Compare commits

..
Author SHA1 Message Date
Niels Lohmann a3c2b0897e Fix -Weffc++ errors in the #5198 locale test
GCC's -Weffc++ (an error in ci_test_gcc and ci_test_standards_gcc)
rejected LocaleSwitchingSax: it has a pointer data member but does not
declare its copy operations, and its vectors are not initialized in the
member initializer list. Store the locale name as a std::string and give
the vectors brace initializers, like SaxEventLogger in
unit-deserialization.cpp.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
2026-09-28 04:19:20 +02:00
Niels Lohmann cfe7c9e732 Merge branch 'develop' into claude/issue-5198-84c9fe
Signed-off-by: Niels Lohmann <mail@nlohmann.me>
2026-09-27 20:57:56 +02:00
Niels Lohmann f37492a6d6 Stop the strtod retry loop when the decimal point is unchanged
convert_float_locale_aware() repeated the conversion until strtod
consumed the whole token, assuming an early stop can only mean a locale
change. Under a locale whose decimal point is not a single character
(e.g. the two-byte U+066B of ar_EG.UTF-8, ar_SA.UTF-8, or fa_IR.UTF-8,
all available on macOS), the in-place substitution can never succeed,
so parsing any float that reaches the strtod fallback (for example
3.14159265358979323846 at C++11) hung forever. Before this branch, the
same input was truncated.

Retry only if the decimal point changed since the previous attempt;
otherwise keep the value strtod parsed so far, as before. Add a test
that parses such numbers under a multi-byte decimal point locale; it
hangs without this change.

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
2026-09-27 18:22:43 +02:00
Niels Lohmann b69794bd80 Look up the locale decimal point at conversion time, not lexer construction
The lexer read localeconv()->decimal_point once in its constructor and wrote
that character into token_buffer in place of '.'. The strtod fallback then
used the locale current at conversion time, so an LC_NUMERIC change in
between (parser callback, SAX handler, another thread) truncated the value
in release builds and fired the endptr assertion in debug builds.

token_buffer now always holds '.'. Only the strtof/strtod/strtold fallback
depends on the locale: it looks up the decimal point right before the call,
restores '.' afterwards, and repeats the conversion if the locale changed in
between. As a side effect, std::from_chars and Clinger's fast path now also
apply under locales whose decimal point is not '.'.

Fixes #5198

Signed-off-by: Niels Lohmann <mail@nlohmann.me>
2026-09-27 17:14:00 +02:00
9 changed files with 393 additions and 165 deletions
+3 -11
View File
@@ -86,9 +86,6 @@ Strong exception safety: if an exception occurs, the original value stays intact
- Throws [`out_of_range.410`](../../home/exceptions.md#jsonexceptionout_of_range410) if an array index in the passed
JSON pointer `ptr` exceeds the range of `size_type` (e.g., on 32-bit platforms).
For the **const** version, an object key or array index in `ptr` that does not exist is not reported by an
exception, but is undefined behavior (see the notes below). Use [`at`](at.md) for checked access.
## Complexity
1. Constant if `idx` is in the range of the array. Otherwise, linear in `idx - size()`.
@@ -103,12 +100,9 @@ Strong exception safety: if an exception occurs, the original value stays intact
The following cases apply to the **const** overloads; the non-const overloads instead insert the missing element
(see the notes below).
1. If the element at index `idx` does not exist, the behavior is undefined and is **guarded by a
[runtime assertion](../../features/assertions.md)**!
1. If the element at index `idx` does not exist, the behavior is undefined.
2. If the element with key `key` does not exist, the behavior is undefined and is **guarded by a
[runtime assertion](../../features/assertions.md)**!
3. If the JSON pointer `ptr` refers to an object key or an array index that does not exist, the behavior is
undefined and is **guarded by a [runtime assertion](../../features/assertions.md)**!
1. The non-const version may add values: If `idx` is beyond the range of the array (i.e., `idx >= size()`), then the
array is silently filled up with `#!json null` values to make `idx` a valid reference to the last stored element. In
@@ -263,11 +257,9 @@ Strong exception safety: if an exception occurs, the original value stays intact
## Version history
1. Added in version 1.0.0. A missing index in the const version is guarded by a runtime assertion since version
3.13.0.
1. Added in version 1.0.0.
2. Added in version 1.0.0. Added overloads for `T* key` in version 1.1.0. Removed overloads for `T* key` (replaced by 3)
in version 3.11.0.
3. Added in version 3.11.0. Fixed in version 3.13.0 to consistently accept `std::string_view`-convertible keys, as
already supported by [`at`](at.md), [`value`](value.md), [`find`](find.md), and other lookup functions.
4. Added in version 2.0.0. A missing array index in the const version is guarded by a runtime assertion since
version 3.13.0.
4. Added in version 2.0.0.
+7 -31
View File
@@ -16,15 +16,14 @@ before including the `json.hpp` header.
## Function with runtime assertions
### Unchecked access to a const value
### Unchecked object access to a const value
Function [`operator[]`](../api/basic_json/operator%5B%5D.md) implements unchecked access for arrays and objects. Whereas
a missing element is added in the case of non-const values, accessing a const value with a missing object key or an
invalid array index is undefined behavior (think of a dereferenced null pointer) and yields a runtime assertion. This
also applies to a [JSON pointer](json_pointer.md) that refers to a missing key or an invalid index.
Function [`operator[]`](../api/basic_json/operator%5B%5D.md) implements unchecked access for objects. Whereas a missing
key is added in the case of non-const objects, accessing a const object with a missing key is undefined behavior (think
of a dereferenced null pointer) and yields a runtime assertion.
If you are not sure whether an element exists, use checked access with the [`at` function](../api/basic_json/at.md)
or call the [`contains` function](../api/basic_json/contains.md) before.
If you are not sure whether an element in an object exists, use checked access with the
[`at` function](../api/basic_json/at.md) or call the [`contains` function](../api/basic_json/contains.md) before.
See also the documentation on [element access](element_access/index.md).
@@ -47,30 +46,7 @@ See also the documentation on [element access](element_access/index.md).
Output:
```
Assertion failed: (it != m_data.m_value.object->end()), function operator[], file json.hpp, line 28795.
```
??? example "Example 2: Invalid array index in a JSON pointer"
The following code will trigger an assertion at runtime:
```cpp
#include <nlohmann/json.hpp>
using json = nlohmann::json;
using namespace nlohmann::literals;
int main()
{
const json j = {{"array", {1, 2, 3}}};
auto v = j["/array/5"_json_pointer];
}
```
Output:
```
Assertion failed: (idx < m_data.m_value.array->size()), function operator[], file json.hpp, line 28758.
Assertion failed: (m_value.object->find(key) != m_value.object->end()), function operator[], file json.hpp, line 2144.
```
### Constructing from an uninitialized iterator range
+76 -39
View File
@@ -206,7 +206,6 @@ class lexer : public lexer_base<BasicJsonType>
explicit lexer(InputAdapterType&& adapter, bool ignore_comments_ = false, bool discard_number_values_ = false) noexcept
: ia(std::move(adapter))
, ignore_comments(ignore_comments_)
, decimal_point_char(static_cast<char_int_type>(get_decimal_point()))
, discard_number_values(discard_number_values_)
{}
@@ -222,8 +221,7 @@ class lexer : public lexer_base<BasicJsonType>
// locales
/////////////////////
/// return the locale-dependent decimal point
JSON_HEDLEY_PURE
/// return the decimal point of the current locale
static char get_decimal_point() noexcept
{
const auto* loc = localeconv();
@@ -1092,9 +1090,10 @@ class lexer : public lexer_base<BasicJsonType>
token_type::value_float if number could be successfully scanned,
token_type::parse_error otherwise
@note The scanner is independent of the current locale. Internally, the
locale's decimal point is used instead of `.` to work with the
locale-dependent converters.
@note The scanner is independent of the current locale: token_buffer
always holds `.`. Only the std::strtod fallback of convert_number()
depends on the locale, and it looks up the decimal point right
before converting (see convert_float_locale_aware()).
*/
token_type scan_number() // lgtm [cpp/use-of-goto] `goto` is used in this function to implement the number-parsing state machine described above. By design, any finite input will eventually reach the "done" state or return token_type::parse_error. In each intermediate state, 1 byte of the input is appended to the token_buffer vector, and only the already initialized variables token_buffer, number_type, and error_message are manipulated.
{
@@ -1183,7 +1182,7 @@ scan_number_zero:
{
case '.':
{
add(decimal_point_char);
add(current);
decimal_point_position = token_buffer.size() - 1;
goto scan_number_decimal1;
}
@@ -1220,7 +1219,7 @@ scan_number_any1:
case '.':
{
add(decimal_point_char);
add(current);
decimal_point_position = token_buffer.size() - 1;
goto scan_number_decimal1;
}
@@ -1462,9 +1461,9 @@ scan_number_done:
// Only a number below 1 can carry further insignificant zeros, and only
// while the count stays at the limit does removing them change the
// answer - so this loop is skipped for all but a few tokens. Note
// token_buffer holds the locale's decimal point, so the fraction is
// located through decimal_point_position rather than by searching '.'.
// answer - so this loop is skipped for all but a few tokens. The
// fraction is located through decimal_point_position rather than by
// searching '.'.
if (lead_zero != 0)
{
JSON_ASSERT(has_dot != 0); // an integer "0" cannot reach the limit
@@ -1482,8 +1481,8 @@ scan_number_done:
@brief convert the number text in token_buffer to its value and token type
The digit sequence in token_buffer has already been validated (by the
scan_number() state machine or by the contiguous fast path) and holds the
locale decimal point in place of '.'. Integers are parsed first and fall
scan_number() state machine or by the contiguous fast path) and holds '.'
as decimal point, independent of the locale. Integers are parsed first and fall
back to floating point on overflow. This is shared so both scanners produce
identical results.
@@ -1563,7 +1562,7 @@ scan_number_done:
// integer conversion above overflowed. Prefer std::from_chars
// (Eisel-Lemire, locale-independent, correctly rounded) when available;
// otherwise the exact Clinger fast path (double only); otherwise the
// locale-aware strtof/strtod.
// locale-aware strtof/strtod/strtold.
if (parse_float_from_chars(num_begin, num_end, value_float))
{
return token_type::value_float;
@@ -1572,26 +1571,75 @@ scan_number_done:
// extra pass over the token's bytes, which otherwise shows up on
// high-precision inputs such as canada.json
if (mantissa_fits_clinger(mantissa_end)
&& parse_float_fast(num_begin, num_end, decimal_point_char, value_float))
&& parse_float_fast(num_begin, num_end, value_float))
{
return token_type::value_float;
}
char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
strtof(value_float, token_buffer.data(), &endptr);
// we checked the number format before
JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
convert_float_locale_aware();
return token_type::value_float;
}
/*!
@brief convert the float in token_buffer with strtof/strtod/strtold
These functions expect the decimal point of the *current* locale, so it is
looked up right before the conversion instead of once when the lexer is
constructed: a locale change in between (by a parser callback, a SAX
handler, or another thread) must not truncate the value (#5198). The
token has been validated before, so if the conversion stops early and the
decimal point changed in the meantime, the locale changed between the
lookup and the call, and the conversion is repeated with the new decimal
point. If the decimal point did not change, a retry cannot succeed: the
locale's decimal point is not a single character (e.g., the two-byte
U+066B of ar_EG.UTF-8 or fa_IR.UTF-8) and cannot be substituted in place.
The value strtod parsed up to that point is kept, as before this change.
Note that changing the locale in another thread *while* strtod runs is
undefined behavior of the C library, which this function cannot prevent.
*/
void convert_float_locale_aware()
{
const bool has_dot = decimal_point_position != std::string::npos;
char decimal_point = get_decimal_point();
for (;;)
{
const bool substitute = has_dot && decimal_point != '.';
if (substitute)
{
token_buffer[decimal_point_position] = static_cast<typename string_t::value_type>(decimal_point);
}
char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
strtof(value_float, token_buffer.data(), &endptr);
if (substitute)
{
// get_string() hands the token to the SAX interface with '.'
token_buffer[decimal_point_position] = '.';
}
if (JSON_HEDLEY_LIKELY(endptr == token_buffer.data() + token_buffer.size()))
{
return;
}
// retry only if the locale changed; otherwise, this would loop forever
const char current_decimal_point = get_decimal_point();
if (current_decimal_point == decimal_point)
{
return;
}
decimal_point = current_decimal_point;
}
}
/*!
@brief contiguous fast path for scanning a number
Parses the whole number token straight from the input buffer, avoiding the
per-character get()/add() of scan_number(). On success it fills token_buffer
(with the locale decimal point substituted, as scan_number() does) and
(as scan_number() does) and
returns the token type. On anything it does not fully recognize as a
well-formed number it makes no state change and returns
token_type::uninitialized, so the caller falls back to scan_number(), which
@@ -1707,16 +1755,11 @@ scan_number_done:
}
#endif
// materialize the token exactly as scan_number() would, substituting the
// locale decimal point so convert_number()'s strtof fallback stays valid.
// reset() already cleared token_buffer, so append() fills it (assign() is
// avoided because custom string_t types need not provide it)
// materialize the token exactly as scan_number() would. reset() already
// cleared token_buffer, so append() fills it (assign() is avoided
// because custom string_t types need not provide it)
token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(data), len);
if (dot_index != std::string::npos)
{
token_buffer[dot_index] = static_cast<typename string_t::value_type>(decimal_point_char);
decimal_point_position = dot_index;
}
decimal_point_position = dot_index;
ia.bulk_skip(len - 1);
position.chars_read_total += (len - 1);
@@ -1983,11 +2026,7 @@ scan_number_done:
/// return current string value (implicitly resets the token; useful only once)
string_t& get_string()
{
// translate decimal points from locale back to '.' (#4084)
if (decimal_point_char != '.' && decimal_point_position != std::string::npos)
{
token_buffer[decimal_point_position] = '.';
}
// a number token holds '.' regardless of the locale (#4084)
return token_buffer;
}
@@ -2283,9 +2322,7 @@ scan_number_done:
number_unsigned_t value_unsigned = 0;
number_float_t value_float = 0;
/// the decimal point
const char_int_type decimal_point_char = '.';
/// the position of the decimal point in the input
/// the position of the decimal point in token_buffer
std::size_t decimal_point_position = std::string::npos;
/// whether the caller (e.g. accept()/json_sax_acceptor) only needs the
+8 -13
View File
@@ -118,14 +118,12 @@ std::strtod. The parser only activates for number_float_t == double; float and
long double keep the std::strtof/std::strtold paths (see the templated overload
below).
@param[in] first pointer to the first character of the number
@param[in] last pointer past the last character
@param[in] decimal_point the (locale-dependent) decimal point character
@param[out] out the parsed value on success
@param[in] first pointer to the first character of the number
@param[in] last pointer past the last character
@param[out] out the parsed value on success
@return true if the value was parsed exactly; false to fall back to strtod
*/
template<typename DecimalPointType>
bool parse_float_fast(const char* first, const char* last, DecimalPointType decimal_point, double& out) noexcept
inline bool parse_float_fast(const char* first, const char* last, double& out) noexcept
{
#if defined(FLT_EVAL_METHOD) && FLT_EVAL_METHOD != 0
// Clinger's fast path is only exact when double operations are evaluated in
@@ -136,7 +134,6 @@ bool parse_float_fast(const char* first, const char* last, DecimalPointType deci
// std::from_chars / std::strtod path.
static_cast<void>(first);
static_cast<void>(last);
static_cast<void>(decimal_point);
static_cast<void>(out);
return false;
#else
@@ -175,7 +172,7 @@ bool parse_float_fast(const char* first, const char* last, DecimalPointType deci
++num_digits;
fractional_digits += static_cast<int>(seen_dot);
}
else if (static_cast<DecimalPointType>(c) == decimal_point)
else if (c == '.')
{
if (JSON_HEDLEY_UNLIKELY(seen_dot))
{
@@ -260,8 +257,8 @@ bool parse_float_fast(const char* first, const char* last, DecimalPointType deci
}
/// fast float path is only exact for `double`; decline for float/long double
template<typename DecimalPointType, typename FloatType>
bool parse_float_fast(const char* /*first*/, const char* /*last*/, DecimalPointType /*decimal_point*/, FloatType& /*out*/) noexcept
template<typename FloatType>
bool parse_float_fast(const char* /*first*/, const char* /*last*/, FloatType& /*out*/) noexcept
{
return false;
}
@@ -273,9 +270,7 @@ std::from_chars is locale-independent, correctly rounded, and - via the
Eisel-Lemire algorithm in modern standard libraries - much faster than strtod
over the whole value range (not just the Clinger subset). It is used only when
__cpp_lib_to_chars indicates full floating-point support and only when it
consumes the entire token ([first, last)); a partial parse means the buffer
uses a non-'.' locale decimal point, in which case the caller falls back to the
locale-aware path. An under-/overflow (result_out_of_range) also declines, so
consumes the entire token ([first, last)). An under-/overflow (result_out_of_range) also declines, so
the caller's strtod fallback supplies the well-defined ±inf/0 result the parser
expects (side-stepping the P4168 divergence between implementations).
+2 -8
View File
@@ -535,10 +535,6 @@ class json_pointer
@return const reference to the JSON value pointed to by the JSON
pointer
@pre Every object key and array index the pointer refers to exists.
Like the const operator[] for keys and indices, a missing one is
undefined behavior, guarded by a runtime assertion.
@throw parse_error.106 if an array index begins with '0'
@throw parse_error.109 if an array index was not a number
@throw out_of_range.402 if the array index '-' is used
@@ -553,8 +549,7 @@ class json_pointer
{
case detail::value_t::object:
{
// use unchecked object access; the const operator[]
// asserts that the key exists
// use unchecked object access
ptr = &ptr->operator[](reference_token);
break;
}
@@ -567,8 +562,7 @@ class json_pointer
JSON_THROW(detail::out_of_range::create(402, detail::concat("array index '-' (", std::to_string(ptr->m_data.m_value.array->size()), ") is out of range"), ptr));
}
// use unchecked array access; the const operator[]
// asserts that the index exists
// use unchecked array access
ptr = &ptr->operator[](array_index<BasicJsonType>(reference_token));
break;
}
-1
View File
@@ -2866,7 +2866,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
// const operator[] only works for arrays
if (JSON_HEDLEY_LIKELY(is_array()))
{
JSON_ASSERT(idx < m_data.m_value.array->size());
return m_data.m_value.array->operator[](idx);
}
+86 -61
View File
@@ -8605,14 +8605,12 @@ std::strtod. The parser only activates for number_float_t == double; float and
long double keep the std::strtof/std::strtold paths (see the templated overload
below).
@param[in] first pointer to the first character of the number
@param[in] last pointer past the last character
@param[in] decimal_point the (locale-dependent) decimal point character
@param[out] out the parsed value on success
@param[in] first pointer to the first character of the number
@param[in] last pointer past the last character
@param[out] out the parsed value on success
@return true if the value was parsed exactly; false to fall back to strtod
*/
template<typename DecimalPointType>
bool parse_float_fast(const char* first, const char* last, DecimalPointType decimal_point, double& out) noexcept
inline bool parse_float_fast(const char* first, const char* last, double& out) noexcept
{
#if defined(FLT_EVAL_METHOD) && FLT_EVAL_METHOD != 0
// Clinger's fast path is only exact when double operations are evaluated in
@@ -8623,7 +8621,6 @@ bool parse_float_fast(const char* first, const char* last, DecimalPointType deci
// std::from_chars / std::strtod path.
static_cast<void>(first);
static_cast<void>(last);
static_cast<void>(decimal_point);
static_cast<void>(out);
return false;
#else
@@ -8662,7 +8659,7 @@ bool parse_float_fast(const char* first, const char* last, DecimalPointType deci
++num_digits;
fractional_digits += static_cast<int>(seen_dot);
}
else if (static_cast<DecimalPointType>(c) == decimal_point)
else if (c == '.')
{
if (JSON_HEDLEY_UNLIKELY(seen_dot))
{
@@ -8747,8 +8744,8 @@ bool parse_float_fast(const char* first, const char* last, DecimalPointType deci
}
/// fast float path is only exact for `double`; decline for float/long double
template<typename DecimalPointType, typename FloatType>
bool parse_float_fast(const char* /*first*/, const char* /*last*/, DecimalPointType /*decimal_point*/, FloatType& /*out*/) noexcept
template<typename FloatType>
bool parse_float_fast(const char* /*first*/, const char* /*last*/, FloatType& /*out*/) noexcept
{
return false;
}
@@ -8760,9 +8757,7 @@ std::from_chars is locale-independent, correctly rounded, and - via the
Eisel-Lemire algorithm in modern standard libraries - much faster than strtod
over the whole value range (not just the Clinger subset). It is used only when
__cpp_lib_to_chars indicates full floating-point support and only when it
consumes the entire token ([first, last)); a partial parse means the buffer
uses a non-'.' locale decimal point, in which case the caller falls back to the
locale-aware path. An under-/overflow (result_out_of_range) also declines, so
consumes the entire token ([first, last)). An under-/overflow (result_out_of_range) also declines, so
the caller's strtod fallback supplies the well-defined ±inf/0 result the parser
expects (side-stepping the P4168 divergence between implementations).
@@ -9303,7 +9298,6 @@ class lexer : public lexer_base<BasicJsonType>
explicit lexer(InputAdapterType&& adapter, bool ignore_comments_ = false, bool discard_number_values_ = false) noexcept
: ia(std::move(adapter))
, ignore_comments(ignore_comments_)
, decimal_point_char(static_cast<char_int_type>(get_decimal_point()))
, discard_number_values(discard_number_values_)
{}
@@ -9319,8 +9313,7 @@ class lexer : public lexer_base<BasicJsonType>
// locales
/////////////////////
/// return the locale-dependent decimal point
JSON_HEDLEY_PURE
/// return the decimal point of the current locale
static char get_decimal_point() noexcept
{
const auto* loc = localeconv();
@@ -10189,9 +10182,10 @@ class lexer : public lexer_base<BasicJsonType>
token_type::value_float if number could be successfully scanned,
token_type::parse_error otherwise
@note The scanner is independent of the current locale. Internally, the
locale's decimal point is used instead of `.` to work with the
locale-dependent converters.
@note The scanner is independent of the current locale: token_buffer
always holds `.`. Only the std::strtod fallback of convert_number()
depends on the locale, and it looks up the decimal point right
before converting (see convert_float_locale_aware()).
*/
token_type scan_number() // lgtm [cpp/use-of-goto] `goto` is used in this function to implement the number-parsing state machine described above. By design, any finite input will eventually reach the "done" state or return token_type::parse_error. In each intermediate state, 1 byte of the input is appended to the token_buffer vector, and only the already initialized variables token_buffer, number_type, and error_message are manipulated.
{
@@ -10280,7 +10274,7 @@ scan_number_zero:
{
case '.':
{
add(decimal_point_char);
add(current);
decimal_point_position = token_buffer.size() - 1;
goto scan_number_decimal1;
}
@@ -10317,7 +10311,7 @@ scan_number_any1:
case '.':
{
add(decimal_point_char);
add(current);
decimal_point_position = token_buffer.size() - 1;
goto scan_number_decimal1;
}
@@ -10559,9 +10553,9 @@ scan_number_done:
// Only a number below 1 can carry further insignificant zeros, and only
// while the count stays at the limit does removing them change the
// answer - so this loop is skipped for all but a few tokens. Note
// token_buffer holds the locale's decimal point, so the fraction is
// located through decimal_point_position rather than by searching '.'.
// answer - so this loop is skipped for all but a few tokens. The
// fraction is located through decimal_point_position rather than by
// searching '.'.
if (lead_zero != 0)
{
JSON_ASSERT(has_dot != 0); // an integer "0" cannot reach the limit
@@ -10579,8 +10573,8 @@ scan_number_done:
@brief convert the number text in token_buffer to its value and token type
The digit sequence in token_buffer has already been validated (by the
scan_number() state machine or by the contiguous fast path) and holds the
locale decimal point in place of '.'. Integers are parsed first and fall
scan_number() state machine or by the contiguous fast path) and holds '.'
as decimal point, independent of the locale. Integers are parsed first and fall
back to floating point on overflow. This is shared so both scanners produce
identical results.
@@ -10660,7 +10654,7 @@ scan_number_done:
// integer conversion above overflowed. Prefer std::from_chars
// (Eisel-Lemire, locale-independent, correctly rounded) when available;
// otherwise the exact Clinger fast path (double only); otherwise the
// locale-aware strtof/strtod.
// locale-aware strtof/strtod/strtold.
if (parse_float_from_chars(num_begin, num_end, value_float))
{
return token_type::value_float;
@@ -10669,26 +10663,75 @@ scan_number_done:
// extra pass over the token's bytes, which otherwise shows up on
// high-precision inputs such as canada.json
if (mantissa_fits_clinger(mantissa_end)
&& parse_float_fast(num_begin, num_end, decimal_point_char, value_float))
&& parse_float_fast(num_begin, num_end, value_float))
{
return token_type::value_float;
}
char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
strtof(value_float, token_buffer.data(), &endptr);
// we checked the number format before
JSON_ASSERT(endptr == token_buffer.data() + token_buffer.size());
convert_float_locale_aware();
return token_type::value_float;
}
/*!
@brief convert the float in token_buffer with strtof/strtod/strtold
These functions expect the decimal point of the *current* locale, so it is
looked up right before the conversion instead of once when the lexer is
constructed: a locale change in between (by a parser callback, a SAX
handler, or another thread) must not truncate the value (#5198). The
token has been validated before, so if the conversion stops early and the
decimal point changed in the meantime, the locale changed between the
lookup and the call, and the conversion is repeated with the new decimal
point. If the decimal point did not change, a retry cannot succeed: the
locale's decimal point is not a single character (e.g., the two-byte
U+066B of ar_EG.UTF-8 or fa_IR.UTF-8) and cannot be substituted in place.
The value strtod parsed up to that point is kept, as before this change.
Note that changing the locale in another thread *while* strtod runs is
undefined behavior of the C library, which this function cannot prevent.
*/
void convert_float_locale_aware()
{
const bool has_dot = decimal_point_position != std::string::npos;
char decimal_point = get_decimal_point();
for (;;)
{
const bool substitute = has_dot && decimal_point != '.';
if (substitute)
{
token_buffer[decimal_point_position] = static_cast<typename string_t::value_type>(decimal_point);
}
char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
strtof(value_float, token_buffer.data(), &endptr);
if (substitute)
{
// get_string() hands the token to the SAX interface with '.'
token_buffer[decimal_point_position] = '.';
}
if (JSON_HEDLEY_LIKELY(endptr == token_buffer.data() + token_buffer.size()))
{
return;
}
// retry only if the locale changed; otherwise, this would loop forever
const char current_decimal_point = get_decimal_point();
if (current_decimal_point == decimal_point)
{
return;
}
decimal_point = current_decimal_point;
}
}
/*!
@brief contiguous fast path for scanning a number
Parses the whole number token straight from the input buffer, avoiding the
per-character get()/add() of scan_number(). On success it fills token_buffer
(with the locale decimal point substituted, as scan_number() does) and
(as scan_number() does) and
returns the token type. On anything it does not fully recognize as a
well-formed number it makes no state change and returns
token_type::uninitialized, so the caller falls back to scan_number(), which
@@ -10804,16 +10847,11 @@ scan_number_done:
}
#endif
// materialize the token exactly as scan_number() would, substituting the
// locale decimal point so convert_number()'s strtof fallback stays valid.
// reset() already cleared token_buffer, so append() fills it (assign() is
// avoided because custom string_t types need not provide it)
// materialize the token exactly as scan_number() would. reset() already
// cleared token_buffer, so append() fills it (assign() is avoided
// because custom string_t types need not provide it)
token_buffer.append(reinterpret_cast<const typename string_t::value_type*>(data), len);
if (dot_index != std::string::npos)
{
token_buffer[dot_index] = static_cast<typename string_t::value_type>(decimal_point_char);
decimal_point_position = dot_index;
}
decimal_point_position = dot_index;
ia.bulk_skip(len - 1);
position.chars_read_total += (len - 1);
@@ -11080,11 +11118,7 @@ scan_number_done:
/// return current string value (implicitly resets the token; useful only once)
string_t& get_string()
{
// translate decimal points from locale back to '.' (#4084)
if (decimal_point_char != '.' && decimal_point_position != std::string::npos)
{
token_buffer[decimal_point_position] = '.';
}
// a number token holds '.' regardless of the locale (#4084)
return token_buffer;
}
@@ -11380,9 +11414,7 @@ scan_number_done:
number_unsigned_t value_unsigned = 0;
number_float_t value_float = 0;
/// the decimal point
const char_int_type decimal_point_char = '.';
/// the position of the decimal point in the input
/// the position of the decimal point in token_buffer
std::size_t decimal_point_position = std::string::npos;
/// whether the caller (e.g. accept()/json_sax_acceptor) only needs the
@@ -19179,10 +19211,6 @@ class json_pointer
@return const reference to the JSON value pointed to by the JSON
pointer
@pre Every object key and array index the pointer refers to exists.
Like the const operator[] for keys and indices, a missing one is
undefined behavior, guarded by a runtime assertion.
@throw parse_error.106 if an array index begins with '0'
@throw parse_error.109 if an array index was not a number
@throw out_of_range.402 if the array index '-' is used
@@ -19197,8 +19225,7 @@ class json_pointer
{
case detail::value_t::object:
{
// use unchecked object access; the const operator[]
// asserts that the key exists
// use unchecked object access
ptr = &ptr->operator[](reference_token);
break;
}
@@ -19211,8 +19238,7 @@ class json_pointer
JSON_THROW(detail::out_of_range::create(402, detail::concat("array index '-' (", std::to_string(ptr->m_data.m_value.array->size()), ") is out of range"), ptr));
}
// use unchecked array access; the const operator[]
// asserts that the index exists
// use unchecked array access
ptr = &ptr->operator[](array_index<BasicJsonType>(reference_token));
break;
}
@@ -28755,7 +28781,6 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
// const operator[] only works for arrays
if (JSON_HEDLEY_LIKELY(is_array()))
{
JSON_ASSERT(idx < m_data.m_value.array->size());
return m_data.m_value.array->operator[](idx);
}
+1 -1
View File
@@ -666,7 +666,7 @@ TEST_CASE("parse_float_fast declines what it cannot convert exactly")
// always safe: the caller then falls back to a slower, exact conversion.
const auto fast = [](const std::string & s, double & out)
{
return nlohmann::detail::parse_float_fast(s.data(), s.data() + s.size(), '.', out);
return nlohmann::detail::parse_float_fast(s.data(), s.data() + s.size(), out);
};
double out = 0;
+210
View File
@@ -12,7 +12,12 @@
#include <nlohmann/json.hpp>
using nlohmann::json;
#include <array>
#include <clocale>
#include <map>
#include <string>
#include <utility>
#include <vector>
struct ParserImpl final: public nlohmann::json_sax<json>
{
@@ -175,3 +180,208 @@ TEST_CASE("locale-dependent test (LC_NUMERIC=de_DE)")
MESSAGE("locale de_DE is not usable");
}
}
namespace
{
// records the numbers of a flat array and switches LC_NUMERIC to the given
// locale once the array opens - after the lexer was constructed, but before
// any number in the array is lexed
struct LocaleSwitchingSax final: public nlohmann::json_sax<json>
{
explicit LocaleSwitchingSax(const char* switch_to)
: locale_after_open(switch_to)
{}
bool null() override
{
return true;
}
bool boolean(bool /*val*/) override
{
return true;
}
bool number_integer(json::number_integer_t /*val*/) override
{
return true;
}
bool number_unsigned(json::number_unsigned_t /*val*/) override
{
return true;
}
bool number_float(json::number_float_t val, const json::string_t& s) override
{
values.push_back(val);
strings.push_back(s);
return true;
}
bool string(json::string_t& /*val*/) override
{
return true;
}
bool binary(json::binary_t& /*val*/) override
{
return true;
}
bool start_object(std::size_t /*val*/) override
{
return true;
}
bool key(json::string_t& /*val*/) override
{
return true;
}
bool end_object() override
{
return true;
}
bool start_array(std::size_t /*val*/) override
{
switched = std::setlocale(LC_NUMERIC, locale_after_open.c_str()) != nullptr;
return true;
}
bool end_array() override
{
return true;
}
bool parse_error(std::size_t /*val*/, const std::string& /*val*/, const nlohmann::detail::exception& /*val*/) override
{
return false;
}
std::string locale_after_open;
bool switched = false;
std::vector<json::number_float_t> values {}; // NOLINT(readability-redundant-member-init)
std::vector<json::string_t> strings {}; // NOLINT(readability-redundant-member-init)
};
} // namespace
TEST_CASE("locale changes between lexer construction and number conversion (#5198)")
{
// The numbers are chosen so that the conversion also takes the strtod
// fallback, which honors the locale that is current at conversion time:
// too many significant digits for Clinger's fast path, an underflow that
// std::from_chars rejects, and a plain value.
const std::vector<std::string> numbers = {"3.14159265358979323846", "1.5e-400", "12.34", "-0.000123456789012345678"};
std::string text = "[";
for (const auto& n : numbers)
{
text += (text.size() == 1 ? "" : ",") + n;
}
text += "]";
using long_double_json = nlohmann::basic_json<std::map, std::vector, std::string, bool, std::int64_t, std::uint64_t, long double>;
// reference values, parsed without a locale switch
REQUIRE(std::setlocale(LC_NUMERIC, "C") != nullptr);
const json expected = json::parse(text);
const long_double_json expected_ld = long_double_json::parse(text);
const std::array<std::pair<const char*, const char*>, 2> transitions =
{
{
{"C", "de_DE"},
{"de_DE", "C"}
}
};
for (const auto& transition : transitions)
{
CAPTURE(transition.first);
CAPTURE(transition.second);
if (std::setlocale(LC_NUMERIC, transition.first) == nullptr)
{
MESSAGE("locale is not usable");
continue;
}
// SAX parsing
{
LocaleSwitchingSax sax(transition.second);
CHECK(json::sax_parse(text, &sax));
if (sax.switched)
{
CHECK(sax.values == expected.get<std::vector<json::number_float_t>>());
CHECK(sax.strings == numbers);
}
}
// DOM parsing with a callback
{
bool switched = false;
const auto cb = [&](int /*depth*/, json::parse_event_t event, json& /*parsed*/)
{
if (event == json::parse_event_t::array_start)
{
switched = std::setlocale(LC_NUMERIC, transition.second) != nullptr;
}
return true;
};
const json j = json::parse(text, cb);
if (switched)
{
CHECK(j == expected);
}
}
// a long double goes through std::strtold unless std::from_chars supports it
{
bool switched = false;
const auto cb = [&](int /*depth*/, long_double_json::parse_event_t event, long_double_json& /*parsed*/)
{
if (event == long_double_json::parse_event_t::array_start)
{
switched = std::setlocale(LC_NUMERIC, transition.second) != nullptr;
}
return true;
};
const long_double_json j = long_double_json::parse(text, cb);
if (switched)
{
CHECK(j == expected_ld);
}
}
}
std::setlocale(LC_NUMERIC, "C");
}
TEST_CASE("locale with a multi-byte decimal point")
{
// Some locales use a decimal point that is not a single character, e.g.
// U+066B ARABIC DECIMAL SEPARATOR (two bytes in UTF-8). It cannot be
// substituted in place for '.', so the strtod fallback stops early. The
// conversion must still terminate rather than retry forever.
const std::array<const char*, 6> names = {{"ar_EG.UTF-8", "ar_SA.UTF-8", "fa_IR.UTF-8", "ps_AF.UTF-8", "ar_EG", "fa_IR"}};
bool tested = false;
for (const char* name : names)
{
if (std::setlocale(LC_NUMERIC, name) == nullptr)
{
continue;
}
const std::string decimal_point = std::localeconv()->decimal_point;
if (decimal_point.size() < 2)
{
continue;
}
CAPTURE(name);
tested = true;
// too many significant digits for Clinger's fast path, and an underflow
// that std::from_chars rejects: both reach the strtod fallback
json j;
CHECK_NOTHROW(j = json::parse("[3.14159265358979323846, 1.5e-400, -0.000123456789012345678]"));
CHECK(j.is_array());
CHECK(json::accept("3.14159265358979323846"));
// a value the locale-independent paths convert is not affected
CHECK(json::parse("12.5") == 12.5);
}
if (!tested)
{
MESSAGE("no locale with a multi-byte decimal point is usable");
}
std::setlocale(LC_NUMERIC, "C");
}