Extract current_is_whitespace() to deduplicate skip_whitespace()'s two whitespace checks

Signed-off-by: Niels Lohmann <mail@nlohmann.me>

Makefile

@@ -1,4 +1,4 @@
-.PHONY: pretty clean ChangeLog.md release update_hedley update_hedley_undef
+.PHONY: pretty clean ChangeLog.md release
 
 ##########################################################################
 # configuration
@@ -41,8 +41,6 @@ all:
 	@echo "fuzz_testing_ubjson - prepare fuzz testing of the UBJSON parser"
 	@echo "pretty - beautify code with Artistic Style"
 	@echo "run_benchmarks - build and run benchmarks"
-	@echo "update_hedley - download Hedley and regenerate hedley.hpp / hedley_undef.hpp"
-	@echo "update_hedley_undef - rebuild hedley_undef.hpp from the JSON_HEDLEY_* #define names in hedley.hpp"
 
 
 ##########################################################################
@@ -243,24 +241,11 @@ update_hedley:
 	rm -f include/nlohmann/thirdparty/hedley/hedley.hpp include/nlohmann/thirdparty/hedley/hedley_undef.hpp
 	curl https://raw.githubusercontent.com/nemequ/hedley/master/hedley.h -o include/nlohmann/thirdparty/hedley/hedley.hpp
 	$(SED) -i 's/HEDLEY_/JSON_HEDLEY_/g' include/nlohmann/thirdparty/hedley/hedley.hpp
+	grep "[[:blank:]]*#[[:blank:]]*undef" include/nlohmann/thirdparty/hedley/hedley.hpp | grep -v "__" | sort | uniq | $(SED) 's/ //g' | $(SED) 's/undef/undef /g' > include/nlohmann/thirdparty/hedley/hedley_undef.hpp
 	$(SED) -i '1s/^/#pragma once\n\n/' include/nlohmann/thirdparty/hedley/hedley.hpp
-	$(MAKE) update_hedley_undef
+	$(SED) -i '1s/^/#pragma once\n\n/' include/nlohmann/thirdparty/hedley/hedley_undef.hpp
 	$(MAKE) amalgamate
 
-# Rebuild hedley_undef.hpp from every JSON_HEDLEY_* name that hedley.hpp
-# #defines. Hedley does not #undef all of its public macros internally (see
-# #5408), so grepping those #undef lines misses names such as
-# JSON_HEDLEY_PRAGMA. cmake/scripts/gen_hedley_undef_check.cmake is the
-# single source of truth for this extraction (tests/CMakeLists.txt uses the
-# same script, in MODE=checks, to generate the matching leak-check test), so
-# the vendored header, the generated #undef list, and the regression test
-# cannot drift apart.
-update_hedley_undef:
-	cmake -DHEDLEY_HPP=include/nlohmann/thirdparty/hedley/hedley.hpp \
-	      -DOUTPUT=include/nlohmann/thirdparty/hedley/hedley_undef.hpp \
-	      -DMODE=undef \
-	      -P cmake/scripts/gen_hedley_undef_check.cmake
-
 ##########################################################################
 # serve_header.py
 ##########################################################################

cmake/scripts/gen_hedley_undef_check.cmake

@@ -1,105 +0,0 @@
-# Shared extractor for the JSON_HEDLEY_* macro names defined in hedley.hpp.
-#
-# Every macro that hedley.hpp #defines must be #undef-ed again once json.hpp
-# has been fully processed (see include/nlohmann/detail/macro_unscope.hpp
-# and https://github.com/nlohmann/json/issues/5408). Deriving the macro list
-# straight from hedley.hpp here -- instead of hand-maintaining it in two
-# places -- means hedley_undef.hpp and the regression test that checks for
-# leaked macros can never drift apart, even after a future `make
-# update_hedley` pulls in new macros from upstream Hedley.
-#
-#   MODE=undef  (default): write hedley_undef.hpp (SPDX header, #pragma once,
-#               one #undef per macro name) -- used by `make update_hedley_undef`
-#   MODE=checks:            write one #ifdef/FAIL_CHECK/#endif per macro name,
-#               meant to be #include-d inside a TEST_CASE -- used by
-#               tests/CMakeLists.txt to (re)generate the include for
-#               tests/src/unit-no-macro-leak.cpp
-#
-# Required variables:
-#   HEDLEY_HPP  path to include/nlohmann/thirdparty/hedley/hedley.hpp
-#   OUTPUT      path of the file to (over)write
-# Optional:
-#   MODE        "undef" (default) or "checks"
-
-if(NOT DEFINED HEDLEY_HPP OR NOT DEFINED OUTPUT)
-    message(FATAL_ERROR "HEDLEY_HPP and OUTPUT must be set")
-endif()
-
-if(NOT EXISTS "${HEDLEY_HPP}")
-    message(FATAL_ERROR "Hedley header not found: ${HEDLEY_HPP}")
-endif()
-
-if(NOT DEFINED MODE)
-    set(MODE undef)
-endif()
-
-if(NOT MODE STREQUAL "undef" AND NOT MODE STREQUAL "checks")
-    message(FATAL_ERROR "MODE must be undef or checks, got: ${MODE}")
-endif()
-
-# Line-anchored, like `grep -oE "^[[:blank:]]*#[[:blank:]]*define[[:blank:]]+JSON_HEDLEY_[A-Za-z0-9_]+"`.
-# Unanchored matching would also pick up JSON_HEDLEY_* mentions inside
-# comments or string literals elsewhere in the file, which must not turn
-# into #undef lines.
-file(STRINGS "${HEDLEY_HPP}" hedley_lines)
-set(macro_names)
-foreach(line IN LISTS hedley_lines)
-    if("${line}" MATCHES "^[ \t]*#[ \t]*define[ \t]+(JSON_HEDLEY_[A-Za-z0-9_]+)")
-        list(APPEND macro_names "${CMAKE_MATCH_1}")
-    endif()
-endforeach()
-
-if(NOT macro_names)
-    message(FATAL_ERROR "No JSON_HEDLEY_* macros found in ${HEDLEY_HPP}")
-endif()
-
-list(REMOVE_DUPLICATES macro_names)
-# Lexicographic, locale-independent (ASCII-only names) -- matches `LC_ALL=C sort`.
-list(SORT macro_names COMPARE STRING)
-list(LENGTH macro_names macro_count)
-
-set(generated "")
-if(MODE STREQUAL "undef")
-    # Same banner `make update_hedley_undef` would stamp by hand, so the
-    # recipe is self-contained and its output is byte-stable across reruns.
-    string(APPEND generated "//     __ _____ _____ _____\n")
-    string(APPEND generated "//  __|  |   __|     |   | |  JSON for Modern C++\n")
-    string(APPEND generated "// |  |  |__   |  |  | | | |  version 3.12.0\n")
-    string(APPEND generated "// |_____|_____|_____|_|___|  https://github.com/nlohmann/json\n")
-    string(APPEND generated "//\n")
-    string(APPEND generated "// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>\n")
-    string(APPEND generated "// SPDX-License-Identifier: MIT\n")
-    string(APPEND generated "\n")
-    string(APPEND generated "#pragma once\n")
-    string(APPEND generated "\n")
-    foreach(name IN LISTS macro_names)
-        string(APPEND generated "#undef ${name}\n")
-    endforeach()
-else()
-    string(APPEND generated "// This file is generated by cmake/scripts/gen_hedley_undef_check.cmake\n")
-    string(APPEND generated "// from include/nlohmann/thirdparty/hedley/hedley.hpp. Do not edit it by\n")
-    string(APPEND generated "// hand -- it is regenerated on every build. ${macro_count} macros checked.\n\n")
-    foreach(name IN LISTS macro_names)
-        string(APPEND generated "#ifdef ${name}\n")
-        string(APPEND generated "    FAIL_CHECK(\"${name} leaked after including nlohmann/json.hpp\");\n")
-        string(APPEND generated "#endif\n")
-    endforeach()
-endif()
-
-get_filename_component(output_dir "${OUTPUT}" DIRECTORY)
-if(output_dir)
-    file(MAKE_DIRECTORY "${output_dir}")
-endif()
-
-# Avoid rewriting the file (and busting downstream incremental rebuilds)
-# when the content has not actually changed.
-set(write_output TRUE)
-if(EXISTS "${OUTPUT}")
-    file(READ "${OUTPUT}" existing_content)
-    if(existing_content STREQUAL generated)
-        set(write_output FALSE)
-    endif()
-endif()
-if(write_output)
-    file(WRITE "${OUTPUT}" "${generated}")
-endif()

include/nlohmann/detail/input/lexer.hpp

@@ -149,10 +149,11 @@ class lexer : public lexer_base<BasicJsonType>
   public:
     using token_type = typename lexer_base<BasicJsonType>::token_type;
 
-    explicit lexer(InputAdapterType&& adapter, bool ignore_comments_ = false) noexcept
+    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_)
     {}
 
     // deleted because of pointer members
@@ -1279,6 +1280,58 @@ scan_number_done:
         // we are done scanning a number)
         unget();
 
+        // If the caller does not need the converted value (only whether the
+        // input is syntactically valid; see json_sax_acceptor/accept()), an
+        // unsigned/integer token can be reported without calling
+        // strtoull()/strtoll() at all, *provided* we can already tell from
+        // the digit count alone that the conversion cannot overflow 64 bits.
+        // Such tokens are always finite and are accepted unconditionally by
+        // the parser regardless of their actual value (parser::sax_parse_internal()
+        // never checks finiteness for value_unsigned/value_integer), so the
+        // classification below is all that is needed.
+        //
+        // A decimal number with up to 18 digits is always representable in
+        // both std::uint64_t and std::int64_t (18 nines is ~1e18, well below
+        // both UINT64_MAX ~1.8e19 and INT64_MAX ~9.2e18), so strtoull()/strtoll()
+        // could not have set errno to ERANGE for it. Numbers with more digits
+        // (rare in practice) fall through to the exact code below, unchanged,
+        // so their handling -- including reclassification to value_float when
+        // the value overflows 64 bits, and rejection when it is not even
+        // finite as a double -- is bit-for-bit identical to before this
+        // optimization.
+        //
+        // Note this reasons about std::uint64_t/std::int64_t, not about
+        // number_unsigned_t/number_integer_t (BasicJsonType's own, possibly
+        // narrower, template parameters -- e.g. std::uint32_t). That is fine
+        // *only* because discard_number_values is exclusively set by
+        // accept() (see json.hpp), and accept() always parses through the
+        // library's own json_sax_acceptor -- never a user-supplied SAX
+        // consumer -- whose number_unsigned()/number_integer()/number_float()
+        // callbacks unconditionally discard their argument and return true.
+        // So for every caller that can reach this branch, neither the token
+        // classification below nor the eventual (possibly narrowed, and on
+        // this fast path left stale/unset) value_unsigned/value_integer is
+        // ever consulted -- an unsigned/integer token is accepted outright,
+        // and even a >18-digit token that this fast path deliberately falls
+        // through for is, once reclassified to value_float, still finite
+        // (and thus accepted) for any digit count that fits in number_unsigned_t
+        // or number_integer_t regardless of that type's width. If this
+        // function is ever taught to run with discard_number_values true for
+        // a caller that *does* read the converted value, this reasoning (and
+        // the fast path below) would need to be revisited.
+        if (discard_number_values)
+        {
+            constexpr std::size_t safe_digit_count = 18;
+            if (number_type == token_type::value_unsigned && token_buffer.size() <= safe_digit_count)
+            {
+                return token_type::value_unsigned;
+            }
+            if (number_type == token_type::value_integer && token_buffer.size() - 1 <= safe_digit_count)
+            {
+                return token_type::value_integer;
+            }
+        }
+
         char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
         errno = 0;
 
@@ -1393,8 +1446,7 @@ scan_number_done:
     */
     char_int_type get()
     {
-        ++position.chars_read_total;
-        ++position.chars_read_current_line;
+        advance_position();
 
         if (next_unget)
         {
@@ -1406,6 +1458,23 @@ scan_number_done:
             current = ia.get_character();
         }
 
+        return track_after_read();
+    }
+
+    /// shared head of get() / get_ignoring_pending_unget(): bump the
+    /// per-character position counters (line-count-on-'\n' bookkeeping is
+    /// handled afterwards, in track_after_read(), once `current` is known)
+    void advance_position() noexcept
+    {
+        ++position.chars_read_total;
+        ++position.chars_read_current_line;
+    }
+
+    /// shared tail of get() / get_ignoring_pending_unget(): capture the
+    /// character for error messages (if needed) and update line/column
+    /// bookkeeping for the character now in `current`
+    char_int_type track_after_read()
+    {
         // seekable adapters reconstruct the token lazily on error (see
         // get_token_string), so the eager per-character copy is skipped
         capture_char(std::integral_constant<bool, lazy_token_string> {});
@@ -1419,6 +1488,29 @@ scan_number_done:
         return current;
     }
 
+    /*!
+    @brief like get(), but for call sites that can prove no unget() is pending
+
+    get() has to check the `next_unget` flag on every call, because a
+    previous token may have ended with unget() (e.g. scan_number() always
+    ungets the character that terminated the number, so the next call to
+    scan() can see it again). skip_whitespace() reads that first,
+    possibly-ungotten character via a plain get(), but every further
+    character it reads is guaranteed to be a fresh read: nothing between
+    those calls invokes unget(). This variant skips the (otherwise always
+    false) next_unget branch for those calls; it is not a general
+    replacement for get().
+    */
+    char_int_type get_ignoring_pending_unget()
+    {
+        JSON_ASSERT(!next_unget);
+
+        advance_position();
+        current = ia.get_character();
+
+        return track_after_read();
+    }
+
     /// seekable adapter: nothing to capture, the token is rebuilt on error
     void capture_char(std::true_type /*lazy*/) const noexcept {}
 
@@ -1612,13 +1704,37 @@ scan_number_done:
         return true;
     }
 
+    /// whether `current` is one of the four JSON whitespace characters
+    bool current_is_whitespace() const noexcept
+    {
+        return current == ' ' || current == '\t' || current == '\n' || current == '\r';
+    }
+
     void skip_whitespace()
     {
+        // the first character may be a pending unget() left over from the
+        // previous token (see get_ignoring_pending_unget()); every
+        // subsequent character read by this loop is guaranteed fresh, since
+        // nothing below calls unget()
+        get();
+
+        if (!current_is_whitespace())
+        {
+            return;
+        }
+
+        // this is written as an if-guarded do-while (rather than a plain
+        // while loop) because that shape is what lets both GCC and Clang
+        // keep the input adapter's read pointer in a register across
+        // iterations; the equivalent while-loop measurably defeated that
+        // optimization in testing, turning long whitespace runs (e.g. the
+        // indentation of pretty-printed JSON) from a register-only loop
+        // into one that reloads the pointer from memory every character
         do
         {
-            get();
+            get_ignoring_pending_unget();
         }
-        while (current == ' ' || current == '\t' || current == '\n' || current == '\r');
+        while (current_is_whitespace());
     }
 
     token_type scan()
@@ -1754,6 +1870,13 @@ scan_number_done:
     const char_int_type decimal_point_char = '.';
     /// the position of the decimal point in the input
     std::size_t decimal_point_position = std::string::npos;
+
+    /// whether the caller (e.g. accept()/json_sax_acceptor) only needs the
+    /// token classification and never looks at the converted numeric value;
+    /// when set, scan_number() may skip strtoull()/strtoll() for
+    /// value_unsigned/value_integer tokens whose digit count guarantees they
+    /// fit into 64 bits (see scan_number())
+    const bool discard_number_values = false;
 };
 
 }  // namespace detail

include/nlohmann/detail/input/parser.hpp

@@ -72,9 +72,10 @@ class parser
                     parser_callback_t<BasicJsonType> cb = nullptr,
                     const bool allow_exceptions_ = true,
                     const bool ignore_comments = false,
-                    const bool ignore_trailing_commas_ = false)
+                    const bool ignore_trailing_commas_ = false,
+                    const bool discard_number_values_ = false)
         : callback(std::move(cb))
-        , m_lexer(std::move(adapter), ignore_comments)
+        , m_lexer(std::move(adapter), ignore_comments, discard_number_values_)
         , allow_exceptions(allow_exceptions_)
         , ignore_trailing_commas(ignore_trailing_commas_)
     {

include/nlohmann/json.hpp

@@ -164,11 +164,12 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         detail::parser_callback_t<basic_json>cb = nullptr,
         const bool allow_exceptions = true,
         const bool ignore_comments = false,
-        const bool ignore_trailing_commas = false
+        const bool ignore_trailing_commas = false,
+        const bool discard_number_values = false
     )
     {
         return ::nlohmann::detail::parser<basic_json, InputAdapterType>(std::move(adapter),
-               std::move(cb), allow_exceptions, ignore_comments, ignore_trailing_commas);
+               std::move(cb), allow_exceptions, ignore_comments, ignore_trailing_commas, discard_number_values);
     }
 
   private:
@@ -4133,7 +4134,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                        const bool ignore_comments = false,
                        const bool ignore_trailing_commas = false)
     {
-        return parser(detail::input_adapter(std::forward<InputType>(i)), nullptr, false, ignore_comments, ignore_trailing_commas).accept(true);
+        return parser(detail::input_adapter(std::forward<InputType>(i)), nullptr, false, ignore_comments, ignore_trailing_commas, true).accept(true);
     }
 
     /// @brief check if the input is valid JSON (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -4144,7 +4145,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                        const bool ignore_comments = false,
                        const bool ignore_trailing_commas = false)
     {
-        return parser(detail::input_adapter(std::move(first), std::move(last)), nullptr, false, ignore_comments, ignore_trailing_commas).accept(true);
+        return parser(detail::input_adapter(std::move(first), std::move(last)), nullptr, false, ignore_comments, ignore_trailing_commas, true).accept(true);
     }
 
     JSON_HEDLEY_WARN_UNUSED_RESULT
@@ -4153,7 +4154,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                        const bool ignore_comments = false,
                        const bool ignore_trailing_commas = false)
     {
-        return parser(i.get(), nullptr, false, ignore_comments, ignore_trailing_commas).accept(true);
+        return parser(i.get(), nullptr, false, ignore_comments, ignore_trailing_commas, true).accept(true);
     }
 
     /// @brief generate SAX events

include/nlohmann/thirdparty/hedley/hedley_undef.hpp

@@ -17,7 +17,7 @@
 #undef JSON_HEDLEY_CLANG_HAS_ATTRIBUTE
 #undef JSON_HEDLEY_CLANG_HAS_BUILTIN
 #undef JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE
-#undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_ATTRIBUTE
+#undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE
 #undef JSON_HEDLEY_CLANG_HAS_EXTENSION
 #undef JSON_HEDLEY_CLANG_HAS_FEATURE
 #undef JSON_HEDLEY_CLANG_HAS_WARNING
@@ -108,10 +108,7 @@
 #undef JSON_HEDLEY_PELLES_VERSION_CHECK
 #undef JSON_HEDLEY_PGI_VERSION
 #undef JSON_HEDLEY_PGI_VERSION_CHECK
-#undef JSON_HEDLEY_PRAGMA
 #undef JSON_HEDLEY_PREDICT
-#undef JSON_HEDLEY_PREDICT_FALSE
-#undef JSON_HEDLEY_PREDICT_TRUE
 #undef JSON_HEDLEY_PRINTF_FORMAT
 #undef JSON_HEDLEY_PRIVATE
 #undef JSON_HEDLEY_PUBLIC

single_include/nlohmann/json.hpp

@@ -7932,10 +7932,11 @@ class lexer : public lexer_base<BasicJsonType>
   public:
     using token_type = typename lexer_base<BasicJsonType>::token_type;
 
-    explicit lexer(InputAdapterType&& adapter, bool ignore_comments_ = false) noexcept
+    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_)
     {}
 
     // deleted because of pointer members
@@ -9062,6 +9063,58 @@ scan_number_done:
         // we are done scanning a number)
         unget();
 
+        // If the caller does not need the converted value (only whether the
+        // input is syntactically valid; see json_sax_acceptor/accept()), an
+        // unsigned/integer token can be reported without calling
+        // strtoull()/strtoll() at all, *provided* we can already tell from
+        // the digit count alone that the conversion cannot overflow 64 bits.
+        // Such tokens are always finite and are accepted unconditionally by
+        // the parser regardless of their actual value (parser::sax_parse_internal()
+        // never checks finiteness for value_unsigned/value_integer), so the
+        // classification below is all that is needed.
+        //
+        // A decimal number with up to 18 digits is always representable in
+        // both std::uint64_t and std::int64_t (18 nines is ~1e18, well below
+        // both UINT64_MAX ~1.8e19 and INT64_MAX ~9.2e18), so strtoull()/strtoll()
+        // could not have set errno to ERANGE for it. Numbers with more digits
+        // (rare in practice) fall through to the exact code below, unchanged,
+        // so their handling -- including reclassification to value_float when
+        // the value overflows 64 bits, and rejection when it is not even
+        // finite as a double -- is bit-for-bit identical to before this
+        // optimization.
+        //
+        // Note this reasons about std::uint64_t/std::int64_t, not about
+        // number_unsigned_t/number_integer_t (BasicJsonType's own, possibly
+        // narrower, template parameters -- e.g. std::uint32_t). That is fine
+        // *only* because discard_number_values is exclusively set by
+        // accept() (see json.hpp), and accept() always parses through the
+        // library's own json_sax_acceptor -- never a user-supplied SAX
+        // consumer -- whose number_unsigned()/number_integer()/number_float()
+        // callbacks unconditionally discard their argument and return true.
+        // So for every caller that can reach this branch, neither the token
+        // classification below nor the eventual (possibly narrowed, and on
+        // this fast path left stale/unset) value_unsigned/value_integer is
+        // ever consulted -- an unsigned/integer token is accepted outright,
+        // and even a >18-digit token that this fast path deliberately falls
+        // through for is, once reclassified to value_float, still finite
+        // (and thus accepted) for any digit count that fits in number_unsigned_t
+        // or number_integer_t regardless of that type's width. If this
+        // function is ever taught to run with discard_number_values true for
+        // a caller that *does* read the converted value, this reasoning (and
+        // the fast path below) would need to be revisited.
+        if (discard_number_values)
+        {
+            constexpr std::size_t safe_digit_count = 18;
+            if (number_type == token_type::value_unsigned && token_buffer.size() <= safe_digit_count)
+            {
+                return token_type::value_unsigned;
+            }
+            if (number_type == token_type::value_integer && token_buffer.size() - 1 <= safe_digit_count)
+            {
+                return token_type::value_integer;
+            }
+        }
+
         char* endptr = nullptr; // NOLINT(misc-const-correctness,cppcoreguidelines-pro-type-vararg,hicpp-vararg)
         errno = 0;
 
@@ -9176,8 +9229,7 @@ scan_number_done:
     */
     char_int_type get()
     {
-        ++position.chars_read_total;
-        ++position.chars_read_current_line;
+        advance_position();
 
         if (next_unget)
         {
@@ -9189,6 +9241,23 @@ scan_number_done:
             current = ia.get_character();
         }
 
+        return track_after_read();
+    }
+
+    /// shared head of get() / get_ignoring_pending_unget(): bump the
+    /// per-character position counters (line-count-on-'\n' bookkeeping is
+    /// handled afterwards, in track_after_read(), once `current` is known)
+    void advance_position() noexcept
+    {
+        ++position.chars_read_total;
+        ++position.chars_read_current_line;
+    }
+
+    /// shared tail of get() / get_ignoring_pending_unget(): capture the
+    /// character for error messages (if needed) and update line/column
+    /// bookkeeping for the character now in `current`
+    char_int_type track_after_read()
+    {
         // seekable adapters reconstruct the token lazily on error (see
         // get_token_string), so the eager per-character copy is skipped
         capture_char(std::integral_constant<bool, lazy_token_string> {});
@@ -9202,6 +9271,29 @@ scan_number_done:
         return current;
     }
 
+    /*!
+    @brief like get(), but for call sites that can prove no unget() is pending
+
+    get() has to check the `next_unget` flag on every call, because a
+    previous token may have ended with unget() (e.g. scan_number() always
+    ungets the character that terminated the number, so the next call to
+    scan() can see it again). skip_whitespace() reads that first,
+    possibly-ungotten character via a plain get(), but every further
+    character it reads is guaranteed to be a fresh read: nothing between
+    those calls invokes unget(). This variant skips the (otherwise always
+    false) next_unget branch for those calls; it is not a general
+    replacement for get().
+    */
+    char_int_type get_ignoring_pending_unget()
+    {
+        JSON_ASSERT(!next_unget);
+
+        advance_position();
+        current = ia.get_character();
+
+        return track_after_read();
+    }
+
     /// seekable adapter: nothing to capture, the token is rebuilt on error
     void capture_char(std::true_type /*lazy*/) const noexcept {}
 
@@ -9395,13 +9487,37 @@ scan_number_done:
         return true;
     }
 
+    /// whether `current` is one of the four JSON whitespace characters
+    bool current_is_whitespace() const noexcept
+    {
+        return current == ' ' || current == '\t' || current == '\n' || current == '\r';
+    }
+
     void skip_whitespace()
     {
+        // the first character may be a pending unget() left over from the
+        // previous token (see get_ignoring_pending_unget()); every
+        // subsequent character read by this loop is guaranteed fresh, since
+        // nothing below calls unget()
+        get();
+
+        if (!current_is_whitespace())
+        {
+            return;
+        }
+
+        // this is written as an if-guarded do-while (rather than a plain
+        // while loop) because that shape is what lets both GCC and Clang
+        // keep the input adapter's read pointer in a register across
+        // iterations; the equivalent while-loop measurably defeated that
+        // optimization in testing, turning long whitespace runs (e.g. the
+        // indentation of pretty-printed JSON) from a register-only loop
+        // into one that reloads the pointer from memory every character
         do
         {
-            get();
+            get_ignoring_pending_unget();
         }
-        while (current == ' ' || current == '\t' || current == '\n' || current == '\r');
+        while (current_is_whitespace());
     }
 
     token_type scan()
@@ -9537,6 +9653,13 @@ scan_number_done:
     const char_int_type decimal_point_char = '.';
     /// the position of the decimal point in the input
     std::size_t decimal_point_position = std::string::npos;
+
+    /// whether the caller (e.g. accept()/json_sax_acceptor) only needs the
+    /// token classification and never looks at the converted numeric value;
+    /// when set, scan_number() may skip strtoull()/strtoll() for
+    /// value_unsigned/value_integer tokens whose digit count guarantees they
+    /// fit into 64 bits (see scan_number())
+    const bool discard_number_values = false;
 };
 
 }  // namespace detail
@@ -14043,9 +14166,10 @@ class parser
                     parser_callback_t<BasicJsonType> cb = nullptr,
                     const bool allow_exceptions_ = true,
                     const bool ignore_comments = false,
-                    const bool ignore_trailing_commas_ = false)
+                    const bool ignore_trailing_commas_ = false,
+                    const bool discard_number_values_ = false)
         : callback(std::move(cb))
-        , m_lexer(std::move(adapter), ignore_comments)
+        , m_lexer(std::move(adapter), ignore_comments, discard_number_values_)
         , allow_exceptions(allow_exceptions_)
         , ignore_trailing_commas(ignore_trailing_commas_)
     {
@@ -21592,11 +21716,12 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
         detail::parser_callback_t<basic_json>cb = nullptr,
         const bool allow_exceptions = true,
         const bool ignore_comments = false,
-        const bool ignore_trailing_commas = false
+        const bool ignore_trailing_commas = false,
+        const bool discard_number_values = false
                                  )
     {
         return ::nlohmann::detail::parser<basic_json, InputAdapterType>(std::move(adapter),
-            std::move(cb), allow_exceptions, ignore_comments, ignore_trailing_commas);
+            std::move(cb), allow_exceptions, ignore_comments, ignore_trailing_commas, discard_number_values);
     }
 
   private:
@@ -25561,7 +25686,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                        const bool ignore_comments = false,
                        const bool ignore_trailing_commas = false)
     {
-        return parser(detail::input_adapter(std::forward<InputType>(i)), nullptr, false, ignore_comments, ignore_trailing_commas).accept(true);
+        return parser(detail::input_adapter(std::forward<InputType>(i)), nullptr, false, ignore_comments, ignore_trailing_commas, true).accept(true);
     }
 
     /// @brief check if the input is valid JSON (iterator pair, or iterator+sentinel pair for C++20 ranges support)
@@ -25572,7 +25697,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                        const bool ignore_comments = false,
                        const bool ignore_trailing_commas = false)
     {
-        return parser(detail::input_adapter(std::move(first), std::move(last)), nullptr, false, ignore_comments, ignore_trailing_commas).accept(true);
+        return parser(detail::input_adapter(std::move(first), std::move(last)), nullptr, false, ignore_comments, ignore_trailing_commas, true).accept(true);
     }
 
     JSON_HEDLEY_WARN_UNUSED_RESULT
@@ -25581,7 +25706,7 @@ class basic_json // NOLINT(cppcoreguidelines-special-member-functions,hicpp-spec
                        const bool ignore_comments = false,
                        const bool ignore_trailing_commas = false)
     {
-        return parser(i.get(), nullptr, false, ignore_comments, ignore_trailing_commas).accept(true);
+        return parser(i.get(), nullptr, false, ignore_comments, ignore_trailing_commas, true).accept(true);
     }
 
     /// @brief generate SAX events
@@ -26965,7 +27090,7 @@ struct formatter<nlohmann::NLOHMANN_BASIC_JSON_TPL, char> // NOLINT(cert-dcl58-c
 #undef JSON_HEDLEY_CLANG_HAS_ATTRIBUTE
 #undef JSON_HEDLEY_CLANG_HAS_BUILTIN
 #undef JSON_HEDLEY_CLANG_HAS_CPP_ATTRIBUTE
-#undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_ATTRIBUTE
+#undef JSON_HEDLEY_CLANG_HAS_DECLSPEC_DECLSPEC_ATTRIBUTE
 #undef JSON_HEDLEY_CLANG_HAS_EXTENSION
 #undef JSON_HEDLEY_CLANG_HAS_FEATURE
 #undef JSON_HEDLEY_CLANG_HAS_WARNING
@@ -27056,10 +27181,7 @@ struct formatter<nlohmann::NLOHMANN_BASIC_JSON_TPL, char> // NOLINT(cert-dcl58-c
 #undef JSON_HEDLEY_PELLES_VERSION_CHECK
 #undef JSON_HEDLEY_PGI_VERSION
 #undef JSON_HEDLEY_PGI_VERSION_CHECK
-#undef JSON_HEDLEY_PRAGMA
 #undef JSON_HEDLEY_PREDICT
-#undef JSON_HEDLEY_PREDICT_FALSE
-#undef JSON_HEDLEY_PREDICT_TRUE
 #undef JSON_HEDLEY_PRINTF_FORMAT
 #undef JSON_HEDLEY_PRIVATE
 #undef JSON_HEDLEY_PUBLIC

tests/CMakeLists.txt

@@ -125,51 +125,6 @@ json_test_set_test_options(test-unicode4 TEST_PROPERTIES TIMEOUT 3000)
 # add unit tests
 #############################################################################
 
-# Generate the leak checks for every JSON_HEDLEY_* macro defined in
-# hedley.hpp; tests/src/unit-no-macro-leak.cpp #include-s the result after
-# nlohmann/json.hpp (see issue #5408). Using the shared
-# cmake/scripts/gen_hedley_undef_check.cmake script (also used by `make
-# update_hedley_undef`) instead of a hand-maintained list of macro names
-# means this test can never go stale after a future `make update_hedley`.
-set(hedley_hpp "${PROJECT_SOURCE_DIR}/include/nlohmann/thirdparty/hedley/hedley.hpp")
-set(hedley_undef_check_script "${PROJECT_SOURCE_DIR}/cmake/scripts/gen_hedley_undef_check.cmake")
-set(hedley_undef_checks "${PROJECT_BINARY_DIR}/include/hedley_undef_checks.inc")
-
-# Reconfigure whenever the vendored header or the generator script changes,
-# so a `cmake --build` after `make update_hedley` does not silently keep a
-# stale generated file around.
-set_property(DIRECTORY APPEND PROPERTY CMAKE_CONFIGURE_DEPENDS
-    "${hedley_hpp}"
-    "${hedley_undef_check_script}")
-
-# Generate once at configure time, so the very first build (before any
-# custom-command build step has run) already has an up-to-date file.
-execute_process(
-    COMMAND ${CMAKE_COMMAND}
-        "-DHEDLEY_HPP=${hedley_hpp}"
-        "-DOUTPUT=${hedley_undef_checks}"
-        -DMODE=checks
-        -P "${hedley_undef_check_script}"
-    RESULT_VARIABLE hedley_undef_check_result
-)
-if(NOT hedley_undef_check_result EQUAL 0)
-    message(FATAL_ERROR "Failed to generate ${hedley_undef_checks}")
-endif()
-
-# Also (re)generate as a build step, so an incremental build after editing
-# hedley.hpp without a full reconfigure still picks up the change.
-add_custom_command(
-    OUTPUT "${hedley_undef_checks}"
-    COMMAND ${CMAKE_COMMAND}
-        "-DHEDLEY_HPP=${hedley_hpp}"
-        "-DOUTPUT=${hedley_undef_checks}"
-        -DMODE=checks
-        -P "${hedley_undef_check_script}"
-    DEPENDS "${hedley_hpp}" "${hedley_undef_check_script}"
-    COMMENT "Generating Hedley undef leak checks"
-    VERBATIM)
-add_custom_target(generate_hedley_undef_checks DEPENDS "${hedley_undef_checks}")
-
 if("${JSON_TestStandards}" STREQUAL "")
     set(test_cxx_standards 11 14 17 20 23)
     unset(test_force)
@@ -208,14 +163,6 @@ foreach(file ${files})
     json_test_add_test_for(${file} MAIN test_main CXX_STANDARDS ${test_cxx_standards} ${test_force})
 endforeach()
 
-# tests/src/unit-no-macro-leak.cpp #include-s the generated leak-check file,
-# so its test targets must be built after generate_hedley_undef_checks.
-foreach(cxx_standard ${test_cxx_standards})
-    if(TARGET test-no-macro-leak_cpp${cxx_standard})
-        add_dependencies(test-no-macro-leak_cpp${cxx_standard} generate_hedley_undef_checks)
-    endif()
-endforeach()
-
 if(json_32bit_test_only)
     # Skip all other tests in this file
     return()

tests/src/unit-class_parser.cpp

@@ -930,6 +930,94 @@ TEST_CASE("parser class")
                 CHECK(accept_helper("+1") == false);
                 CHECK(accept_helper("+0") == false);
             }
+
+            SECTION("issue #5411 - skip conversion when accept() does not need the numeric value")
+            {
+                // lexer::scan_number() may skip strtoull()/strtoll() for
+                // value_unsigned/value_integer tokens when the caller (e.g.
+                // json::accept()) does not need the converted value, as long
+                // as the digit count alone guarantees no 64-bit overflow (see
+                // the "safe_digit_count" fast path in scan_number()). This
+                // differential test checks that json::accept() (which enables
+                // the fast path) and json::parse() (which never does) always
+                // agree, over a corpus that exercises both the fast path
+                // (<=18 digits) and the untouched, exact fallback path (>=19
+                // digits) -- including reclassification of huge digit-only
+                // integers to a (possibly non-finite) floating-point value.
+                const std::vector<std::pair<std::string, bool>> cases =
+                {
+                    // normal small/large integers, both signs
+                    {"0", true}, {"1", true}, {"-1", true}, {"42", true}, {"-42", true},
+                    {"123456789", true}, {"-123456789", true},
+
+                    // digit-count boundary around the 18-digit safe cutoff (both signs)
+                    {std::string(17, '9'), true},
+                    {std::string(18, '9'), true},
+                    {std::string(19, '9'), true},
+                    {std::string(20, '9'), true},
+                    {"-" + std::string(17, '9'), true},
+                    {"-" + std::string(18, '9'), true},
+                    {"-" + std::string(19, '9'), true},
+                    {"-" + std::string(20, '9'), true},
+
+                    // 64-bit boundaries
+                    {"9223372036854775807", true},    // INT64_MAX
+                    {"-9223372036854775808", true},   // INT64_MIN
+                    {"18446744073709551615", true},   // UINT64_MAX
+                    {"18446744073709551616", true},   // UINT64_MAX + 1 (overflows uint64_t, finite double)
+
+                    // the 28-digit example from the issue: overflows uint64_t
+                    // but is finite as a double, so the scanner reclassifies
+                    // it to value_float and it is accepted
+                    {"9999999999999999999999999999", true},
+
+                    // huge digit-only integers that overflow even a double -> rejected
+                    {std::string(309, '9'), false},
+                    {std::string(400, '9'), false},
+                    {"1" + std::string(400, '0'), false},
+
+                    // 1e999 / 1e400 style overflow -> rejected
+                    {"1e999", false},
+                    {"1e400", false},
+                    {"-1e999", false},
+                    {"1E999", false},
+
+                    // values straddling DBL_MAX
+                    {"1.7976931348623157e308", true},   // <= DBL_MAX, finite
+                    {"1.7976931348623159e308", false},  // > DBL_MAX, overflows to inf
+
+                    // a mix of other valid/invalid numeric syntax
+                    {"3.14159", true},
+                    {"-0.0", true},
+                    {"1.0e10", true},
+                    {"01", false},
+                    {"-", false},
+                    {"1.", false},
+                    {"1e", false},
+                    {"+1", false},
+                };
+
+                for (const auto& c : cases)
+                {
+                    const std::string& number = c.first;
+                    const bool expected = c.second;
+                    CAPTURE(number)
+                    CAPTURE(expected)
+
+                    // accept() takes the fast path (skips conversion when possible)
+                    CHECK(json::accept(number) == expected);
+
+                    // parse() always performs the full conversion; it must agree
+                    json j;
+                    CHECK_NOTHROW(json::parser(nlohmann::detail::input_adapter(number), nullptr, false).parse(true, j));
+                    CHECK(!j.is_discarded() == expected);
+
+                    // wrap in an array so get_token() is exercised beyond the
+                    // very first (constructor-time) scan as well
+                    const std::string wrapped = "[" + number + "," + number + "]";
+                    CHECK(json::accept(wrapped) == expected);
+                }
+            }
         }
     }
 
@@ -1394,6 +1482,62 @@ TEST_CASE("parser class")
         CHECK(accept_helper("\"\\uD80C\\uFFFF\"") == false);
     }
 
+    SECTION("issue #5412 - whitespace skipping bookkeeping (compact vs. pretty-printed)")
+    {
+        // lexer::skip_whitespace() reads its first character with get() (to
+        // honor a possibly pending unget() from the previous token) and every
+        // further whitespace character with get_ignoring_pending_unget() (a
+        // get() variant that skips the then-always-false next_unget check).
+        // This must not change the reported byte offset, line, or column of
+        // a syntax error, even when a long run of whitespace containing
+        // multiple newlines is skipped beforehand (as with pretty-printed
+        // input). The expected values below were captured from the
+        // unmodified do-while(get()) loop, so any regression that miscounts
+        // characters or newlines while skipping whitespace changes them.
+        const auto check_error = [](const std::string & input, std::size_t expected_byte,
+                                    const std::string & expected_what)
+        {
+            CAPTURE(input)
+            try
+            {
+                json _ = json::parse(input);
+                FAIL_CHECK("expected a parse_error, but parsing succeeded");
+            }
+            catch (const json::parse_error& e)
+            {
+                CHECK(e.byte == expected_byte);
+                CHECK(std::string(e.what()) == expected_what);
+            }
+        };
+
+        // a nested document, serialized both compactly and pretty-printed
+        // (dump(4)), each truncated right before the final closing '}' so
+        // that the parser hits EOF after skipping all of the (in the
+        // pretty-printed case, substantial) indentation whitespace
+        const json doc =
+        {
+            {"a", 1},
+            {"b", json::array({true, false, nullptr, "x"})},
+            {"c", json::object({{"d", 3.14}, {"e", json::array({1, 2, 3})}})}
+        };
+
+        const std::string compact = doc.dump();
+        const std::string pretty = doc.dump(4);
+
+        check_error(compact.substr(0, compact.size() - 1), 60,
+                    "[json.exception.parse_error.101] parse error at line 1, column 60: syntax error while parsing object - unexpected end of input; expected '}'");
+        check_error(pretty.substr(0, pretty.size() - 1), 193,
+                    "[json.exception.parse_error.101] parse error at line 17, column 1: syntax error while parsing object - unexpected end of input; expected '}'");
+
+        // an invalid token appearing after several indented, multi-line
+        // whitespace runs vs. the same document without any of that
+        // whitespace
+        check_error("{\n    \"a\": 1,\n    \"b\": [\n        true,\n        false\n    ],\n    \"c\": @\n}", 70,
+                    "[json.exception.parse_error.101] parse error at line 7, column 10: syntax error while parsing value - invalid literal; last read: '\"c\": @'");
+        check_error("{\"a\":1,\"b\":[true,false],\"c\":@}", 29,
+                    "[json.exception.parse_error.101] parse error at line 1, column 29: syntax error while parsing value - invalid literal; last read: '\"c\":@'");
+    }
+
     SECTION("tests found by mutate++")
     {
         // test case to make sure no comma precedes the first key

tests/src/unit-no-macro-leak.cpp

@@ -1,34 +0,0 @@
-//     __ _____ _____ _____
-//  __|  |   __|     |   | |  JSON for Modern C++
-// |  |  |__   |  |  | | | |  version 3.12.0
-// |_____|_____|_____|_|___|  https://github.com/nlohmann/json
-//
-// SPDX-FileCopyrightText: 2013-2026 Niels Lohmann <https://nlohmann.me>
-// SPDX-License-Identifier: MIT
-
-// This file makes sure that none of the internal JSON_HEDLEY_* macros (vendored
-// from https://nemequ.github.io/hedley/, see
-// include/nlohmann/thirdparty/hedley/hedley.hpp) leak into the including
-// translation unit. include/nlohmann/detail/macro_unscope.hpp is supposed to
-// #undef every JSON_HEDLEY_* macro (via hedley_undef.hpp) once json.hpp has
-// been fully processed. See https://github.com/nlohmann/json/issues/5408,
-// where JSON_HEDLEY_PRAGMA, JSON_HEDLEY_PREDICT_TRUE, JSON_HEDLEY_PREDICT_FALSE,
-// and JSON_HEDLEY_CLANG_HAS_DECLSPEC_ATTRIBUTE escaped this cleanup because
-// hedley_undef.hpp had no matching #undef for them.
-//
-// hedley_undef_checks.inc (included below) is generated at CMake configure/
-// build time by cmake/scripts/gen_hedley_undef_check.cmake, which derives the
-// full list of JSON_HEDLEY_* macro names directly from hedley.hpp. That way
-// this test covers every macro Hedley actually defines -- not a hardcoded
-// snapshot that would silently go stale the next time `make update_hedley`
-// runs -- and can never drift from the vendored header.
-
-#include "doctest_compatibility.h"
-
-#include <nlohmann/json.hpp>
-
-TEST_CASE("JSON_HEDLEY macros do not leak after including json.hpp")
-{
-#include "hedley_undef_checks.inc"
-    CHECK(true); // keep an assertion when nothing leaked
-}