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 == EOF && stream->eof(stream->data))
141 /* multi-byte UTF-8 sequence */
144 count = utf8_check_first(c);
150 for(i = 1; i < count; i++)
151 stream->buffer[i] = stream->get(stream->data);
153 if(!utf8_check_full(stream->buffer, count))
156 stream->stream_pos += count;
157 stream->buffer[count] = '\0';
160 stream->buffer[1] = '\0';
161 stream->stream_pos++;
165 return stream->buffer[stream->buffer_pos++];
168 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
169 (unsigned char)c, stream->stream_pos);
171 stream->buffer[0] = EOF;
172 stream->buffer[1] = '\0';
173 stream->buffer_pos = 1;
178 static void stream_unget(stream_t *stream, char c)
180 assert(stream->buffer_pos > 0);
181 stream->buffer_pos--;
182 assert(stream->buffer[stream->buffer_pos] == c);
186 static int lex_get(lex_t *lex, json_error_t *error)
188 return stream_get(&lex->stream, error);
191 static int lex_eof(lex_t *lex)
193 return lex->stream.eof(lex->stream.data);
196 static void lex_save(lex_t *lex, char c)
198 strbuffer_append_byte(&lex->saved_text, c);
201 static int lex_get_save(lex_t *lex, json_error_t *error)
203 char c = stream_get(&lex->stream, error);
208 static void lex_unget_unsave(lex_t *lex, char c)
211 stream_unget(&lex->stream, c);
212 d = strbuffer_pop(&lex->saved_text);
216 static void lex_save_cached(lex_t *lex)
218 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
220 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
221 lex->stream.buffer_pos++;
225 /* assumes that str points to 'u' plus at least 4 valid hex digits */
226 static int decode_unicode_escape(const char *str)
231 assert(str[0] == 'u');
233 for(i = 1; i <= 4; i++) {
239 value += c - 'a' + 10;
241 value += c - 'A' + 10;
249 static void lex_scan_string(lex_t *lex, json_error_t *error)
256 lex->value.string = NULL;
257 lex->token = TOKEN_INVALID;
260 c = lex_get_save(lex, error);
265 error_set(error, lex, "premature end of input");
269 else if(0 <= c && c <= 0x1F) {
270 /* control character */
271 lex_unget_unsave(lex, c);
273 error_set(error, lex, "unexpected newline", c);
275 error_set(error, lex, "control character 0x%x", c);
280 c = lex_get_save(lex, error);
282 c = lex_get_save(lex, error);
283 for(i = 0; i < 4; i++) {
285 lex_unget_unsave(lex, c);
286 error_set(error, lex, "invalid escape");
289 c = lex_get_save(lex, error);
292 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
293 c == 'f' || c == 'n' || c == 'r' || c == 't')
294 c = lex_get_save(lex, error);
296 lex_unget_unsave(lex, c);
297 error_set(error, lex, "invalid escape");
302 c = lex_get_save(lex, error);
305 /* the actual value is at most of the same length as the source
307 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
308 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
309 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
310 are converted to 4 bytes
312 lex->value.string = malloc(lex->saved_text.length + 1);
313 if(!lex->value.string) {
314 /* this is not very nice, since TOKEN_INVALID is returned */
319 t = lex->value.string;
321 /* + 1 to skip the " */
322 p = strbuffer_value(&lex->saved_text) + 1;
332 value = decode_unicode_escape(p);
335 if(0xD800 <= value && value <= 0xDBFF) {
337 if(*p == '\\' && *(p + 1) == 'u') {
338 int value2 = decode_unicode_escape(++p);
341 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
342 /* valid second surrogate */
343 value = ((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 void lex_scan_number(lex_t *lex, char c, json_error_t *error)
406 const char *saved_text;
409 lex->token = TOKEN_INVALID;
412 c = lex_get_save(lex, error);
415 c = lex_get_save(lex, error);
417 lex_unget_unsave(lex, c);
421 else /* c != '0' */ {
422 c = lex_get_save(lex, error);
424 c = lex_get_save(lex, error);
427 if(c != '.' && c != 'E' && c != 'e') {
428 lex_unget_unsave(lex, c);
429 lex->token = TOKEN_INTEGER;
431 saved_text = strbuffer_value(&lex->saved_text);
432 lex->value.integer = strtol(saved_text, &end, 10);
433 assert(end == saved_text + lex->saved_text.length);
439 c = lex_get(lex, error);
444 c = lex_get_save(lex, error);
446 c = lex_get_save(lex, error);
449 if(c == 'E' || c == 'e') {
450 c = lex_get_save(lex, error);
451 if(c == '+' || c == '-')
452 c = lex_get_save(lex, error);
455 lex_unget_unsave(lex, c);
459 c = lex_get_save(lex, error);
461 c = lex_get_save(lex, error);
464 lex_unget_unsave(lex, c);
465 lex->token = TOKEN_REAL;
467 saved_text = strbuffer_value(&lex->saved_text);
468 lex->value.real = strtod(saved_text, &end);
469 assert(end == saved_text + lex->saved_text.length);
475 static int lex_scan(lex_t *lex, json_error_t *error)
479 strbuffer_clear(&lex->saved_text);
481 if(lex->token == TOKEN_STRING) {
482 free(lex->value.string);
483 lex->value.string = NULL;
486 c = lex_get(lex, error);
487 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
492 c = lex_get(lex, error);
497 lex->token = TOKEN_EOF;
499 lex->token = TOKEN_INVALID;
505 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
509 lex_scan_string(lex, error);
511 else if(isdigit(c) || c == '-')
512 lex_scan_number(lex, c, error);
514 else if(isupper(c) || islower(c)) {
515 /* eat up the whole identifier for clearer error messages */
516 const char *saved_text;
518 c = lex_get_save(lex, error);
519 while(isupper(c) || islower(c))
520 c = lex_get_save(lex, error);
521 lex_unget_unsave(lex, c);
523 saved_text = strbuffer_value(&lex->saved_text);
525 if(strcmp(saved_text, "true") == 0)
526 lex->token = TOKEN_TRUE;
527 else if(strcmp(saved_text, "false") == 0)
528 lex->token = TOKEN_FALSE;
529 else if(strcmp(saved_text, "null") == 0)
530 lex->token = TOKEN_NULL;
532 lex->token = TOKEN_INVALID;
536 /* save the rest of the input UTF-8 sequence to get an error
537 message of valid UTF-8 */
538 lex_save_cached(lex);
539 lex->token = TOKEN_INVALID;
546 static char *lex_steal_string(lex_t *lex)
549 if(lex->token == TOKEN_STRING)
551 result = lex->value.string;
552 lex->value.string = NULL;
557 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
559 stream_init(&lex->stream, get, eof, data);
560 if(strbuffer_init(&lex->saved_text))
563 lex->token = TOKEN_INVALID;
569 static void lex_close(lex_t *lex)
571 if(lex->token == TOKEN_STRING)
572 free(lex->value.string);
573 strbuffer_close(&lex->saved_text);
579 static json_t *parse_value(lex_t *lex, json_error_t *error);
581 static json_t *parse_object(lex_t *lex, json_error_t *error)
583 json_t *object = json_object();
587 lex_scan(lex, error);
588 if(lex->token == '}')
595 if(lex->token != TOKEN_STRING) {
596 error_set(error, lex, "string or '}' expected");
600 key = lex_steal_string(lex);
604 lex_scan(lex, error);
605 if(lex->token != ':') {
607 error_set(error, lex, "':' expected");
611 lex_scan(lex, error);
612 value = parse_value(lex, error);
618 if(json_object_set_nocheck(object, key, value)) {
627 lex_scan(lex, error);
628 if(lex->token != ',')
631 lex_scan(lex, error);
634 if(lex->token != '}') {
635 error_set(error, lex, "'}' expected");
646 static json_t *parse_array(lex_t *lex, json_error_t *error)
648 json_t *array = json_array();
652 lex_scan(lex, error);
653 if(lex->token == ']')
657 json_t *elem = parse_value(lex, error);
661 if(json_array_append(array, elem)) {
667 lex_scan(lex, error);
668 if(lex->token != ',')
671 lex_scan(lex, error);
674 if(lex->token != ']') {
675 error_set(error, lex, "']' expected");
686 static json_t *parse_value(lex_t *lex, json_error_t *error)
692 json = json_string_nocheck(lex->value.string);
696 case TOKEN_INTEGER: {
697 json = json_integer(lex->value.integer);
702 json = json_real(lex->value.real);
719 json = parse_object(lex, error);
723 json = parse_array(lex, error);
727 error_set(error, lex, "invalid token");
731 error_set(error, lex, "unexpected token");
741 json_t *parse_json(lex_t *lex, json_error_t *error)
745 lex_scan(lex, error);
746 if(lex->token != '[' && lex->token != '{') {
747 error_set(error, lex, "'[' or '{' expected");
751 return parse_value(lex, error);
760 static int string_get(void *data)
763 string_data_t *stream = (string_data_t *)data;
764 c = stream->data[stream->pos];
774 static int string_eof(void *data)
776 string_data_t *stream = (string_data_t *)data;
777 return (stream->data[stream->pos] == '\0');
780 json_t *json_loads(const char *string, json_error_t *error)
785 string_data_t stream_data = {
790 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
793 result = parse_json(&lex, error);
797 lex_scan(&lex, error);
798 if(lex.token != TOKEN_EOF) {
799 error_set(error, &lex, "end of file expected");
809 json_t *json_loadf(FILE *input, json_error_t *error)
814 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
817 result = parse_json(&lex, error);
821 lex_scan(&lex, error);
822 if(lex.token != TOKEN_EOF) {
823 error_set(error, &lex, "end of file expected");
833 json_t *json_load_file(const char *path, json_error_t *error)
838 fp = fopen(path, "r");
841 error_set(error, NULL, "unable to open %s: %s",
842 path, strerror(errno));
846 result = json_loadf(fp, error);