Compare commits

..
Author SHA1 Message Date
Niels Lohmann 0395ad0992 Pass the complete text of a UBJSON/BJData high-precision number to SAX
For a high-precision number that is a floating-point number, the
UBJSON/BJData reader passed lexer::get_token_string() as the source
text to SAX number_float(). That helper is meant for error messages and
shows a token of more than 8192 characters only by its first and last
4096 characters around "...", so a SAX handler received a shortened
string for a long number.

Pass the number text the lexer collected while scanning the number
(lexer::get_string()) instead. It is complete and, as before, holds
only the number, without any whitespace or byte order mark around it
in the payload. Error messages still use get_token_string().

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
2026-11-16 08:27:07 +01:00
6 changed files with 52 additions and 112 deletions

No files matched your search

@@ -4096,6 +4096,10 @@ class binary_reader
auto number_lexer = detail::lexer<BasicJsonType, ia_type>(detail::input_adapter(number_vector), false);
const auto result_number = number_lexer.scan();
const auto number_string = number_lexer.get_token_string();
// the number as the lexer read it, for SAX number_float(): unlike
// get_token_string(), which is meant for error messages and shortens
// long tokens, it is complete; the next scan() replaces it
const string_t number_text = number_lexer.get_string();
const auto result_remainder = number_lexer.scan();
using token_type = typename detail::lexer_base<BasicJsonType>::token_type;
@@ -4122,10 +4126,7 @@ class binary_reader
number_string,
out_of_range::create(exception_id::number_overflow, concat("number overflow parsing '", number_string, '\''), nullptr));
}
// number_string is a std::string, while the SAX interface takes a
// string_t; convert explicitly, as the two are only implicitly
// convertible for some string types
return sax->number_float(parsed_float, string_t(number_string.data(), number_string.size()));
return sax->number_float(parsed_float, number_text);
}
case token_type::uninitialized:
case token_type::literal_true:
+2 -41
View File
@@ -636,7 +636,7 @@ class serializer
put_literal("\"subtype\": ");
if (binary.has_subtype())
{
dump_subtype(binary.subtype());
dump_integer(binary.subtype());
}
else
{
@@ -662,7 +662,7 @@ class serializer
put_literal("],\"subtype\":");
if (binary.has_subtype())
{
dump_subtype(binary.subtype());
dump_integer(binary.subtype());
put_char('}');
}
else
@@ -1883,40 +1883,6 @@ class serializer
write_buffer_pos = pos;
}
/*!
@brief dump the subtype of a binary value
The subtype is a std::uint64_t, whatever the number types are. @ref
dump_integer converts its argument to the magnitude type of
number_unsigned_t, so it writes the subtype only if that type has at least
64 bits; otherwise (it would truncate the subtype) the digits are written
one by one.
@param[in] subtype the subtype to dump
*/
void dump_subtype(std::uint64_t subtype)
{
dump_subtype(subtype, std::integral_constant<bool, magnitude_has_at_least_64_bits> {});
}
void dump_subtype(std::uint64_t subtype, std::true_type /*magnitude has at least 64 bits*/)
{
dump_integer(subtype);
}
void dump_subtype(std::uint64_t subtype, std::false_type /*magnitude has less than 64 bits*/)
{
std::array<char, 20> digits{{}}; // 2^64 - 1 has 20 digits
std::size_t start = digits.size();
do
{
digits[--start] = static_cast<char>('0' + (subtype % 10));
subtype /= 10;
}
while (subtype != 0);
put_string(digits, start, digits.size());
}
/*!
@brief dump an integer
@@ -2170,11 +2136,6 @@ class serializer
static constexpr bool magnitude_has_at_most_64_bits = std::numeric_limits<number_magnitude_t>::is_specialized
&& std::numeric_limits<number_magnitude_t>::digits <= 64;
/// whether every std::uint64_t fits number_magnitude_t (then dump_integer
/// can write a binary subtype)
static constexpr bool magnitude_has_at_least_64_bits = std::numeric_limits<number_magnitude_t>::is_specialized
&& std::numeric_limits<number_magnitude_t>::digits >= 64;
/// whether @a x fits 64 bits: always if number_magnitude_t has at most
/// 64 bits (selected by type, as the comparison would be reported as
/// always true), else compared
+7 -45
View File
@@ -22774,6 +22774,10 @@ class binary_reader
auto number_lexer = detail::lexer<BasicJsonType, ia_type>(detail::input_adapter(number_vector), false);
const auto result_number = number_lexer.scan();
const auto number_string = number_lexer.get_token_string();
// the number as the lexer read it, for SAX number_float(): unlike
// get_token_string(), which is meant for error messages and shortens
// long tokens, it is complete; the next scan() replaces it
const string_t number_text = number_lexer.get_string();
const auto result_remainder = number_lexer.scan();
using token_type = typename detail::lexer_base<BasicJsonType>::token_type;
@@ -22800,10 +22804,7 @@ class binary_reader
number_string,
out_of_range::create(exception_id::number_overflow, concat("number overflow parsing '", number_string, '\''), nullptr));
}
// number_string is a std::string, while the SAX interface takes a
// string_t; convert explicitly, as the two are only implicitly
// convertible for some string types
return sax->number_float(parsed_float, string_t(number_string.data(), number_string.size()));
return sax->number_float(parsed_float, number_text);
}
case token_type::uninitialized:
case token_type::literal_true:
@@ -38122,7 +38123,7 @@ class serializer
put_literal("\"subtype\": ");
if (binary.has_subtype())
{
dump_subtype(binary.subtype());
dump_integer(binary.subtype());
}
else
{
@@ -38148,7 +38149,7 @@ class serializer
put_literal("],\"subtype\":");
if (binary.has_subtype())
{
dump_subtype(binary.subtype());
dump_integer(binary.subtype());
put_char('}');
}
else
@@ -39369,40 +39370,6 @@ class serializer
write_buffer_pos = pos;
}
/*!
@brief dump the subtype of a binary value
The subtype is a std::uint64_t, whatever the number types are. @ref
dump_integer converts its argument to the magnitude type of
number_unsigned_t, so it writes the subtype only if that type has at least
64 bits; otherwise (it would truncate the subtype) the digits are written
one by one.
@param[in] subtype the subtype to dump
*/
void dump_subtype(std::uint64_t subtype)
{
dump_subtype(subtype, std::integral_constant<bool, magnitude_has_at_least_64_bits> {});
}
void dump_subtype(std::uint64_t subtype, std::true_type /*magnitude has at least 64 bits*/)
{
dump_integer(subtype);
}
void dump_subtype(std::uint64_t subtype, std::false_type /*magnitude has less than 64 bits*/)
{
std::array<char, 20> digits{{}}; // 2^64 - 1 has 20 digits
std::size_t start = digits.size();
do
{
digits[--start] = static_cast<char>('0' + (subtype % 10));
subtype /= 10;
}
while (subtype != 0);
put_string(digits, start, digits.size());
}
/*!
@brief dump an integer
@@ -39656,11 +39623,6 @@ class serializer
static constexpr bool magnitude_has_at_most_64_bits = std::numeric_limits<number_magnitude_t>::is_specialized
&& std::numeric_limits<number_magnitude_t>::digits <= 64;
/// whether every std::uint64_t fits number_magnitude_t (then dump_integer
/// can write a binary subtype)
static constexpr bool magnitude_has_at_least_64_bits = std::numeric_limits<number_magnitude_t>::is_specialized
&& std::numeric_limits<number_magnitude_t>::digits >= 64;
/// whether @a x fits 64 bits: always if number_magnitude_t has at most
/// 64 bits (selected by type, as the comparison would be reported as
/// always true), else compared
+19
View File
@@ -1278,6 +1278,25 @@ TEST_CASE("BJData")
CHECK(j.dump() == "3.141592653589793");
}
SECTION("long floating-point number")
{
// the SAX parser receives the number text as it is in the
// input, also when it is longer than the part of a token
// that error messages show
const std::vector<std::size_t> lengths = {8192, 8193, 10000};
for (const auto length : lengths)
{
CAPTURE(length)
const std::string number = "1." + std::string(length - 2, '3');
std::vector<std::uint8_t> vec = {'H', 'I', static_cast<std::uint8_t>(length & 0xFF), static_cast<std::uint8_t>(length >> 8)};
vec.insert(vec.end(), number.begin(), number.end());
SaxEventLogger logger;
CHECK(json::sax_parse(vec, &logger, json::input_format_t::bjdata));
CHECK(logger.events == std::vector<std::string>({"number_float(" + number + ")"}));
CHECK(json::from_bjdata(vec).is_number_float());
}
}
SECTION("errors")
{
// error while parsing length
-22
View File
@@ -483,25 +483,3 @@ TEST_CASE("custom number types: 32-bit integers in the binary formats")
CHECK(json32::to_msgpack(values) == nlohmann::json::to_msgpack(values64));
CHECK(json32::to_ubjson(values) == nlohmann::json::to_ubjson(values64));
}
TEST_CASE("custom number types: binary subtypes wider than the integer types")
{
// the subtype of a binary value is a std::uint64_t, whatever the number types are
using json8 = nlohmann::json::with_integers_t<std::int8_t, std::uint8_t>;
using json16 = nlohmann::json::with_integers_t<std::int16_t, std::uint16_t>;
using json32 = nlohmann::json::with_integers_t<std::int32_t, std::uint32_t>;
const std::uint64_t max_uint64 = (std::numeric_limits<std::uint64_t>::max)();
CHECK(json8::binary({1}, 300).dump() == "{\"bytes\":[1],\"subtype\":300}");
CHECK(json8::binary({200}, 255).dump() == "{\"bytes\":[200],\"subtype\":255}");
CHECK(json16::binary({1}, 65543).dump() == "{\"bytes\":[1],\"subtype\":65543}");
CHECK(json32::binary({1}, 4294967301u).dump() == "{\"bytes\":[1],\"subtype\":4294967301}");
CHECK(json32::binary({1}, max_uint64).dump() == "{\"bytes\":[1],\"subtype\":18446744073709551615}");
CHECK(json32::binary({1}, 0).dump() == "{\"bytes\":[1],\"subtype\":0}");
CHECK(json32::binary({1}, 4294967301u).dump(1) == "{\n \"bytes\": [1],\n \"subtype\": 4294967301\n}");
CHECK(json8::binary({1}).dump() == "{\"bytes\":[1],\"subtype\":null}");
// unchanged for the default number types
CHECK(nlohmann::json::binary({1}, 4294967301u).dump() == "{\"bytes\":[1],\"subtype\":4294967301}");
CHECK(nlohmann::json::binary({1}, max_uint64).dump(1) == "{\n \"bytes\": [1],\n \"subtype\": 18446744073709551615\n}");
}
+19
View File
@@ -793,6 +793,25 @@ TEST_CASE("UBJSON")
CHECK(j.dump() == "3.141592653589793");
}
SECTION("long floating-point number")
{
// the SAX parser receives the number text as it is in the
// input, also when it is longer than the part of a token
// that error messages show
const std::vector<std::size_t> lengths = {8192, 8193, 10000};
for (const auto length : lengths)
{
CAPTURE(length)
const std::string number = "1." + std::string(length - 2, '3');
std::vector<std::uint8_t> vec = {'H', 'I', static_cast<std::uint8_t>(length >> 8), static_cast<std::uint8_t>(length & 0xFF)};
vec.insert(vec.end(), number.begin(), number.end());
SaxEventLogger logger;
CHECK(json::sax_parse(vec, &logger, json::input_format_t::ubjson));
CHECK(logger.events == std::vector<std::string>({"number_float(" + number + ")"}));
CHECK(json::from_ubjson(vec).is_number_float());
}
}
SECTION("errors")
{
// error while parsing length