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.
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 ***/
116 void stream_init(stream_t *stream, get_func get, eof_func eof, void *data)
121 stream->stream_pos = 0;
122 stream->buffer[0] = '\0';
123 stream->buffer_pos = 0;
126 static char stream_get(stream_t *stream, json_error_t *error)
130 if(!stream->buffer[stream->buffer_pos])
132 stream->buffer[0] = stream->get(stream->data);
133 stream->buffer_pos = 0;
135 c = stream->buffer[0];
137 if(c == EOF && stream->eof(stream->data))
142 /* multi-byte UTF-8 sequence */
145 count = utf8_check_first(c);
151 for(i = 1; i < count; i++)
152 stream->buffer[i] = stream->get(stream->data);
154 if(!utf8_check_full(stream->buffer, count))
157 stream->stream_pos += count;
158 stream->buffer[count] = '\0';
161 stream->buffer[1] = '\0';
162 stream->stream_pos++;
166 return stream->buffer[stream->buffer_pos++];
169 error_set(error, NULL, "unable to decode byte 0x%x at position %d",
170 (unsigned char)c, stream->stream_pos);
172 stream->buffer[0] = EOF;
173 stream->buffer[1] = '\0';
174 stream->buffer_pos = 1;
179 static void stream_unget(stream_t *stream, char c)
181 assert(stream->buffer_pos > 0);
182 stream->buffer_pos--;
183 assert(stream->buffer[stream->buffer_pos] == c);
187 static int lex_get(lex_t *lex, json_error_t *error)
189 return stream_get(&lex->stream, error);
192 static int lex_eof(lex_t *lex)
194 return lex->stream.eof(lex->stream.data);
197 static void lex_save(lex_t *lex, char c)
199 strbuffer_append_byte(&lex->saved_text, c);
202 static int lex_get_save(lex_t *lex, json_error_t *error)
204 char c = stream_get(&lex->stream, error);
209 static void lex_unget_unsave(lex_t *lex, char c)
212 stream_unget(&lex->stream, c);
213 d = strbuffer_pop(&lex->saved_text);
217 static void lex_save_cached(lex_t *lex)
219 while(lex->stream.buffer[lex->stream.buffer_pos] != '\0')
221 lex_save(lex, lex->stream.buffer[lex->stream.buffer_pos]);
222 lex->stream.buffer_pos++;
226 /* assumes that str points to 'u' plus at least 4 valid hex digits */
227 static int decode_unicode_escape(const char *str)
232 assert(str[0] == 'u');
234 for(i = 1; i <= 4; i++) {
240 value += c - 'a' + 10;
242 value += c - 'A' + 10;
250 static void lex_scan_string(lex_t *lex, json_error_t *error)
257 lex->value.string = NULL;
258 lex->token = TOKEN_INVALID;
261 c = lex_get_save(lex, error);
266 error_set(error, lex, "premature end of input");
270 else if(0 <= c && c <= 0x1F) {
271 /* control character */
272 lex_unget_unsave(lex, c);
274 error_set(error, lex, "unexpected newline", c);
276 error_set(error, lex, "control character 0x%x", c);
281 c = lex_get_save(lex, error);
283 c = lex_get_save(lex, error);
284 for(i = 0; i < 4; i++) {
286 lex_unget_unsave(lex, c);
287 error_set(error, lex, "invalid escape");
290 c = lex_get_save(lex, error);
293 else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
294 c == 'f' || c == 'n' || c == 'r' || c == 't')
295 c = lex_get_save(lex, error);
297 lex_unget_unsave(lex, c);
298 error_set(error, lex, "invalid escape");
303 c = lex_get_save(lex, error);
306 /* the actual value is at most of the same length as the source
308 - shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
309 - a single \uXXXX escape (length 6) is converted to at most 3 bytes
310 - two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
311 are converted to 4 bytes
313 lex->value.string = malloc(lex->saved_text.length + 1);
314 if(!lex->value.string) {
315 /* this is not very nice, since TOKEN_INVALID is returned */
320 t = lex->value.string;
322 /* + 1 to skip the " */
323 p = strbuffer_value(&lex->saved_text) + 1;
333 value = decode_unicode_escape(p);
336 if(0xD800 <= value && value <= 0xDBFF) {
338 if(*p == '\\' && *(p + 1) == 'u') {
339 int value2 = decode_unicode_escape(++p);
342 if(0xDC00 <= value2 && value2 <= 0xDFFF) {
343 /* valid second surrogate */
344 value = ((value - 0xD800) << 10) +
349 /* invalid second surrogate */
350 error_set(error, lex,
351 "invalid Unicode '\\u%04X\\u%04X'",
357 /* no second surrogate */
358 error_set(error, lex, "invalid Unicode '\\u%04X'",
363 else if(0xDC00 <= value && value <= 0xDFFF) {
364 error_set(error, lex, "invalid Unicode '\\u%04X'", value);
369 error_set(error, lex, "\\u0000 is not allowed");
373 if(utf8_encode(value, buffer, &length))
376 memcpy(t, buffer, length);
381 case '"': case '\\': case '/':
383 case 'b': *t = '\b'; break;
384 case 'f': *t = '\f'; break;
385 case 'n': *t = '\n'; break;
386 case 'r': *t = '\r'; break;
387 case 't': *t = '\t'; break;
398 lex->token = TOKEN_STRING;
402 free(lex->value.string);
405 static void lex_scan_number(lex_t *lex, char c, json_error_t *error)
407 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 /* c != '0' */ {
423 c = lex_get_save(lex, error);
425 c = lex_get_save(lex, error);
428 if(c != '.' && c != 'E' && c != 'e') {
429 lex_unget_unsave(lex, c);
430 lex->token = TOKEN_INTEGER;
432 saved_text = strbuffer_value(&lex->saved_text);
433 lex->value.integer = strtol(saved_text, &end, 10);
434 assert(end == saved_text + lex->saved_text.length);
440 c = lex_get(lex, error);
445 c = lex_get_save(lex, error);
447 c = lex_get_save(lex, error);
450 if(c == 'E' || c == 'e') {
451 c = lex_get_save(lex, error);
452 if(c == '+' || c == '-')
453 c = lex_get_save(lex, error);
456 lex_unget_unsave(lex, c);
460 c = lex_get_save(lex, error);
462 c = lex_get_save(lex, error);
465 lex_unget_unsave(lex, c);
466 lex->token = TOKEN_REAL;
468 saved_text = strbuffer_value(&lex->saved_text);
469 lex->value.real = strtod(saved_text, &end);
470 assert(end == saved_text + lex->saved_text.length);
476 static int lex_scan(lex_t *lex, json_error_t *error)
480 strbuffer_clear(&lex->saved_text);
482 if(lex->token == TOKEN_STRING) {
483 free(lex->value.string);
484 lex->value.string = NULL;
487 c = lex_get(lex, error);
488 while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
493 c = lex_get(lex, error);
498 lex->token = TOKEN_EOF;
500 lex->token = TOKEN_INVALID;
506 if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
510 lex_scan_string(lex, error);
512 else if(isdigit(c) || c == '-')
513 lex_scan_number(lex, c, error);
515 else if(isupper(c) || islower(c)) {
516 /* eat up the whole identifier for clearer error messages */
517 const char *saved_text;
519 c = lex_get_save(lex, error);
520 while(isupper(c) || islower(c))
521 c = lex_get_save(lex, error);
522 lex_unget_unsave(lex, c);
524 saved_text = strbuffer_value(&lex->saved_text);
526 if(strcmp(saved_text, "true") == 0)
527 lex->token = TOKEN_TRUE;
528 else if(strcmp(saved_text, "false") == 0)
529 lex->token = TOKEN_FALSE;
530 else if(strcmp(saved_text, "null") == 0)
531 lex->token = TOKEN_NULL;
533 lex->token = TOKEN_INVALID;
537 /* save the rest of the input UTF-8 sequence to get an error
538 message of valid UTF-8 */
539 lex_save_cached(lex);
540 lex->token = TOKEN_INVALID;
547 static int lex_init(lex_t *lex, get_func get, eof_func eof, void *data)
549 stream_init(&lex->stream, get, eof, data);
550 if(strbuffer_init(&lex->saved_text))
553 lex->token = TOKEN_INVALID;
559 static void lex_close(lex_t *lex)
561 if(lex->token == TOKEN_STRING)
562 free(lex->value.string);
563 strbuffer_close(&lex->saved_text);
569 static json_t *parse_value(lex_t *lex, json_error_t *error);
571 static json_t *parse_object(lex_t *lex, json_error_t *error)
573 json_t *object = json_object();
577 lex_scan(lex, error);
578 if(lex->token == '}')
585 if(lex->token != TOKEN_STRING) {
586 error_set(error, lex, "string or '}' expected");
590 key = strdup(lex->value.string);
594 lex_scan(lex, error);
595 if(lex->token != ':') {
597 error_set(error, lex, "':' expected");
601 lex_scan(lex, error);
602 value = parse_value(lex, error);
608 if(json_object_set_nocheck(object, key, value)) {
617 lex_scan(lex, error);
618 if(lex->token != ',')
621 lex_scan(lex, error);
624 if(lex->token != '}') {
625 error_set(error, lex, "'}' expected");
636 static json_t *parse_array(lex_t *lex, json_error_t *error)
638 json_t *array = json_array();
642 lex_scan(lex, error);
643 if(lex->token == ']')
647 json_t *elem = parse_value(lex, error);
651 if(json_array_append(array, elem)) {
657 lex_scan(lex, error);
658 if(lex->token != ',')
661 lex_scan(lex, error);
664 if(lex->token != ']') {
665 error_set(error, lex, "']' expected");
676 static json_t *parse_value(lex_t *lex, json_error_t *error)
682 json = json_string_nocheck(lex->value.string);
686 case TOKEN_INTEGER: {
687 json = json_integer(lex->value.integer);
692 json = json_real(lex->value.real);
709 json = parse_object(lex, error);
713 json = parse_array(lex, error);
717 error_set(error, lex, "invalid token");
721 error_set(error, lex, "unexpected token");
731 json_t *parse_json(lex_t *lex, json_error_t *error)
735 lex_scan(lex, error);
736 if(lex->token != '[' && lex->token != '{') {
737 error_set(error, lex, "'[' or '{' expected");
741 return parse_value(lex, error);
750 static int string_get(void *data)
753 string_data_t *stream = (string_data_t *)data;
754 c = stream->data[stream->pos];
764 static int string_eof(void *data)
766 string_data_t *stream = (string_data_t *)data;
767 return (stream->data[stream->pos] == '\0');
770 json_t *json_loads(const char *string, json_error_t *error)
775 string_data_t stream_data = {
780 if(lex_init(&lex, string_get, string_eof, (void *)&stream_data))
783 result = parse_json(&lex, error);
787 lex_scan(&lex, error);
788 if(lex.token != TOKEN_EOF) {
789 error_set(error, &lex, "end of file expected");
799 json_t *json_loadf(FILE *input, json_error_t *error)
804 if(lex_init(&lex, (get_func)fgetc, (eof_func)feof, input))
807 result = parse_json(&lex, error);
811 lex_scan(&lex, error);
812 if(lex.token != TOKEN_EOF) {
813 error_set(error, &lex, "end of file expected");
823 json_t *json_load_file(const char *path, json_error_t *error)
828 fp = fopen(path, "r");
831 error_set(error, NULL, "unable to open %s: %s",
832 path, strerror(errno));
836 result = json_loadf(fp, error);