json_tokener.c

/*
 * Copyright (c) 2004, 2005 Metaparadigm Pte. Ltd.
 * Michael Clark <michael@metaparadigm.com>
 * Copyright (c) 2016 Rainer Gerhards
 *
 * This library is free software; you can redistribute it and/or modify
 * it under the terms of the MIT license. See COPYING for details.
 *
 *
 * Copyright (c) 2008-2009 Yahoo! Inc.  All rights reserved.
 * The copyrights to the contents of this file are licensed under the MIT License
 * (http://www.opensource.org/licenses/mit-license.php)
 */

#include "config.h"

/* this is a work-around until we manage to fix configure.ac */
#ifndef _AIX
#pragma GCC diagnostic ignored "-Wdeclaration-after-statement"
#endif

#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <ctype.h>
#include <string.h>
#include <limits.h>
#include <inttypes.h>

#include "debug.h"
#include "printbuf.h"
#include "arraylist.h"
#include "json_object.h"
#include "json_tokener.h"
#include "json_util.h"

#ifdef HAVE_LOCALE_H
#include <locale.h>
#endif				/* HAVE_LOCALE_H */

#define jt_hexdigit(x) (((x) <= '9') ? (x) - '0' : ((x) & 7) + 9)

#if !HAVE_STRDUP
#error You do not have strdup on your system.
#endif				/* HAVE_STRDUP */

#if !HAVE_STRNCASECMP
#error You do not have strncasecmp on your system.
#endif				/* HAVE_STRNCASECMP */

/* Use C99 NAN by default; if not available, nan("") should work too. */
#ifndef NAN
#define NAN nan("")
#endif				/* !NAN */

static const char fjson_null_str[] = "null";
static const int fjson_null_str_len = sizeof(fjson_null_str) - 1;
static const char fjson_inf_str[] = "Infinity";
static const int fjson_inf_str_len = sizeof(fjson_inf_str) - 1;
static const char fjson_nan_str[] = "NaN";
static const int fjson_nan_str_len = sizeof(fjson_nan_str) - 1;
static const char fjson_true_str[] = "true";
static const int fjson_true_str_len = sizeof(fjson_true_str) - 1;
static const char fjson_false_str[] = "false";
static const int fjson_false_str_len = sizeof(fjson_false_str) - 1;

const char *fjson_tokener_errors[15] = {
	"success",
	"continue",
	"nesting too deep",
	"unexpected end of data",
	"unexpected character",
	"null expected",
	"boolean expected",
	"number expected",
	"array value separator ',' expected",
	"quoted object property name expected",
	"object property name separator ':' expected",
	"object value separator ',' expected",
	"invalid string sequence",
	"expected comment",
	"buffer size overflow"
};

const char *fjson_tokener_error_desc(enum fjson_tokener_error jerr)
{
	int jerr_int = (int)jerr;
	if (jerr_int < 0 || jerr_int >= (int)(sizeof(fjson_tokener_errors) / sizeof(fjson_tokener_errors[0])))
		return "Unknown error, invalid fjson_tokener_error value passed to fjson_tokener_error_desc()";
	return fjson_tokener_errors[jerr];
}

enum fjson_tokener_error fjson_tokener_get_error(fjson_tokener * tok)
{
	return tok->err;
}

/* Stuff for decoding unicode sequences */
#define IS_HIGH_SURROGATE(uc) (((uc) & 0xFC00) == 0xD800)
#define IS_LOW_SURROGATE(uc)  (((uc) & 0xFC00) == 0xDC00)
#define DECODE_SURROGATE_PAIR(hi,lo) ((((hi) & 0x3FF) << 10) + ((lo) & 0x3FF) + 0x10000)
static unsigned char utf8_replacement_char[3] = { 0xEF, 0xBF, 0xBD };

struct fjson_tokener *fjson_tokener_new_ex(int depth)
{
	struct fjson_tokener *tok;

	tok = (struct fjson_tokener *)calloc(1, sizeof(struct fjson_tokener));
	if (!tok)
		return NULL;
	tok->stack = (struct fjson_tokener_srec *)calloc(depth, sizeof(struct fjson_tokener_srec));
	if (!tok->stack) {
		free(tok);
		return NULL;
	}
	tok->pb = printbuf_new();
	tok->max_depth = depth;
	fjson_tokener_reset(tok);
	return tok;
}

struct fjson_tokener *fjson_tokener_new(void)
{
	return fjson_tokener_new_ex(FJSON_TOKENER_DEFAULT_DEPTH);
}

void fjson_tokener_free(struct fjson_tokener *tok)
{
	fjson_tokener_reset(tok);
	if (tok->pb)
		printbuf_free(tok->pb);
	if (tok->stack)
		free(tok->stack);
	free(tok);
}

static void fjson_tokener_reset_level(struct fjson_tokener *tok, int depth)
{
	tok->stack[depth].state = fjson_tokener_state_eatws;
	tok->stack[depth].saved_state = fjson_tokener_state_start;
	fjson_object_put(tok->stack[depth].current);
	tok->stack[depth].current = NULL;
	free(tok->stack[depth].obj_field_name);
	tok->stack[depth].obj_field_name = NULL;
}

void fjson_tokener_reset(struct fjson_tokener *tok)
{
	int i;
	if (!tok)
		return;

	for (i = tok->depth; i >= 0; i--)
		fjson_tokener_reset_level(tok, i);
	tok->depth = 0;
	tok->err = fjson_tokener_success;
}

struct fjson_object *fjson_tokener_parse(const char *str)
{
	enum fjson_tokener_error jerr_ignored;
	struct fjson_object *obj;
	obj = fjson_tokener_parse_verbose(str, &jerr_ignored);
	return obj;
}

struct fjson_object *fjson_tokener_parse_verbose(const char *str, enum fjson_tokener_error *error)
{
	struct fjson_tokener *tok;
	struct fjson_object *obj;

	tok = fjson_tokener_new();
	if (!tok)
		return NULL;
	obj = fjson_tokener_parse_ex(tok, str, -1);
	*error = tok->err;
	if (tok->err != fjson_tokener_success) {
		if (obj != NULL)
			fjson_object_put(obj);
		obj = NULL;
	}

	fjson_tokener_free(tok);
	return obj;
}

#define state  tok->stack[tok->depth].state
#define saved_state  tok->stack[tok->depth].saved_state
#define current tok->stack[tok->depth].current
#define obj_field_name tok->stack[tok->depth].obj_field_name

/* Optimization:
 * fjson_tokener_parse_ex() consumed a lot of CPU in its main loop,
 * iterating character-by character.  A large performance boost is
 * achieved by using tighter loops to locally handle units such as
 * comments and strings.  Loops that handle an entire token within
 * their scope also gather entire strings and pass them to
 * printbuf_memappend() in a single call, rather than calling
 * printbuf_memappend() one char at a time.
 *
 * PEEK_CHAR() and ADVANCE_CHAR() macros are used for code that is
 * common to both the main loop and the tighter loops.
 */

/* PEEK_CHAR(dest, tok) macro:
 *   Peeks at the current char and stores it in dest.
 *   Returns 1 on success, sets tok->err and returns 0 if no more chars.
 *   Implicit inputs:  str, len vars
 */
#define PEEK_CHAR(dest, tok)                                                \
	(((tok)->char_offset == len) ?                                      \
		(((tok)->depth == 0 && state == fjson_tokener_state_eatws && saved_state == fjson_tokener_state_finish) ? \
			(((tok)->err = fjson_tokener_success), 0)           \
		:                                                           \
			(((tok)->err = fjson_tokener_continue), 0)          \
	) :                                                                 \
		(((dest) = *str), 1)                                        \
	)

/* ADVANCE_CHAR() macro:
 *   Incrementes str & tok->char_offset.
 *   For convenience of existing conditionals, returns the old value of c (0 on eof)
 *   Implicit inputs:  c var
 */
#define ADVANCE_CHAR(str, tok) \
	( ++(str), ((tok)->char_offset)++, c)

/* End optimization macro defs */

struct fjson_object *fjson_tokener_parse_ex(struct fjson_tokener *tok, const char *str, int len)
{
	struct fjson_object *obj = NULL;
	char c = '\1';
#ifdef HAVE_SETLOCALE
	char *oldlocale = NULL, *tmplocale;

	tmplocale = setlocale(LC_NUMERIC, NULL);
	if (tmplocale)
		oldlocale = strdup(tmplocale);
	setlocale(LC_NUMERIC, "C");
#endif

	tok->char_offset = 0;
	tok->err = fjson_tokener_success;

	/* this interface is presently not 64-bit clean due to the int len argument
	   and the internal printbuf interface that takes 32-bit int len arguments
	   so the function limits the maximum string size to INT32_MAX (2GB).
	   If the function is called with len == -1 then strlen is called to check
	   the string length is less than INT32_MAX (2GB) */
	if ((len < -1) || (len == -1 && strlen(str) > INT32_MAX)) {
		tok->err = fjson_tokener_error_size;
		return NULL;
	}

	while (PEEK_CHAR(c, tok)) {

redo_char:
		switch (state) {

		case fjson_tokener_state_eatws:
			/* Advance until we change state */
			while (isspace((int)c)) {
				if ((!ADVANCE_CHAR(str, tok)) || (!PEEK_CHAR(c, tok)))
					goto out;
			}
			if (c == '/' && !(tok->flags & FJSON_TOKENER_STRICT)) {
				printbuf_reset(tok->pb);
				printbuf_memappend_fast(tok->pb, &c, 1);
				state = fjson_tokener_state_comment_start;
			} else {
				state = saved_state;
				goto redo_char;
			}
			break;

		case fjson_tokener_state_start:
			switch (c) {
			case '{':
				state = fjson_tokener_state_eatws;
				saved_state = fjson_tokener_state_object_field_start;
				current = fjson_object_new_object();
				break;
			case '[':
				state = fjson_tokener_state_eatws;
				saved_state = fjson_tokener_state_array;
				current = fjson_object_new_array();
				break;
			case 'I':
			case 'i':
				state = fjson_tokener_state_inf;
				printbuf_reset(tok->pb);
				tok->st_pos = 0;
				goto redo_char;
			case 'N':
			case 'n':
				state = fjson_tokener_state_null;	// or NaN
				printbuf_reset(tok->pb);
				tok->st_pos = 0;
				goto redo_char;
			case '\'':
				if (tok->flags & FJSON_TOKENER_STRICT) {
					/* in STRICT mode only double-quote are allowed */
					tok->err = fjson_tokener_error_parse_unexpected;
					goto out;
				}
			case '"':
				state = fjson_tokener_state_string;
				printbuf_reset(tok->pb);
				tok->quote_char = c;
				break;
			case 'T':
			case 't':
			case 'F':
			case 'f':
				state = fjson_tokener_state_boolean;
				printbuf_reset(tok->pb);
				tok->st_pos = 0;
				goto redo_char;
#if defined(__GNUC__)
			case '0' ... '9':
#else
			case '0':
			case '1':
			case '2':
			case '3':
			case '4':
			case '5':
			case '6':
			case '7':
			case '8':
			case '9':
#endif
			case '-':
				state = fjson_tokener_state_number;
				printbuf_reset(tok->pb);
				tok->is_double = 0;
				goto redo_char;
			default:
				tok->err = fjson_tokener_error_parse_unexpected;
				goto out;
			}
			break;

		case fjson_tokener_state_finish:
			if (tok->depth == 0)
				goto out;
			obj = fjson_object_get(current);
			fjson_tokener_reset_level(tok, tok->depth);
			tok->depth--;
			goto redo_char;

		case fjson_tokener_state_inf:	/* aka starts with 'i' */
			{
				size_t size_inf;
				int is_negative = 0;

				printbuf_memappend_fast(tok->pb, &c, 1);
				size_inf = fjson_min(tok->st_pos + 1, fjson_inf_str_len);
				char *infbuf = tok->pb->buf;
				if (*infbuf == '-') {
					infbuf++;
					is_negative = 1;
				}
				if ((!(tok->flags & FJSON_TOKENER_STRICT) &&
				     strncasecmp(fjson_inf_str, infbuf, size_inf) == 0) ||
				    (strncmp(fjson_inf_str, infbuf, size_inf) == 0)
				    ) {
					if (tok->st_pos == fjson_inf_str_len) {
						current = fjson_object_new_double(is_negative ? -INFINITY : INFINITY);
						saved_state = fjson_tokener_state_finish;
						state = fjson_tokener_state_eatws;
						goto redo_char;
					}
				} else {
					tok->err = fjson_tokener_error_parse_unexpected;
					goto out;
				}
				tok->st_pos++;
			}
			break;
		case fjson_tokener_state_null:	/* aka starts with 'n' */
			{
				int size;
				int size_nan;
				printbuf_memappend_fast(tok->pb, &c, 1);
				size = fjson_min(tok->st_pos + 1, fjson_null_str_len);
				size_nan = fjson_min(tok->st_pos + 1, fjson_nan_str_len);
				if ((!(tok->flags & FJSON_TOKENER_STRICT) &&
				     strncasecmp(fjson_null_str, tok->pb->buf, size) == 0)
				    || (strncmp(fjson_null_str, tok->pb->buf, size) == 0)
				    ) {
					if (tok->st_pos == fjson_null_str_len) {
						current = NULL;
						saved_state = fjson_tokener_state_finish;
						state = fjson_tokener_state_eatws;
						goto redo_char;
					}
				} else if ((!(tok->flags & FJSON_TOKENER_STRICT) &&
					    strncasecmp(fjson_nan_str, tok->pb->buf, size_nan) == 0) ||
					   (strncmp(fjson_nan_str, tok->pb->buf, size_nan) == 0)
				    ) {
					if (tok->st_pos == fjson_nan_str_len) {
						current = fjson_object_new_double(NAN);
						saved_state = fjson_tokener_state_finish;
						state = fjson_tokener_state_eatws;
						goto redo_char;
					}
				} else {
					tok->err = fjson_tokener_error_parse_null;
					goto out;
				}
				tok->st_pos++;
			}
			break;

		case fjson_tokener_state_comment_start:
			if (c == '*') {
				state = fjson_tokener_state_comment;
			} else if (c == '/') {
				state = fjson_tokener_state_comment_eol;
			} else {
				tok->err = fjson_tokener_error_parse_comment;
				goto out;
			}
			printbuf_memappend_fast(tok->pb, &c, 1);
			break;

		case fjson_tokener_state_comment:
			{
				/* Advance until we change state */
				const char *case_start = str;
				while (c != '*') {
					if (!ADVANCE_CHAR(str, tok) || !PEEK_CHAR(c, tok)) {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						goto out;
					}
				}
				printbuf_memappend_fast(tok->pb, case_start, 1 + str - case_start);
				state = fjson_tokener_state_comment_end;
			}
			break;

		case fjson_tokener_state_comment_eol:
			{
				/* Advance until we change state */
				const char *case_start = str;
				while (c != '\n') {
					if (!ADVANCE_CHAR(str, tok) || !PEEK_CHAR(c, tok)) {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						goto out;
					}
				}
				printbuf_memappend_fast(tok->pb, case_start, str - case_start);
				MC_DEBUG("fjson_tokener_comment: %s\n", tok->pb->buf);
				state = fjson_tokener_state_eatws;
			}
			break;

		case fjson_tokener_state_comment_end:
			printbuf_memappend_fast(tok->pb, &c, 1);
			if (c == '/') {
				MC_DEBUG("fjson_tokener_comment: %s\n", tok->pb->buf);
				state = fjson_tokener_state_eatws;
			} else {
				state = fjson_tokener_state_comment;
			}
			break;

		case fjson_tokener_state_string:
			{
				/* Advance until we change state */
				const char *case_start = str;
				while (1) {
					if (c == tok->quote_char) {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						current = fjson_object_new_string_len(tok->pb->buf, tok->pb->bpos);
						saved_state = fjson_tokener_state_finish;
						state = fjson_tokener_state_eatws;
						break;
					} else if (c == '\\') {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						saved_state = fjson_tokener_state_string;
						state = fjson_tokener_state_string_escape;
						break;
					}
					if (!ADVANCE_CHAR(str, tok) || !PEEK_CHAR(c, tok)) {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						goto out;
					}
				}
			}
			break;

		case fjson_tokener_state_string_escape:
			switch (c) {
			case '"':
			case '\\':
			case '/':
				printbuf_memappend_fast(tok->pb, &c, 1);
				state = saved_state;
				break;
			case 'b':
			case 'n':
			case 'r':
			case 't':
			case 'f':
				if (c == 'b')
					printbuf_memappend_fast(tok->pb, "\b", 1);
				else if (c == 'n')
					printbuf_memappend_fast(tok->pb, "\n", 1);
				else if (c == 'r')
					printbuf_memappend_fast(tok->pb, "\r", 1);
				else if (c == 't')
					printbuf_memappend_fast(tok->pb, "\t", 1);
				else if (c == 'f')
					printbuf_memappend_fast(tok->pb, "\f", 1);
				state = saved_state;
				break;
			case 'u':
				tok->ucs_char = 0;
				tok->st_pos = 0;
				state = fjson_tokener_state_escape_unicode;
				break;
			default:
				tok->err = fjson_tokener_error_parse_string;
				goto out;
			}
			break;

		case fjson_tokener_state_escape_unicode:
			{
				unsigned int got_hi_surrogate = 0;

				/* Handle a 4-byte sequence, or two sequences if a surrogate pair */
				while (1) {
					if (strchr(fjson_hex_chars, c)) {
						tok->ucs_char +=
						    ((unsigned int)jt_hexdigit(c) << ((3 - tok->st_pos++) * 4));
						if (tok->st_pos == 4) {
							unsigned char unescaped_utf[4];

							if (got_hi_surrogate) {
								if (IS_LOW_SURROGATE(tok->ucs_char)) {
									/* Recalculate the ucs_char, then fall thru to process normally */
									tok->ucs_char =
									    DECODE_SURROGATE_PAIR(got_hi_surrogate,
												  tok->ucs_char);
								} else {
									/* Hi surrogate was not followed by a low surrogate */
									/* Replace the hi and process the rest normally */
									printbuf_memappend_fast(tok->pb,
												(char *)
												utf8_replacement_char,
												3);
								}
								got_hi_surrogate = 0;
								/* clang static analyzer thins that got_hi_surrogate is never read,
								 * however, it is read on each iteration. So I assume clang has a false
								 * positive. We use the otherwise nonsense statement below to make it
								 * happy.
								 */
								if (got_hi_surrogate) {
								};
							}

							if (tok->ucs_char < 0x80) {
								unescaped_utf[0] = tok->ucs_char;
								printbuf_memappend_fast(tok->pb, (char *)unescaped_utf,
											1);
							} else if (tok->ucs_char < 0x800) {
								unescaped_utf[0] = 0xc0 | (tok->ucs_char >> 6);
								unescaped_utf[1] = 0x80 | (tok->ucs_char & 0x3f);
								printbuf_memappend_fast(tok->pb, (char *)unescaped_utf,
											2);
							} else if (IS_HIGH_SURROGATE(tok->ucs_char)) {
								/* Got a high surrogate.  Remember it and look for the
								 * the beginning of another sequence, which should be the
								 * low surrogate.
								 */
								got_hi_surrogate = tok->ucs_char;
								/* Not at end, and the next two chars should be "\u" */
								if ((tok->char_offset + 1 != len) &&
								    (tok->char_offset + 2 != len) &&
								    (str[1] == '\\') && (str[2] == 'u')) {
									/* Advance through the 16 bit surrogate, and move on to the
									 * next sequence. The next step is to process the following
									 * characters.
									 */
									if (!ADVANCE_CHAR(str, tok)
									    || !ADVANCE_CHAR(str, tok)) {
										printbuf_memappend_fast(tok->pb,
													(char *)
													utf8_replacement_char,
													3);
									}
									/* Advance to the first char of the next sequence and
									 * continue processing with the next sequence.
									 */
									if (!ADVANCE_CHAR(str, tok)
									    || !PEEK_CHAR(c, tok)) {
										printbuf_memappend_fast(tok->pb,
													(char *)
													utf8_replacement_char,
													3);
										goto out;
									}
									tok->ucs_char = 0;
									tok->st_pos = 0;
									continue;	/* other fjson_tokener_state_escape_unicode */
								} else {
									/* Got a high surrogate without another sequence following
									 * it.  Put a replacement char in for the hi surrogate
									 * and pretend we finished.
									 */
									printbuf_memappend_fast(tok->pb,
												(char *)
												utf8_replacement_char,
												3);
								}
							} else if (IS_LOW_SURROGATE(tok->ucs_char)) {
								/* Got a low surrogate not preceded by a high */
								printbuf_memappend_fast(tok->pb,
											(char *)utf8_replacement_char,
											3);
							} else if (tok->ucs_char < 0x10000) {
								unescaped_utf[0] = 0xe0 | (tok->ucs_char >> 12);
								unescaped_utf[1] = 0x80 | ((tok->ucs_char >> 6) & 0x3f);
								unescaped_utf[2] = 0x80 | (tok->ucs_char & 0x3f);
								printbuf_memappend_fast(tok->pb, (char *)unescaped_utf,
											3);
							} else if (tok->ucs_char < 0x110000) {
								unescaped_utf[0] =
								    0xf0 | ((tok->ucs_char >> 18) & 0x07);
								unescaped_utf[1] =
								    0x80 | ((tok->ucs_char >> 12) & 0x3f);
								unescaped_utf[2] = 0x80 | ((tok->ucs_char >> 6) & 0x3f);
								unescaped_utf[3] = 0x80 | (tok->ucs_char & 0x3f);
								printbuf_memappend_fast(tok->pb, (char *)unescaped_utf,
											4);
							} else {
								/* Don't know what we got--insert the replacement char */
								printbuf_memappend_fast(tok->pb,
											(char *)utf8_replacement_char,
											3);
							}
							state = saved_state;
							break;
						}
					} else {
						tok->err = fjson_tokener_error_parse_string;
						goto out;
					}
					if (!ADVANCE_CHAR(str, tok) || !PEEK_CHAR(c, tok)) {
						if (got_hi_surrogate)	/* Clean up any pending chars */
							printbuf_memappend_fast(tok->pb, (char *)utf8_replacement_char,
										3);
						goto out;
					}
				}
			}
			break;

		case fjson_tokener_state_boolean:
			{
				int size1, size2;
				printbuf_memappend_fast(tok->pb, &c, 1);
				size1 = fjson_min(tok->st_pos + 1, fjson_true_str_len);
				size2 = fjson_min(tok->st_pos + 1, fjson_false_str_len);
				if ((!(tok->flags & FJSON_TOKENER_STRICT) &&
				     strncasecmp(fjson_true_str, tok->pb->buf, size1) == 0)
				    || (strncmp(fjson_true_str, tok->pb->buf, size1) == 0)
				    ) {
					if (tok->st_pos == fjson_true_str_len) {
						current = fjson_object_new_boolean(1);
						saved_state = fjson_tokener_state_finish;
						state = fjson_tokener_state_eatws;
						goto redo_char;
					}
				} else if ((!(tok->flags & FJSON_TOKENER_STRICT) &&
					    strncasecmp(fjson_false_str, tok->pb->buf, size2) == 0)
					   || (strncmp(fjson_false_str, tok->pb->buf, size2) == 0)) {
					if (tok->st_pos == fjson_false_str_len) {
						current = fjson_object_new_boolean(0);
						saved_state = fjson_tokener_state_finish;
						state = fjson_tokener_state_eatws;
						goto redo_char;
					}
				} else {
					tok->err = fjson_tokener_error_parse_boolean;
					goto out;
				}
				tok->st_pos++;
			}
			break;

		case fjson_tokener_state_number:
			{
				/* Advance until we change state */
				const char *case_start = str;
				int case_len = 0;
				int is_exponent = 0;
				int negativesign_next_possible_location = 1;
				while (c && strchr(fjson_number_chars, c)) {
					++case_len;

					/* non-digit characters checks */
					/* note: since the main loop condition to get here was
					   an input starting with 0-9 or '-', we are
					   protected from input starting with '.' or
					   e/E. */
					if (c == '.') {
						if (tok->is_double != 0) {
							/* '.' can only be found once, and out of the exponent part.
							   Thus, if the input is already flagged as double, it
							   is invalid. */
							tok->err = fjson_tokener_error_parse_number;
							goto out;
						}
						tok->is_double = 1;
					}
					if (c == 'e' || c == 'E') {
						if (is_exponent != 0) {
							/* only one exponent possible */
							tok->err = fjson_tokener_error_parse_number;
							goto out;
						}
						is_exponent = 1;
						tok->is_double = 1;
						/* the exponent part can begin with a negative sign */
						negativesign_next_possible_location = case_len + 1;
					}
					if (c == '-' && case_len != negativesign_next_possible_location) {
						/* If the negative sign is not where expected (ie
						   start of input or start of exponent part), the
						   input is invalid. */
						tok->err = fjson_tokener_error_parse_number;
						goto out;
					}

					if (!ADVANCE_CHAR(str, tok) || !PEEK_CHAR(c, tok)) {
						printbuf_memappend_fast(tok->pb, case_start, case_len);
						goto out;
					}
				}
				if (case_len > 0)
					printbuf_memappend_fast(tok->pb, case_start, case_len);

				// Check for -Infinity
				if (tok->pb->buf[0] == '-' && case_len == 1 && (c == 'i' || c == 'I')) {
					state = fjson_tokener_state_inf;
					goto redo_char;
				}
			}
			{
				int64_t num64;
				double numd;
				if (!tok->is_double && fjson_parse_int64(tok->pb->buf, &num64) == 0) {
					if (num64 && tok->pb->buf[0] == '0' && (tok->flags & FJSON_TOKENER_STRICT)) {
						/* in strict mode, number must not start with 0 */
						tok->err = fjson_tokener_error_parse_number;
						goto out;
					}
					current = fjson_object_new_int64(num64);
				} else if (tok->is_double && fjson_parse_double(tok->pb->buf, &numd) == 0) {
					current = fjson_object_new_double_s(numd, tok->pb->buf);
				} else {
					tok->err = fjson_tokener_error_parse_number;
					goto out;
				}
				saved_state = fjson_tokener_state_finish;
				state = fjson_tokener_state_eatws;
				goto redo_char;
			}
			break;

		case fjson_tokener_state_array_after_sep:
		case fjson_tokener_state_array:
			if (c == ']') {
				if (state == fjson_tokener_state_array_after_sep && (tok->flags & FJSON_TOKENER_STRICT)) {
					tok->err = fjson_tokener_error_parse_unexpected;
					goto out;
				}
				saved_state = fjson_tokener_state_finish;
				state = fjson_tokener_state_eatws;
			} else {
				if (tok->depth >= tok->max_depth - 1) {
					tok->err = fjson_tokener_error_depth;
					goto out;
				}
				state = fjson_tokener_state_array_add;
				tok->depth++;
				fjson_tokener_reset_level(tok, tok->depth);
				goto redo_char;
			}
			break;

		case fjson_tokener_state_array_add:
			fjson_object_array_add(current, obj);
			saved_state = fjson_tokener_state_array_sep;
			state = fjson_tokener_state_eatws;
			goto redo_char;

		case fjson_tokener_state_array_sep:
			if (c == ']') {
				saved_state = fjson_tokener_state_finish;
				state = fjson_tokener_state_eatws;
			} else if (c == ',') {
				saved_state = fjson_tokener_state_array_after_sep;
				state = fjson_tokener_state_eatws;
			} else {
				tok->err = fjson_tokener_error_parse_array;
				goto out;
			}
			break;

		case fjson_tokener_state_object_field_start:
		case fjson_tokener_state_object_field_start_after_sep:
			if (c == '}') {
				if (state == fjson_tokener_state_object_field_start_after_sep &&
				    (tok->flags & FJSON_TOKENER_STRICT)) {
					tok->err = fjson_tokener_error_parse_unexpected;
					goto out;
				}
				saved_state = fjson_tokener_state_finish;
				state = fjson_tokener_state_eatws;
			} else if (c == '"' || c == '\'') {
				tok->quote_char = c;
				printbuf_reset(tok->pb);
				state = fjson_tokener_state_object_field;
			} else {
				tok->err = fjson_tokener_error_parse_object_key_name;
				goto out;
			}
			break;

		case fjson_tokener_state_object_field:
			{
				/* Advance until we change state */
				const char *case_start = str;
				while (1) {
					if (c == tok->quote_char) {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						obj_field_name = strdup(tok->pb->buf);
						saved_state = fjson_tokener_state_object_field_end;
						state = fjson_tokener_state_eatws;
						break;
					} else if (c == '\\') {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						saved_state = fjson_tokener_state_object_field;
						state = fjson_tokener_state_string_escape;
						break;
					}
					if (!ADVANCE_CHAR(str, tok) || !PEEK_CHAR(c, tok)) {
						printbuf_memappend_fast(tok->pb, case_start, str - case_start);
						goto out;
					}
				}
			}
			break;

		case fjson_tokener_state_object_field_end:
			if (c == ':') {
				saved_state = fjson_tokener_state_object_value;
				state = fjson_tokener_state_eatws;
			} else {
				tok->err = fjson_tokener_error_parse_object_key_sep;
				goto out;
			}
			break;

		case fjson_tokener_state_object_value:
			if (tok->depth >= tok->max_depth - 1) {
				tok->err = fjson_tokener_error_depth;
				goto out;
			}
			state = fjson_tokener_state_object_value_add;
			tok->depth++;
			fjson_tokener_reset_level(tok, tok->depth);
			goto redo_char;

		case fjson_tokener_state_object_value_add:
			fjson_object_object_add(current, obj_field_name, obj);
			free(obj_field_name);
			obj_field_name = NULL;
			saved_state = fjson_tokener_state_object_sep;
			state = fjson_tokener_state_eatws;
			goto redo_char;

		case fjson_tokener_state_object_sep:
			if (c == '}') {
				saved_state = fjson_tokener_state_finish;
				state = fjson_tokener_state_eatws;
			} else if (c == ',') {
				saved_state = fjson_tokener_state_object_field_start_after_sep;
				state = fjson_tokener_state_eatws;
			} else {
				tok->err = fjson_tokener_error_parse_object_value_sep;
				goto out;
			}
			break;

		default:
			/* TODO: this should not happen, emit error msg? */
			break;
		}
		if (!ADVANCE_CHAR(str, tok))
			goto out;
	}			/* while(POP_CHAR) */

out:
	if (c && (state == fjson_tokener_state_finish) && (tok->depth == 0) && (tok->flags & FJSON_TOKENER_STRICT)) {
		/* unexpected char after JSON data */
		tok->err = fjson_tokener_error_parse_unexpected;
	}
	if (!c) {		/* We hit an eof char (0) */
		if (state != fjson_tokener_state_finish && saved_state != fjson_tokener_state_finish)
			tok->err = fjson_tokener_error_parse_eof;
	}
#ifdef HAVE_SETLOCALE
	setlocale(LC_NUMERIC, oldlocale);
	if (oldlocale)
		free(oldlocale);
#endif

	if (tok->err == fjson_tokener_success) {
		fjson_object *ret = fjson_object_get(current);
		int ii;

		/* Partially reset, so we parse additional objects on subsequent calls. */
		for (ii = tok->depth; ii >= 0; ii--)
			fjson_tokener_reset_level(tok, ii);
		return ret;
	}

	MC_DEBUG("fjson_tokener_parse_ex: error %s at offset %d\n", fjson_tokener_errors[tok->err], tok->char_offset);
	return NULL;
}

void fjson_tokener_set_flags(struct fjson_tokener *tok, int flags)
{
	tok->flags = flags;
}