static int lex_scan(lex_t *lex, json_error_t *error)
{
int c;
strbuffer_clear(&lex->saved_text);
if(lex->token == TOKEN_STRING) {
jsonp_free(lex->value.string);
lex->value.string = NULL;
}
c = lex_get(lex, error);
while(c == ' ' || c == '\t' || c == '\n' || c == '\r')
c = lex_get(lex, error);
if(c == STREAM_STATE_EOF) {
lex->token = TOKEN_EOF;
goto out;
}
if(c == STREAM_STATE_ERROR) {
lex->token = TOKEN_INVALID;
goto out;
}
lex_save(lex, c);
if(c == '{' || c == '}' || c == '[' || c == ']' || c == ':' || c == ',')
lex->token = c;
else if(c == '"')
lex_scan_string(lex, error);
else if(l_isdigit(c) || c == '-') {
if(lex_scan_number(lex, c, error))
goto out;
}
else if(l_isalpha(c)) {
/* eat up the whole identifier for clearer error messages */
const char *saved_text;
c = lex_get_save(lex, error);
while(l_isalpha(c))
c = lex_get_save(lex, error);
lex_unget_unsave(lex, c);
saved_text = strbuffer_value(&lex->saved_text);
if(strcmp(saved_text, "true") == 0)
lex->token = TOKEN_TRUE;
else if(strcmp(saved_text, "false") == 0)
lex->token = TOKEN_FALSE;
else if(strcmp(saved_text, "null") == 0)
lex->token = TOKEN_NULL;
else
lex->token = TOKEN_INVALID;
}
else {
/* save the rest of the input UTF-8 sequence to get an error
message of valid UTF-8 */
lex_save_cached(lex);
lex->token = TOKEN_INVALID;
}
out:
return lex->token;
}
static int lex_scan_number(lex_t *lex, int c, json_error_t *error)
{
const char *saved_text;
char *end;
double value;
lex->token = TOKEN_INVALID;
if(c == '-')
c = lex_get_save(lex, error);
if(c == '0') {
c = lex_get_save(lex, error);
if(l_isdigit(c)) {
lex_unget_unsave(lex, c);
goto out;
}
}
else if(l_isdigit(c)) {
c = lex_get_save(lex, error);
while(l_isdigit(c))
c = lex_get_save(lex, error);
}
else {
lex_unget_unsave(lex, c);
goto out;
}
if(c != '.' && c != 'E' && c != 'e') {
json_int_t value;
lex_unget_unsave(lex, c);
saved_text = strbuffer_value(&lex->saved_text);
errno = 0;
value = json_strtoint(saved_text, &end, 10);
if(errno == ERANGE) {
if(value < 0)
error_set(error, lex, "too big negative integer");
else
error_set(error, lex, "too big integer");
goto out;
}
assert(end == saved_text + lex->saved_text.length);
lex->token = TOKEN_INTEGER;
lex->value.integer = value;
return 0;
}
if(c == '.') {
c = lex_get(lex, error);
if(!l_isdigit(c)) {
lex_unget(lex, c);
goto out;
}
lex_save(lex, c);
c = lex_get_save(lex, error);
while(l_isdigit(c))
c = lex_get_save(lex, error);
}
if(c == 'E' || c == 'e') {
c = lex_get_save(lex, error);
if(c == '+' || c == '-')
c = lex_get_save(lex, error);
if(!l_isdigit(c)) {
lex_unget_unsave(lex, c);
goto out;
}
c = lex_get_save(lex, error);
while(l_isdigit(c))
c = lex_get_save(lex, error);
}
lex_unget_unsave(lex, c);
if(jsonp_strtod(&lex->saved_text, &value)) {
error_set(error, lex, "real number overflow");
goto out;
}
lex->token = TOKEN_REAL;
lex->value.real = value;
return 0;
out:
return -1;
}
static int lex_scan_number(lex_t *lex, char c, json_error_t *error)
{
const char *saved_text;
char *end;
double value;
lex->token = TOKEN_INVALID;
if(c == '-')
c = lex_get_save(lex, error);
if(c == '0') {
c = lex_get_save(lex, error);
if(isdigit(c)) {
lex_unget_unsave(lex, c);
goto out;
}
}
else if(isdigit(c)) {
c = lex_get_save(lex, error);
while(isdigit(c))
c = lex_get_save(lex, error);
}
else {
lex_unget_unsave(lex, c);
goto out;
}
if(c != '.' && c != 'E' && c != 'e') {
long value;
lex_unget_unsave(lex, c);
saved_text = strbuffer_value(&lex->saved_text);
value = strtol(saved_text, &end, 10);
assert(end == saved_text + lex->saved_text.length);
if((value == LONG_MAX && errno == ERANGE) || value > INT_MAX) {
error_set(error, lex, "too big integer");
goto out;
}
else if((value == LONG_MIN && errno == ERANGE) || value < INT_MIN) {
error_set(error, lex, "too big negative integer");
goto out;
}
lex->token = TOKEN_INTEGER;
lex->value.integer = (int)value;
return 0;
}
if(c == '.') {
c = lex_get(lex, error);
if(!isdigit(c))
goto out;
lex_save(lex, c);
c = lex_get_save(lex, error);
while(isdigit(c))
c = lex_get_save(lex, error);
}
if(c == 'E' || c == 'e') {
c = lex_get_save(lex, error);
if(c == '+' || c == '-')
c = lex_get_save(lex, error);
if(!isdigit(c)) {
lex_unget_unsave(lex, c);
goto out;
}
c = lex_get_save(lex, error);
while(isdigit(c))
c = lex_get_save(lex, error);
}
lex_unget_unsave(lex, c);
saved_text = strbuffer_value(&lex->saved_text);
value = strtod(saved_text, &end);
assert(end == saved_text + lex->saved_text.length);
if(value == 0 && errno == ERANGE) {
error_set(error, lex, "real number underflow");
goto out;
}
/* Cannot test for +/-HUGE_VAL because the HUGE_VAL constant is
only defined in C99 mode. So let's trust in sole errno. */
else if(errno == ERANGE) {
error_set(error, lex, "real number overflow");
goto out;
}
lex->token = TOKEN_REAL;
lex->value.real = value;
return 0;
out:
return -1;
}
static void lex_scan_string(lex_t *lex, json_error_t *error)
{
int c;
const char *p;
char *t;
int i;
lex->value.string = NULL;
lex->token = TOKEN_INVALID;
c = lex_get_save(lex, error);
while(c != '"') {
if(c == STREAM_STATE_ERROR)
goto out;
else if(c == STREAM_STATE_EOF) {
error_set(error, lex, "premature end of input");
goto out;
}
else if(0 <= c && c <= 0x1F) {
/* control character */
lex_unget_unsave(lex, c);
if(c == '\n')
error_set(error, lex, "unexpected newline", c);
else
error_set(error, lex, "control character 0x%x", c);
goto out;
}
else if(c == '\\') {
c = lex_get_save(lex, error);
if(c == 'u') {
c = lex_get_save(lex, error);
for(i = 0; i < 4; i++) {
if(!l_isxdigit(c)) {
error_set(error, lex, "invalid escape");
goto out;
}
c = lex_get_save(lex, error);
}
}
else if(c == '"' || c == '\\' || c == '/' || c == 'b' ||
c == 'f' || c == 'n' || c == 'r' || c == 't')
c = lex_get_save(lex, error);
else {
error_set(error, lex, "invalid escape");
goto out;
}
}
else
c = lex_get_save(lex, error);
}
/* the actual value is at most of the same length as the source
string, because:
- shortcut escapes (e.g. "\t") (length 2) are converted to 1 byte
- a single \uXXXX escape (length 6) is converted to at most 3 bytes
- two \uXXXX escapes (length 12) forming an UTF-16 surrogate pair
are converted to 4 bytes
*/
lex->value.string = jsonp_malloc(lex->saved_text.length + 1);
if(!lex->value.string) {
/* this is not very nice, since TOKEN_INVALID is returned */
goto out;
}
/* the target */
t = lex->value.string;
/* + 1 to skip the " */
p = strbuffer_value(&lex->saved_text) + 1;
while(*p != '"') {
if(*p == '\\') {
p++;
if(*p == 'u') {
char buffer[4];
int length;
int32_t value;
value = decode_unicode_escape(p);
p += 5;
if(0xD800 <= value && value <= 0xDBFF) {
/* surrogate pair */
if(*p == '\\' && *(p + 1) == 'u') {
int32_t value2 = decode_unicode_escape(++p);
p += 5;
if(0xDC00 <= value2 && value2 <= 0xDFFF) {
/* valid second surrogate */
value =
((value - 0xD800) << 10) +
(value2 - 0xDC00) +
0x10000;
}
else {
/* invalid second surrogate */
error_set(error, lex,
"invalid Unicode '\\u%04X\\u%04X'",
value, value2);
goto out;
}
}
else {
/* no second surrogate */
error_set(error, lex, "invalid Unicode '\\u%04X'",
value);
goto out;
}
}
else if(0xDC00 <= value && value <= 0xDFFF) {
error_set(error, lex, "invalid Unicode '\\u%04X'", value);
goto out;
}
else if(value == 0)
{
error_set(error, lex, "\\u0000 is not allowed");
goto out;
}
if(utf8_encode(value, buffer, &length))
assert(0);
memcpy(t, buffer, length);
t += length;
}
else {
switch(*p) {
case '"': case '\\': case '/':
*t = *p; break;
case 'b': *t = '\b'; break;
case 'f': *t = '\f'; break;
case 'n': *t = '\n'; break;
case 'r': *t = '\r'; break;
case 't': *t = '\t'; break;
default: assert(0);
}
t++;
p++;
}
}
else
*(t++) = *(p++);
}
*t = '\0';
lex->token = TOKEN_STRING;
return;
out:
jsonp_free(lex->value.string);
}
static int lex_scan_number(lex_t *lex, char c, json_error_t *error)
{
const char *saved_text;
char *end;
double value;
lex->token = TOKEN_INVALID;
if(c == '-')
c = lex_get_save(lex, error);
if(c == '0') {
c = lex_get_save(lex, error);
if(isdigit(c)) {
lex_unget_unsave(lex, c);
goto out;
}
}
else if(isdigit(c)) {
c = lex_get_save(lex, error);
while(isdigit(c))
c = lex_get_save(lex, error);
}
else {
lex_unget_unsave(lex, c);
goto out;
}
if(c != '.' && c != 'E' && c != 'e') {
long value;
lex_unget_unsave(lex, c);
saved_text = strbuffer_value(&lex->saved_text);
value = strtol(saved_text, &end, 10);
assert(end == saved_text + lex->saved_text.length);
if((value == LONG_MAX && errno == ERANGE) || value > INT_MAX)
{
unsigned long uvalue = strtoul(saved_text, &end, 10);
if( !(uvalue>value) )
{
error_set(error, lex, "too big integer");
goto out;
}
else
{
value = uvalue;
}
}
else if((value == LONG_MIN && errno == ERANGE) || value < INT_MIN) {
error_set(error, lex, "too big negative integer");
goto out;
}
lex->token = TOKEN_INTEGER;
lex->value.integer = (int)value;
return 0;
}
if(c == '.') {
c = lex_get(lex, error);
if(!isdigit(c))
goto out;
lex_save(lex, c);
c = lex_get_save(lex, error);
while(isdigit(c))
c = lex_get_save(lex, error);
}
if(c == 'E' || c == 'e') {
c = lex_get_save(lex, error);
if(c == '+' || c == '-')
c = lex_get_save(lex, error);
if(!isdigit(c)) {
lex_unget_unsave(lex, c);
goto out;
}
c = lex_get_save(lex, error);
while(isdigit(c))
c = lex_get_save(lex, error);
}
lex_unget_unsave(lex, c);
saved_text = strbuffer_value(&lex->saved_text);
value = strtod(saved_text, &end);
assert(end == saved_text + lex->saved_text.length);
if(errno == ERANGE && value != 0) {
error_set(error, lex, "real number overflow");
goto out;
}
lex->token = TOKEN_REAL;
lex->value.real = value;
return 0;
out:
return -1;
}