ConversionResultFlags parseJSONNumber(raw_buffer *str, int64_t *integerPortion,
int64_t *exponentPortion, int64_t *decimalPortion, int64_t *decimalLeadingZeros)
{
size_t i = 0;
int integerMultiplier = 1;
int exponentMultiplier = 1;
int64_t exponent = 0;
int64_t fraction = 0;
int64_t fractionFactor = 0;
int64_t trailingZeros = 0;
bool validDecimal;
int64_t temp;
ConversionResultFlags result = CONV_OK;
*integerPortion = 0;
if (str->m_len == 0) {
result = CONV_NOT_A_NUM;
goto fast_stop;
}
switch (str->m_str[i]) {
case '-':
integerMultiplier = -1;
i++;
goto parse_integer_portion;
case '0':
i++;
if (i >= str->m_len)
goto fast_stop;
switch(str->m_str[i])
{
case '.':
goto parse_decimal_portion;
case 'e':
case 'E':
goto parse_exponent_portion;
default:
result = CONV_NOT_A_NUM;
goto fast_stop;
}
case '1'...'9':
goto parse_integer_portion;
default:
goto not_a_number;
}
parse_integer_portion:
assert(integerMultiplier == 1 || integerMultiplier == -1);
assert(exponent == 0);
for (; i < str->m_len; i++) {
switch (str->m_str[i]) {
case 'e':
case 'E':
goto parse_exponent_portion;
case '.':
goto parse_decimal_portion;
case '0'...'9':
if (exponent == 0) {
if (integerMultiplier == 1) {
if (*integerPortion > (INT64_MAX / 10))
exponent = 1;
} else {
if (*integerPortion < (INT64_MIN) / 10)
exponent = 1;
}
if (exponent == 0) {
temp = *integerPortion * 10 + integerMultiplier * (str->m_str[i] - '0');
if (UNLIKELY(temp * integerMultiplier < 0)) {
// sign flipped - overflow
exponent = 1;
} else {
*integerPortion = temp;
}
}
} else {
if (exponent++ == INT64_MAX)
return CONV_PRECISION_LOSS | (integerMultiplier == 1 ? CONV_POSITIVE_INFINITY : CONV_NEGATIVE_INFINITY);
}
break;
default:
PJ_LOG_WARN("PBNJSON_PNUM_UNXPCTD_CHAR1", 1, PMLOGKFV("CHAR", "%c", str->m_str[i]),
"Unexpected character %d('%c') in '%.*s' at %zu", (int)str->m_str[i], str->m_str[i], (int)str->m_len, str->m_str, i);
goto not_a_number;
}
}
goto finish_parse;
parse_decimal_portion:
validDecimal = false;
assert(fraction == 0);
assert(fractionFactor == 0);
assert(trailingZeros == 0);
if (str->m_str[i] != '.') {
assert(false);
PJ_LOG_WARN("PBNJSON_PNUM_UNXPCTD_CHAR2", 1, PMLOGKFV("CHAR", "%c", str->m_str[i]),
"Unexpected character %d('%c') in '%.*s' at %zu", (int)str->m_str[i], str->m_str[i], (int) str->m_len, str->m_str, i);
goto not_a_number;
}
i++;
for (; i < str->m_len; i++) {
switch(str->m_str[i]) {
case 'e':
case 'E':
goto parse_exponent_portion;
case '0'...'9':
validDecimal = true;
if (str->m_str[i] == '0') {
// short-circuit - trailing 0s are ignored if that's what they are.
trailingZeros ++;
break;
}
if (UNLIKELY(fractionFactor == INT64_MAX)) {
assert(false);
// this will only become an issue if 10^INT64_MAX < (2^((sizeof(fraction)*8) - 1) - 1)
// which will never happen
PJ_LOG_ERR("PBNJSON_FRACTION_ERR", 1, PMLOGKS("STRING", str->m_str), "Internal error for input: %.*s", (int)str->m_len, str->m_str);
return CONV_GENERIC_ERROR;
}
while (trailingZeros != 0) {
temp = fraction * 10;
if (temp < 0)
goto skip_remaining_decimal;
trailingZeros--;
fractionFactor++;
fraction = temp;
}
fractionFactor++;
if (fraction != INT64_MAX) {
temp = fraction * 10 + (str->m_str[i] - '0');
if (UNLIKELY(temp < 0)) {
fractionFactor--;
goto skip_remaining_decimal;
} else {
fraction = temp;
}
}
break;
default:
PJ_LOG_WARN("PBNJSON_PNUM_UNXPCTD_CHAR3", 1, PMLOGKFV("CHAR", "%c", str->m_str[i]),
"Unexpected character %d('%c') in '%.*s' at %zu", (int)str->m_str[i], str->m_str[i], (int)str->m_len, str->m_str, i);
goto not_a_number;
}
}
if (UNLIKELY(!validDecimal)) {
PJ_LOG_WARN("PBNJSON_UNXPCTD_EOS", 0, "Unexpected end of string at %zu in '%.*s'", i, (int)str->m_len, str->m_str);
goto not_a_number;
}
goto finish_parse;
skip_remaining_decimal:
assert(str->m_str[i] >= '0');
assert(str->m_str[i] <= '9');
result |= CONV_PRECISION_LOSS;
for (; i < str->m_len; i++) {
if (str->m_str[i] >= '0' && str->m_str[i] <= '9')
continue;
if (str->m_str[i] == 'e' || str->m_str[i] == 'E')
goto parse_exponent_portion;
PJ_LOG_WARN("PBNJSON_PNUM_UNXPCTD_CHAR4", 1, PMLOGKFV("CHAR", "%c", str->m_str[i]),
"Unexpected character %d('%c') in '%.*s' at %zu", (int)str->m_str[i], str->m_str[i], (int)str->m_len, str->m_str, i);
goto not_a_number;
}
assert(i == str->m_len);
goto finish_parse;
parse_exponent_portion:
assert(exponent >= 0);
if (UNLIKELY(str->m_str[i] != 'e' && str->m_str[i] != 'E')) {
// problem with the state machine
assert(false);
PJ_LOG_ERR("PBNJSON_NO_EXPONENT", 1, PMLOGKS("STRING", str->m_str), "Expecting an exponent but didn't get one at %zu in '%.*s'", i, (int)str->m_len, str->m_str);
return CONV_GENERIC_ERROR;
}
i++;
switch (str->m_str[i]) {
case '-':
i++;
exponentMultiplier = -1;
break;
case '+':
i++;
case '0'...'9':
exponentMultiplier = 1;
break;
default:
PJ_LOG_WARN("PBNJSON_PNUM_UNXPCTD_CHAR5", 1, PMLOGKFV("CHAR", "%c", str->m_str[i]),
"Unexpected character %d('%c') in '%.*s' at %zu", (int)str->m_str[i], str->m_str[i], (int)str->m_len, str->m_str, i);
goto not_a_number;
}
assert(exponentMultiplier == 1 || exponentMultiplier == -1);
for (; i < str->m_len; i++) {
switch (str->m_str[i]) {
case '0'...'9':
if (exponentMultiplier == 1) {
if (UNLIKELY(exponent > (INT64_MAX / 10)))
goto exponent_overflow;
} else if (exponentMultiplier == -1) {
if (UNLIKELY(exponent < (INT64_MIN / 10)))
goto exponent_overflow;
}
exponent *= 10;
exponent += exponentMultiplier * (str->m_str[i] - '0');
if (exponent * exponentMultiplier < 0) {
goto exponent_overflow;
}
break;
default:
PJ_LOG_WARN("PBNJSON_PNUM_UNXPCTD_CHAR6", 1, PMLOGKFV("CHAR", "%c", str->m_str[i]),
"Unexpected character %d('%c') in '%.*s' at %zu", (int)str->m_str[i], str->m_str[i], (int)str->m_len, str->m_str, i);
goto not_a_number;
}
}
assert(i == str->m_len);
goto finish_parse;
exponent_overflow:
// overflow of a 64-bit exponent - +/- infinity or 0 it is.
assert(exponent > (INT64_MAX / 10 - 10) || exponent < (INT64_MIN / 10 + 10));
if (exponentMultiplier == 1) {
exponent = INT64_MAX;
if (integerMultiplier == 1) {
*integerPortion = INT64_MAX;
result |= CONV_POSITIVE_INFINITY;
} else {
*integerPortion = INT64_MIN;
result |= CONV_NEGATIVE_INFINITY;
}
} else {
result |= CONV_PRECISION_LOSS;
exponent = INT64_MIN;
*integerPortion = 0;
}
goto finish_parse;
finish_parse:
if (trailingZeros) {
PJ_LOG_DBG("%"PRId64 " unnecessary 0s in fraction portion of '%.*s'", trailingZeros, (int)str->m_len, str->m_str);
}
if (fraction == 0) {
assert(fractionFactor == 0);
}
if (*integerPortion == 0 && (decimalPortion == NULL || fraction == 0)) {
// shortcut - exponent is redundant if the number is 0.something but we're
// ignoring the decimal (or there's no fractional portion)
exponent = 0;
if (fraction != 0) {
result |= CONV_PRECISION_LOSS;
}
}
// can't really do this anyways - it would require us shifting values into or out
// of the fractional component when we adjust the integerPortion by the exponent.
// internally, we would never use this case anyways because if we care what the
// fraction is (i.e. we're converting to a floating point), we'll provide the exponent
// pointer anyways
if (exponentPortion == NULL && exponent != 0 && fraction != 0) {
result |= CONV_PRECISION_LOSS;
fraction = 0;
fractionFactor = 0;
}
if (!exponentPortion) {
if (*integerPortion != 0) {
if (exponent > 0) {
while (exponent) {
if (*integerPortion > INT64_MAX / 10) {
assert(integerMultiplier == 1);
result |= CONV_POSITIVE_OVERFLOW;
*integerPortion = INT64_MAX;
break;
} else if (*integerPortion < INT64_MIN / 10) {
assert(integerMultiplier == -1);
result |= CONV_NEGATIVE_OVERFLOW;
*integerPortion = INT64_MIN;
break;
}
if (*integerPortion != 0)
assert(*integerPortion * 10 > 0);
*integerPortion *= 10;
exponent--;
}
} else if (exponent < 0) {
if (fraction) {
result |= CONV_PRECISION_LOSS;
goto lost_precision;
}
while (exponent) {
if (*integerPortion % 10 != 0) {
result |= CONV_PRECISION_LOSS;
goto lost_precision;
}
*integerPortion /= 10;
exponent++;
}
lost_precision:
while (exponent++ && *integerPortion > 0)
*integerPortion /= 10;
}
}
} else {
*exponentPortion = exponent;
}
if (!decimalPortion) {
if (fraction != 0) {
result |= CONV_PRECISION_LOSS;
}
} else {
*decimalPortion = fraction;
*decimalLeadingZeros = fractionFactor;
}
return result;
not_a_number:
return CONV_NOT_A_NUM;
fast_stop:
if (exponentPortion) *exponentPortion = exponent;
if (decimalPortion) *decimalPortion = fraction;
if (decimalLeadingZeros) *decimalLeadingZeros = fractionFactor;
return result;
}
static bool jsax_parse_internal(PJSAXCallbacks *parser, raw_buffer input, JSchemaInfoRef schemaInfo, void **ctxt, bool logError, bool comments)
{
yajl_status parseResult;
PJ_LOG_TRACE("Parsing '%.*s'", RB_PRINTF(input));
if (parser == NULL)
parser = &no_callbacks;
if (jis_null_schema(schemaInfo->m_schema)) {
PJ_LOG_WARN("Cannot match against schema that matches nothing: Schema pointer = %p", schemaInfo->m_schema);
return false;
}
if (schemaInfo->m_schema == jschema_all()) {
PJ_LOG_DBG("Using default empty schema for matching");
} else {
if (schemaInfo->m_resolver == NULL) {
PJ_LOG_DBG("No resolver specified for the schema. Make sure %p doesn't contain any external references", schemaInfo->m_schema);
}
}
if (schemaInfo->m_errHandler == NULL)
schemaInfo->m_errHandler = &null_err_handler;
#ifdef _DEBUG
logError = true;
#endif
yajl_callbacks yajl_cb = {
(pj_yajl_null)parser->m_null, // yajl_null
(pj_yajl_boolean)parser->m_boolean, // yajl_boolean
NULL, // yajl_integer
NULL, // yajl_double
(pj_yajl_number)parser->m_number, // yajl_number
(pj_yajl_string)parser->m_string, // yajl_stirng
(pj_yajl_start_map)parser->m_objStart, // yajl_start_map
(pj_yajl_map_key)parser->m_objKey, // yajl_map_key
(pj_yajl_end_map)parser->m_objEnd, // yajl_end_map
(pj_yajl_start_array)parser->m_arrStart, // yajl_start_array
(pj_yajl_end_array)parser->m_arrEnd, // yajl_end_array
};
yajl_parser_config yajl_opts = {
comments, // comments are not allowed
0, // currently only UTF-8 will be supported for input.
};
PJSAXContext internalCtxt = {
.ctxt = (ctxt != NULL ? *ctxt : NULL),
.m_handlers = &yajl_cb,
.m_errors = schemaInfo->m_errHandler,
};
#if !BYPASS_SCHEMA
internalCtxt.m_validation = jschema_init(schemaInfo);
if (internalCtxt.m_validation == NULL) {
PJ_LOG_WARN("Failed to initialize validation state machine");
return false;
}
#endif
yajl_handle handle = yajl_alloc(&my_bounce, &yajl_opts, NULL, &internalCtxt);
parseResult = yajl_parse(handle, (unsigned char *)input.m_str, input.m_len);
if (ctxt != NULL) *ctxt = jsax_getContext(&internalCtxt);
switch (parseResult) {
case yajl_status_ok:
break;
case yajl_status_client_canceled:
if (ERR_HANDLER_FAILED(schemaInfo->m_errHandler, m_unknown, &internalCtxt))
goto parse_failure;
PJ_LOG_WARN("Client claims they handled an unknown error in '%.*s'", (int)input.m_len, input.m_str);
break;
case yajl_status_insufficient_data:
if (ERR_HANDLER_FAILED(schemaInfo->m_errHandler, m_parser, &internalCtxt))
goto parse_failure;
PJ_LOG_WARN("Client claims they handled incomplete JSON input provided '%.*s'", (int)input.m_len, input.m_str);
break;
case yajl_status_error:
default:
if (ERR_HANDLER_FAILED(schemaInfo->m_errHandler, m_unknown, &internalCtxt))
goto parse_failure;
PJ_LOG_WARN("Client claims they handled an unknown error in '%.*s'", (int)input.m_len, input.m_str);
break;
}
#if !BYPASS_SCHEMA
jschema_state_release(&internalCtxt.m_validation);
#endif
#ifndef NDEBUG
assert(yajl_get_error(handle, 0, NULL, 0) == NULL);
#endif
yajl_free(handle);
return true;
parse_failure:
if (UNLIKELY(logError)) {
unsigned char *errMsg = yajl_get_error(handle, 1, (unsigned char *)input.m_str, input.m_len);
PJ_LOG_WARN("Parser reason for failure:\n'%s'", errMsg);
yajl_free_error(handle, errMsg);
}
#if !BYPASS_SCHEMA
jschema_state_release(&internalCtxt.m_validation);
#endif
yajl_free(handle);
return false;
}
bool jsax_parse_ex(PJSAXCallbacks *parser, raw_buffer input, JSchemaInfoRef schemaInfo, void **ctxt, bool logError)
{
return jsax_parse_internal(parser, input, schemaInfo, ctxt, logError, false);
}
bool jsax_parse(PJSAXCallbacks *parser, raw_buffer input, JSchemaInfoRef schema)
{
assert(schema != NULL);
return jsax_parse_ex(parser, input, schema, NULL, false);
}
