error);
421 c = lex_get_save(lex, error);
424 if(c != '.' && c != 'E' && c != 'e') {
425 lex_unget_unsave(lex, c);
426 lex->token = TOKEN_INTEGER;
428 saved_text = strbuffer_value(&lex->saved_text);
429 lex->value.integer = strtol(saved_text, &end, 10);
430 assert(end == saved_text + lex->saved_text.length);
436 c = lex_get(lex, error);
441 c = lex_get_save(lex, error);
443 c = lex_get_save(lex, error);
446 if(c == 'E' || c == 'e') {
447 c = lex_get_save(lex, error);
448 if(c == '+' || c == '-')
449 c = lex_get_save(lex, error);
452 lex_unget_unsave(lex, c);
456 c = lex_get_save(lex, error);
458 c = lex_get_save(lex, error);
461 lex_unget_unsave(lex, c);
462 lex->token = TOKEN_REAL;
464 saved_text = strbuffer_value(&lex->saved_text);
465 lex->value.real = strtod(saved_text, &end);
466 assert(end == saved_text + lex->saved_text.length);
472 static int lex_scan(lex_t *lex, json_error_t *error)
476 strbuffer_clear(&lex->saved_text);
478 if(lex->token == TOKEN_STRING) {
479 free(lex->value.string);
480 lex->value.string = NULL;
483 c = lex_get(lex, error);
484 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
489 c = lex_get(lex, error);
494 lex->token = TOKEN_EOF;
496 lex->token = TOKEN_INVALID;
502 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
506 lex_scan_string(lex, error);
508 else if(isdigit(c) || c == '-')
509 lex_scan_number(lex, c, error);
511 else if(isupper(c) || islower(c)) {
512 /* eat up the whole identifier for clearer error messages */
513 const char *saved_text;
515 c = lex_get_save(lex, error);
516 while(isupper(c) || islower(c))
517 c = lex_get_save(lex, error);
518 lex_unget_unsave(lex, c);
520 saved_text = strbuffer_value(&lex->saved_text);
522 if(strcmp(saved_text, "true") == 0)
523 lex->token = TOKEN_TRUE;
524 else if(strcmp(saved_text, "false") == 0)
525 lex->token = TOKEN_FALSE;
526 else if(strcmp(saved_text, "null") == 0)
527 lex->token = TOKEN_NULL;
529 lex->token = TOKEN_INVALID;
533 /* save the rest of the input UTF-8 sequence to get an error
534 message of valid UTF-8 */
535 lex_save_cached(lex);
536 lex->token = TOKEN_INVALID;
543 static char *lex_steal_string(lex_t *lex)
546 if(lex->token == TOKEN_STRING)
548 result = lex->value.string;
549 lex->value.string = NULL;
554 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
556 stream_init(&lex->stream, get, eof, data);
557 if(strbuffer_init(&lex->saved_text))
560 lex->token = TOKEN_INVALID;
566 static void lex_close(lex_t *lex)
568 if(lex->token == TOKEN_STRING)
569 free(lex->value.string);
570 strbuffer_close(&lex->saved_text);
576 static json_t *parse_value(lex_t *lex, json_error_t *error);
578 static json_t *parse_object(lex_t *lex, json_error_t *error)
580 json_t *object = json_object();
584 lex_scan(lex, error);
585 if(lex->token == '}')
592 if(lex->token != TOKEN_STRING) {
593 error_set(error, lex, "string or '}' expected");
597 key = lex_steal_string(lex);
601 lex_scan(lex, error);
602 if(lex->token != ':') {
604 error_set(error, lex, "':' expected");
608 lex_scan(lex, error);
609 value = parse_value(lex, error);
615 if(json_object_set_nocheck(object, key, value)) {
624 lex_scan(lex, error);
625 if(lex->token != ',')
628 lex_scan(lex, error);
631 if(lex->token != '}') {
632 error_set(error, lex, "'}' expected");
643 static json_t *parse_array(lex_t *lex, json_error_t *error)
645 json_t *array = json_array();
649 lex_scan(lex, error);
650 if(lex->token == ']')
654 json_t *elem = parse_value(lex, error);
658 if(json_array_append(array, elem)) {
664 lex_scan(lex, error);
665 if(lex->token != ',')
668 lex_scan(lex, error);
671 if(lex->token != ']') {
672 error_set(error, lex, "']' expected");
683 static json_t *parse_value(lex_t *lex, json_error_t *error)
689 json = json_string_nocheck(lex->value.string);
693 case TOKEN_INTEGER: {
694 json = json_integer(lex->value.integer);
699 json = json_real(lex->value.real);
716 json = parse_object(lex, error);
720 json = parse_array(lex, error);
724 error_set(error, lex, "invalid token");
728 error_set(error, lex, "unexpected token");
738 json_t *parse_json(lex_t *lex, json_error_t *error)
742 lex_scan(lex, error);
743 if(lex->token != '[' && lex->token != '{') {
744 error_set(error, lex, "'[' or '{' expected");
748 return parse_value(lex, error);
757 static int string_get(void *data)
760 string_data_t *stream = (string_data_t *)data;
761 c = stream->data[stream->pos];
771 static int string_eof(void *data)
773 string_data_t *stream = (string_data_t *)data;
774 return (stream->data[stream->pos] == '\0');
777 json_t *json_loads(const char *string, json_error_t *error)
782 string_data_t stream_data = {
787 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
790 result = parse_json(&lex, error);
794 lex_scan(&lex, error);
795 if(lex.token != TOKEN_EOF) {
796 error_set(error, &lex, "end of file expected");
806 json_t *json_loadf(FILE *input, json_error_t *error)
811 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
814 result = parse_json(&lex, error);
818 lex_scan(&lex, error);
819 if(lex.token != TOKEN_EOF) {
820 error_set(error, &lex, "end of file expected");
830 json_t *json_load_file(const char *path, json_error_t *error)
835 fp = fopen(path, "r");
838 error_set(error, NULL, "unable to open %s: %s",
839 path, strerror(errno));
843 result = json_loadf(fp, error);