| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
Support custom object/array types and improve template parameter handling (#5443) * docs: document the implicit requirements on basic_json's template parameters The requirements that basic_json places on its eleven template parameters were only implied by how the library uses the resulting object_t, array_t, string_t, etc. Consumers had to discover them by trial and error. Add "Template Parameter Requirements" collecting them, split into what is always required and what is only required when a particular part of the API is instantiated. Notable findings that were previously undocumented: - ObjectType must provide a key_compare member type (actual_object_comparator names object_t::key_compare in both arms of a std::conditional), and its third template parameter is used as a comparator, so std::unordered_map cannot be used without a wrapper. - ArrayType must provide capacity() -- push_back(), emplace_back(), operator+=(), and operator[](size_type) call it unconditionally -- and needs random-access iterators, so std::deque and std::list do not work. - StringType needs contiguous, null-terminated data(), a one-byte value_type, and either assignability from std::to_string or an ADL int_to_string(). - NumberFloatType must be float, double, or long double for parsing and serialization; the integer types must satisfy std::is_integral. - AllocatorType must be stateless, support incomplete types, and use plain pointers. - BooleanType and the number types are union members and must be trivial. Link the new page from the basic_json overview, the types feature page, and the individual type alias pages, and correct the container examples given for ObjectType (std::unordered_map) and ArrayType (std::list), which do not work. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix object_comparator_t for object types without key_compare detail::actual_object_comparator selected between object_t::key_compare and default_object_comparator_t with std::conditional. Both type arguments of std::conditional are named eagerly, so object_t::key_compare had to exist regardless of the condition, and the has_key_compare guard added in 3.11.0 never took effect: any ObjectType without a key_compare member type failed to compile while instantiating basic_json itself. Use detected_or_t instead, which resolves through a SFINAE partial specialization and only names object_t::key_compare when it exists. The selected type is unchanged for every object type that compiled before, so object_comparator_t -- a public member type -- keeps its meaning and ABI. has_key_compare had no other users and is removed. Add a regression test using an adapter around std::unordered_map, which has no key_compare; it fails to compile without this change. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: list the types that are known to work for each template parameter Follow up on the template parameter requirements page: state, for every template parameter, which concrete types work and where they stop working. Each entry was verified by compiling and running a common workload (DOM access, dump, parse, CBOR/MessagePack round-trip, flatten, hash) against that instantiation. Findings worth calling out: - ObjectType no longer needs a key_compare member type, so the std::unordered_map adapter only has to restore the template argument order. A hash-ordered ObjectType works everywhere except unflatten(), which reconstructs an array only when it meets the reference token 0 before the other indices. - ArrayType: std::deque works when wrapped to add capacity(); std::list does not. - StringType: std::pmr::string and std::basic_string with a custom allocator compile for the DOM, dump, and parse, but not for the binary readers, flatten, or diff, because the library assigns std::string values to string_t and int_to_string cannot be overloaded for a type in namespace std. - NumberFloatType: long double works for dump and parse but not for the binary formats, which have no encoding for it. - BinaryType: std::vector<std::byte> supports assignment, get, and the binary formats, but neither dump nor std::hash<basic_json>. Also record the object_comparator_t fix in its version history. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix unflatten and binary dumping for non-default configurations unflatten() decided between array and object by looking at the first reference token it happened to see for a node: it started an array only when that token was 0. With a sorted object type the token 0 always arrives first, so the result was correct by accident; with an object type whose iteration order is unspecified, {"/c/2":3,"/c/1":2,"/c/0":1} unflattened to an object with the keys "0", "1", and "2" instead of an array. Collect the pointer prefixes that have a reference token 0 among their children before building the result, and let get_and_create() consult that set. The outcome is now independent of the iteration order and matches, for every input, what a sorted object type produced before: a value is restored as an array if and only if one of its keys is 0. Iterating the flattened object in a different order would have been simpler, but it would have changed the key order of the result for insertion-ordered object types. The serializer, std::hash, and the UBJSON writer converted the elements of a binary value to an integer implicitly, which does not compile for a BinaryType whose value type is std::byte, and which made dump() write the bytes of a signed value type as negative numbers. Convert to std::uint8_t explicitly in all three places, so every byte type dumps as 0..255. The default std::vector<std::uint8_t> configuration is unaffected. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: note which Abseil containers can be used as template arguments Checked against Abseil release 20250127.0 with the same workload as the other entries on the page (DOM access, dump, parse, CBOR/MessagePack/UBJSON round-trip, flatten, hash), with and without JSON_DIAGNOSTICS. absl::flat_hash_map and absl::node_hash_map work as ObjectType through an adapter that restores the template argument order and makes erase(iterator) return the following iterator, which Abseil's returns as void. The page now carries that adapter, and notes that absl::flat_hash_map does not keep references to the mapped values valid across insertions while absl::node_hash_map does. Both have a capacity() member, so JSON_DIAGNOSTICS already refreshes the parent pointers conservatively for them. absl::btree_map and absl::InlinedVector cannot be used at all: object_t and array_t are formed while basic_json is still incomplete, and both inspect their value type at class scope. std::map and std::vector are required by the standard to tolerate this, third-party containers generally are not, so the page states the constraint on its own rather than only per container. absl::InlinedVector does work as BinaryType, where it is instantiated with a complete type. absl::FixedArray and absl::Cord are not usable. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Relax the ArrayType and ObjectType requirements Two requirements forced users of otherwise suitable containers to write a wrapper, and neither was load-bearing. array_t::capacity() was read in push_back(), emplace_back(), operator+=(), and operator[](size_type), but set_parent() only looks at the value under JSON_DIAGNOSTICS; without diagnostics it was computed and discarded. Read it through array_capacity(), which reports unknown_size() when diagnostics are off or when the array type has no capacity() at all, and treat an unknown capacity as "the elements may have moved" so the parent pointers are refreshed conservatively. std::deque now works as ArrayType, in both builds, and capacity() is no longer named at all in a default build. Since the capacity is now only meaningful for array insertions, it moves out of set_parent() into set_parent_after_array_insert(). basic_json::erase(iterator) assigned the object's erase() return value, which requires the container to return the following iterator. Abseil's hash maps return void to avoid computing a successor the caller may not need. Detect that and compute the successor before erasing; containers that return an iterator, including the vector-backed ordered_map where a precomputed successor would be wrong, keep the existing path. Together these leave an Abseil hash map needing only an alias that restores the template argument order, and no adapter at all for std::deque. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not require string_t to be convertible from std::string Three places built a std::string and handed it to something expecting a string_t: the UBJSON high-precision number reader, which every binary reader instantiates, and the BSON writer's array element size calculation and write. That silently required string_t to be implicitly convertible from std::string, which std::string itself and types with a string_view conversion satisfy, but many string types do not. Construct the string_t explicitly from the data and size, which the requirements already cover. This makes boost::container::string, eastl::string, std::pmr::string, and std::basic_string with a custom allocator work as StringType, none of which could previously be used with any binary format. Add binary format coverage to the alt_string test, which had none, including a UBJSON high-precision number -- the case that goes through the reader path. BSON stays uncovered there: it additionally needs string_t::find(value_type), which alt_string does not provide. Also record which containers from Boost, Abseil, and EASTL work for each template parameter, and correct two claims: std::pmr::string is usable after this change, and tsl::ordered_map is not usable at all, because its iterators expose the mapped value as const while basic_json modifies it in place. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: record compatibility for the common header-only hash maps ankerl::unordered_dense (map and segmented_map), phmap (flat_hash_map and node_hash_map), and robin_hood::unordered_flat_map all work as ObjectType through the same adapter as Abseil's and Boost's hash maps, which only has to restore the template argument order. phmap::btree_map and robin_hood::unordered_node_map do not: like the other btree containers they require a complete value type. Note that none of these hash maps defines key_compare, so every one of them depends on object_comparator_t falling back to default_object_comparator_t. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: record Folly and the remaining vector replacements Folly works, with the caveat that its headers need C++20: folly::fbstring as StringType, folly::fbvector and folly::small_vector as ArrayType, folly::fbvector<std::uint8_t> as BinaryType, and folly::F14NodeMap as ObjectType through the usual argument-order adapter. folly::F14FastMap is the exception and requires a complete value type. For ArrayType, boost::container::devector, boost::container::static_vector (within its fixed capacity), and std::pmr::vector work as well. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: cover fifo_map, gtl, folly::sorted_vector_map, and Qt nlohmann::fifo_map works through the adapter that has always been documented for it, and preserves the insertion order. Restore its mention in the object order page, which was dropped together with the tsl::ordered_map one: unlike ordered_map it keeps a lookup index, so it is the insertion-ordered option without the quadratic cost. gtl::flat_hash_map and folly::sorted_vector_map work as well, the latter through an alias that drops the allocator, whose value type it disagrees on. gtl::btree_map does not, for the same reason as the other btree containers. None of the Qt containers can be used, each for its own reason: QMap has no value_type, QHash iterators yield the mapped value rather than a pair, QList has no max_size(), QByteArray spells empty() as isEmpty(), and QString is UTF-16. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: qualify the std::pmr::string support claim Listing std::pmr::string as fully supported was an overclaim: it was only ever checked with the default memory resource, which is not what PMR is for. basic_json cannot be given an allocator or a memory resource, so a pmr string inside a value always allocates from std::pmr::get_default_resource(), and assigning an arena-backed string into a value silently drops its resource, because polymorphic_allocator does not propagate on copy construction. Passing polymorphic_allocator as AllocatorType does not compile either. Only the process-global set_default_resource() redirects these allocations. Say so, and separate the row from std::basic_string with a custom stateless allocator, which is unaffected. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: remove a duplicated StringType compatibility section The StringType section carried two 'Compatible types' tables and two copies of the reference-implementation tip. The second table was a stale copy from before the binary format string fixes and still listed std::pmr::string and std::basic_string with a custom allocator as unusable, contradicting the corrected table a few lines above it, and it dragged along the old explanation that blamed int_to_string. Drop the stale copy and put the surviving table before the notes, so the 'see below' in the std::pmr::string row points forwards. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: correct the template parameter requirements after independent verification Every claim on the page was re-checked by compiling and running it, including the rows that say a type cannot be used, which were checked to fail for the documented reason and not merely to fail. Twenty-four claims were wrong. The most consequential: the incomplete-type constraint applies to ObjectType only. object_t is instantiated inside the class definition, because it is probed for key_compare; array_t is only named there and is not instantiated until basic_json is complete. So eastl::vector, QList and QVector are not excluded by incomplete types at all -- they simply have no max_size() -- and absl::InlinedVector is excluded for a subtler reason of its own. Further corrections: ObjectType does not need erase(key), which has a fallback, but does need at(key) for UBJSON output; only == and < are used, or == and <=> under C++20, not all six; the documented adapter does not fit ankerl or robin_hood. ArrayType needs no initializer-list insert, and value_type, the (count, value) constructor and swappability are per-function, not always. BinaryType needs a range insert for CBOR indefinite-length byte strings and does not need push_back. StringType needs append(const StringType&) unconditionally, and does not need operator!= or operator== against const char*; empty(), resize(n) and reserve(n) are per-subsystem; int_to_string is needed by diff, items and std::hash rather than by JSON Pointer or flatten. BooleanType must be implicitly convertible from bool, and JSONSerializer's second parameter need not carry a default. std::pmr::string was wrong in the other direction this time: a moved-in string does keep its memory resource, and later growth allocates from it. Only copies land on the default resource. Five requirement violations are not caught at compile time rather than the two the page claimed; they are now listed together up front. Split every compatibility table into what works and what does not, as the reasons in the second half are the useful part. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Reduce the string_t and array_t members the library requires Several members were required only because of how the library happened to be written, not because the functionality needs them. Dropping them widens the set of usable string and array types, and one of them was also a performance problem. string_t: - c_str() is gone. Every call site already knew the length and passed it along, so data() is enough. The one place that did not, the diagnostics path in exceptions.hpp, now builds the token from data() and size(), which also stops it from truncating keys that contain a null byte. - back() is gone; the serializer indexes the last character instead. - find(str, pos), replace(), and substr() are gone. escape() and unescape() rebuilt the string with one replace() per escaped character, which moves the tail every time: escaping a string of n characters that all need escaping cost O(n^2). Both now scan with find_first_of() -- a member the pointer parser already required -- and append whole runs, so the common case is one search and one copy. Escaping 64000 tildes drops from 717 ms to 20 ms; a string with nothing to escape gets faster too (8.4 ms to 5.8 ms), because the scan is still a single memchr per pass. json_pointer::split() takes its reference tokens with the (const char*, size_type) constructor rather than substr(). - json_pointer::to_string() accumulates with concat<string_t> instead of letting concat default to std::string and converting afterwards, so streaming a json_pointer no longer requires string_t to be assignable from a std::string. array_t: - at(size_type) is gone. basic_json::at(size_type) checked the index by calling array_t::at() and translating std::out_of_range, which also required the array type to throw that exact exception. It now compares against size() and uses operator[]. The thrown exception, its message, and the behaviour under JSON_NOEXCEPTION are unchanged. The BSON writer wrote the terminating null byte out of the string's own buffer (size() + 1). It now writes the byte itself, so string_t::data() need not be null-terminated for to_bson(). The tests pin the reduced API: alt_string loses the five dropped members and gains coverage of the escaping paths, and a std::vector whose at() is hidden is used as an ArrayType. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: record the reduced string_t and array_t requirements Drop c_str(), back(), find(str, pos), replace(), and substr() from the StringType requirements and at(size_type) from the ArrayType ones, and note the string assignment the JSON pointer code performs. Streaming a json_pointer no longer needs assignability from a std::string. Add the non-null-terminated data() to the list of violations that are not diagnosed at compile time -- it was described in the StringType section but missing from the summary at the top -- and correct the QString row, which no longer fails for the c_str() it lacks. JSON_CATCH_USER no longer wraps a catch of std::out_of_range: the last one went away with array_t::at(). Describe what the library actually catches. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use character literals for the signed BinaryType test MSVC rejects char(0xFF) with C4310 (cast truncates constant value), which the Windows workflow treats as an error. The character literals carry the same byte values without a narrowing cast. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not instantiate a hash map with an incomplete basic_json in the tests object_t is probed for key_compare inside the definition of basic_json, so it is instantiated while basic_json is still incomplete. Whether a hash map survives that depends on the standard library: libstdc++ 9 needs the size of the mapped type to instantiate std::unordered_map's node type and rejects the adapter, which broke the GCC 9 builds. The test now derives its no-key_compare object type from std::map -- which does cope -- and shadows the inherited key_compare member type with an entity that is not a type, so the library's probe finds none, exactly as for a hash map. The unflatten() order-independence checks in unit-json_pointer already cover the behaviour that the unordered object type was there for. The limitation is documented for std::unordered_map. Also address two Clang-Tidy findings the earlier commits introduced: erase_from_object() declares its iterator with auto, and at(size_type) checks the type first and then falls through to the return instead of throwing from an else branch. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Keep diagnostic key paths null-terminated Building the token from data() and size() kept an embedded null byte in the key, and since what() hands out a C string, that truncated the whole message rather than just the key: to_bson() on a key containing U+0000 reported "[json.exception.out_of_range.409] (/en" instead of the full explanation. This broke test-bson under JSON_DIAGNOSTICS. Constructing from data() alone stops at the first null byte, which is what c_str() did before, so the message is unchanged -- without requiring string_t to provide c_str(). Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not parse the value in the array-at() test JSON_DIAGNOSTIC_POSITIONS adds the byte range of the value to the exception message, which a parsed value has and an in-memory one does not, so the two message checks failed in that configuration. Build the array in memory instead of parsing it; the test is about at(size_type) not needing array_t::at(), and the byte range is beside the point. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Move the custom BinaryType tests into their own translation unit The two sections added to unit-regression2.cpp brought a third full basic_json instantiation into a translation unit that was already large. With Clang on MinGW that pushed the object over the reach of a 32-bit relocation and test-regression2_cpp20.exe failed to link: relocation truncated to fit: IMAGE_REL_AMD64_REL32 against `.rdata' unit-regression2.cpp is restored to exactly what it was before, and the coverage moves to unit-custom-binary-type.cpp, next to the object and array type tests it belongs with. The signed value type is now also covered in C++11, where std::byte is not available. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not require the container iterators to be nothrow move constructible iter_impl declared its defaulted move operations noexcept. The exception specification a defaulted function gets implicitly follows from its members, here internal_iterator, which holds the object and array iterators. libstdc++ gives std::deque's iterator a user-provided copy constructor without noexcept before version 11, so the implicit specification is noexcept(false) and does not match the declared one. That deletes the function -- and with g++ 4.8, which predates CWG 1778, it is an error outright: error: function 'iter_impl<basic_json<std::map, std::deque> >::iter_impl( iter_impl&&)' defaulted on its first declaration with an exception-specification that differs from the implicit declaration So std::deque, which this branch documents as a usable array type, could not be used with an older standard library. Leaving the specification to be computed cannot mismatch; iteration_proxy_value already spells out the same condition next door. The default configuration is unaffected: json::iterator, json::const_iterator and ordered_json::iterator stay nothrow move constructible and move assignable, which the test now checks so it cannot regress unnoticed. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Address two Clang-Tidy findings the custom container tests exposed Both come from instantiating basic_json with containers other than the default ones, and neither shows up with the Clang-Tidy version available outside CI: - insert(const_iterator, basic_json&&) forwards its by-value iterator to the const-reference overload. performance-unnecessary-value-param asks for the copy to be a move; it only fires for an iterator that is not trivially copyable, as std::deque's is not. The NOLINT on the function does not cover it, because the finding is reported where the parameter is used rather than where it is declared. Move it, which is what the check asks for and is a (very small) improvement in its own right. - cppcoreguidelines-use-enum-class rejects the unnamed enum that shadowed the inherited key_compare member type. An enum class would not do, since it declares a type of that name and the probe would find it again; a member function declaration hides the name just as well. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Assert the iterators' exception specification relative to the container The test pinned that nlohmann::json's iterators stay nothrow movable after iter_impl's defaulted move operations lost their declared noexcept. That is not a property of the library, though: the exception specification is now computed from the container iterators, so it holds only for standard library implementations whose iterators are themselves nothrow movable. MSVC's checked iterators before VS2017 are not -- _Iterator_base12 registers the iterator with the container's debug proxy in a copy constructor that carries no noexcept -- so the assertions fail on a Visual Studio 2015 debug build, which is the one debug configuration in the AppVeyor matrix and has no counterpart in the GitHub Actions matrix. Assert what the change actually guarantees instead: the iterators are nothrow movable exactly when the object and array iterators they are built from are. That still pins the default configuration against a silent regression, and it is true whatever the standard library provides. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Detect a void-returning erase() through a named trait erase_from_object() distinguished its two overloads with a decltype of a member call written inline in a default template argument. Every other detection in the library goes through the detector machinery in detected.hpp instead -- has_erase_with_key_type is the same question about the same member function -- and the inline form is the one shape older compilers are least reliable about. Express it the same way: detect_erase_with_iterator plus is_detected_exact, both of which the library already relies on elsewhere. No behaviour changes. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Give the custom container types only the constructors the library uses The three container types in the new tests inherited every constructor of their base with using Base::Base. That asks for more than the test needs: the library builds an object or an array by default construction, by copy or move, and -- when converting between two basic_json types or from an initializer list -- from an iterator range. Declaring those directly makes the requirement visible in the test, and keeps object types out of a corner where a compiler has to declare std::map's whole constructor set for a derived class while basic_json is still incomplete. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Temporarily disable the new custom container tests AppVeyor is the only CI that builds MSVC 2015 and 2017, and it has now rejected three heads of this branch. Its build log is not reachable from where this is being worked on, so the verdict is a single bit and the cause has to be narrowed down by bisection. Everything else stays: the library changes, the reduced alt_string, and the unflatten() tests. If AppVeyor passes with these three translation units disabled, the cause is one of the six basic_json instantiations they add; if it fails, it is in the library. Either way this commit is reverted. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard the disabled tests with a macro rather than #if 0 Clang-Tidy's readability-avoid-unconditional-preprocessor-if rejects a literal #if 0. Use a macro that is never defined instead, which the check does not look at. Still temporary, and reverted together with the previous commit. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Re-enable the array and binary container tests AppVeyor passed with all three new translation units disabled, so the library changes, the reduced alt_string, and the unflatten() tests are fine on MSVC 2015 and 2017; the cause is one of the six basic_json instantiations the new tests add. Bring back two of the three. If AppVeyor passes again, the cause is in unit-custom-object-type.cpp, which is the one still disabled; if it fails, it is in one of these two and needs one more split. Still temporary. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Diagnose two silently violated template parameter requirements Both were on the list of requirements that are not caught at compile time and corrupt values rather than failing, and both are a plain size comparison: - A BinaryType whose value_type is wider than one byte, which the readers and writers reinterpret as raw bytes anyway. - A NumberUnsignedType too narrow to hold the absolute value of every NumberIntegerType value, which makes basic_json(INT64_MIN).dump() yield -0 for std::int64_t with std::uint32_t. Neither static_assert rejects a configuration that worked before: both only fire where the result was already wrong. Also add the two comments the review asked for, in write_bson_string() and calc_bson_array_size(), matching the ones their counterparts already carry. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Align the template parameter tables and record what is now diagnosed Every table in the page is reformatted so each column is exactly as wide as its widest cell, which is what the review asked for in a dozen places: the separator rows that ran two dashes long, the stray spaces, and the columns padded well past their content. The row listing six containers that require a complete mapped type is split in two so that one cell no longer sets the width of the whole table. Content changes: NumberUnsignedType is described as any unsigned integer type at least as wide as NumberIntegerType rather than any unsigned integer type; the two requirements that are now static_asserts move out of the list of violations that are not caught at compile time; and the two places that require a non-const operator[] say why data() will not do (std::string has no non-const data() before C++17). Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Bisect the other way: only the object container tests The previous head touched only docs/, which AppVeyor's only_commits filter skips, so it produced no build and no status at all -- the pull request looked green without ever having been built on MSVC 2015 or 2017. Swap the guards instead of repeating that step: unit-custom-object-type.cpp is enabled and the array and binary translation units are disabled. AppVeyor already passed with all three disabled, so a failure here pins the cause on no_key_compare_json or void_erase_json, and a pass pins it on the array or binary file. Still temporary. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Split the two object types apart AppVeyor failed with only unit-custom-object-type.cpp enabled and passed with all three new translation units disabled, so the cause is one of the two object types in this file and not the array or binary ones. Guard out void_erase_map and leave no_key_compare_map, which separates the two constructs under suspicion: shadowing the inherited key_compare member type with an entity that is not a type, and hiding the inherited erase with a void-returning overload. A failure here points at the first, a pass at the second. Still temporary. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Build the no-key_compare object type by composition, not inheritance The "object type without key_compare" test failed on AppVeyor's MSVC 2017 jobs (/std:c++17): its no_key_compare_map derived publicly from std::map and shadowed the inherited key_compare type with a same-named member function, relying on ordinary member hiding to make key_compare unreachable as a type for the library's detection trait. MSVC 2017 does not honor that hiding for a typename-qualified lookup performed from outside the class and still resolves key_compare to the base's comparator type, so object_comparator_t incorrectly picked it up instead of falling back to default_object_comparator_t. Wrapping a std::map by composition instead removes the base class entirely, so there is no key_compare to find under any lookup rule, on any compiler. Also drops the now-unneeded JSON_BISECT_CUSTOM_CONTAINER_TESTS guard left over from narrowing this down: the void_erase_map test in the same file was never the cause and is re-enabled unconditionally. Verified locally with clang++ and g++ under C++17 and C++20. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Re-enable the array and binary custom-container tests unit-custom-array-type.cpp and unit-custom-binary-type.cpp were still guarded behind JSON_BISECT_CUSTOM_CONTAINER_TESTS from bisecting the AppVeyor failure fixed in 7c39f3227, which was unrelated to either file. The macro was never defined, so none of these tests actually ran in CI. Verified locally with clang++ and g++ under C++17 and C++20 before removing the guards. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix indentation of custom_object_type per astyle The one-line function bodies in the composition-based no_key_compare_map (7c39f3227) do not match the project's Allman brace style, which the ci_test_amalgamation job enforces with astyle. Reformatted with the pinned astyle 3.4.13; no functional change. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * docs: add compiled reference implementations for the container template parameters Each of ObjectType, ArrayType, StringType, and BinaryType now links to a minimal, self-contained header (docs/mkdocs/docs/examples/custom_*_type.hpp) that wraps the corresponding standard container by composition and satisfies every "Always required" member listed on that page. Unlike the prose requirement lists, these are real code: each header has a companion .cpp that instantiates a basic_json specialization with it and is compiled and run by the existing ci_test_examples check (docs/Makefile's check_output_portable), so the reference implementations cannot silently drift from what the library actually requires. The .output files were generated with that same target. StringType's existing pointer to tests/src/unit-alt-string.cpp's alt_string is kept alongside the new header as a more thorough, battle-tested example. Verified locally: astyle (pinned 3.4.13, project .astylerc) on the new files; clang++/g++ under C++11/17/20 for each example against the amalgamated header; `make check_output_portable` in docs/; `mkdocs build --strict` and scripts/check_structure.py for the page itself. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Declare no_key_compare_map's accessors noexcept The GCC C++20 job builds with -Wnoexcept and -Werror, and the standard library takes noexcept(c.begin()) and noexcept(c.end()) in ranges_base.h and range_access.h. Forwarding to std::map without repeating its noexcept made those expressions false, which the warning reports as an error: error: noexcept-expression evaluates to 'false' because of a call to no_key_compare_map<...>::begin() [-Werror=noexcept] note: but ... does not throw; perhaps it should be declared 'noexcept' Give the accessors the exception specification of what they forward to. std::map declares begin, end, cbegin, cend, empty, size, max_size, and clear noexcept, so the wrapper does too. swap is left alone: std::map's is only conditionally noexcept, and nothing asks for it. void_erase_map is unaffected because it still derives from std::map and inherits accessors that already carry the specification. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Write out what the defaulted constructor of no_key_compare_map implied ci_test_gcc builds with -Weffc++, which asks for data to be initialized in a member initialization list; a defaulted default constructor does not do that: error: 'no_key_compare_map<...>::data' should be initialized in the member initialization list [-Werror=effc++] Writing the constructor out satisfies that but drops the exception specification the defaulted one carried, which -Wnoexcept then objects to where the standard library takes noexcept(construct(...)). Declare it the way the defaulted constructor was: noexcept when the wrapped map's default constructor is. This is the cost of composition -- inheritance carried std::map's exception specifications and initialization for free, and forwarding by hand has to restate them. Checked with the repository's own GCC warning set from cmake/gcc_flags.cmake, all 346 flags, at C++11, C++17 and C++20: no diagnostics for this file, nor for the two custom container translation units that were disabled while the MSVC failure was narrowed down and are built again now. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Mark no_key_compare_map::swap noexcept Clang-Tidy rejects a swap that is not: error: swap functions should be marked noexcept [cppcoreguidelines-noexcept-swap,performance-noexcept-swap] It was left unmarked on the grounds that std::map::swap is only conditionally noexcept, so an unconditional promise would be wrong for a comparator or allocator that can throw while swapping. Both concerns are met by taking the specification from the wrapped map rather than asserting one: noexcept(noexcept(data.swap(other.data))). Clang-Tidy accepts that, and no NOLINT is needed. Last in the series of specifications that inheritance used to supply and composition has to write out by hand. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018hxZxz8svM54c6ATEvXp5E Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Opus 5 <noreply@anthropic.com> | 1 天前 | |
Speed up parsing of contiguous input (numbers, strings, UTF-8) (#5283) * Speed up number parsing in the lexer (fast paths from the fast_float/simdjson world) The number scanner converted its already-validated digit buffer with std::strtoull/std::strtoll/std::strtod. Those pull in locale and errno machinery and dominate number-heavy parsing (strtod runs at ~6 M/s). Replace them with dedicated parsers over the validated buffer: - parse_integer_unsigned / parse_integer_signed: accumulate digits with overflow detection, falling back to the float path on overflow exactly as the strtoull/strtoll round-trip check did. Overflow behavior is unchanged for narrower or wider custom number types. - parse_float_fast: Clinger's exact fast path for `double` (<=19 significant digits, |exp10| <= 22, significand < 2^53), where significand * 10^exp is exact under IEEE round-to-nearest. This is the same fast path used by fast_float/simdjson. It is bit-identical to strtod on this subset and declines (falling back to strtod) otherwise. Only `double` uses it; float and long double keep std::strtof/std::strtold via a templated overload. Measured on representative data (g++ 13, -O3): - integers: DOM parse +11%, SAX +25-34% - floats: DOM parse +37%, SAX +70% (clang: float DOM ~1.9x) No dependencies added; header-only and C++11-clean. Existing parser, lexer, conversion and deserialization unit tests pass unchanged; a 3M-value random-double fuzz matches strtod bit-for-bit. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add SWAR bulk string scanning for contiguous input (simdjson-style) scan_string() read the input one character at a time through the input adapter and classified every byte with a large switch. For contiguous byte buffers we can instead scan 8 bytes at a time with a SWAR word test that finds the first byte needing individual handling (the closing quote, an escape, a control character, or a non-ASCII UTF-8 byte) and bulk-append the ordinary run in one go. - input adapters expose supports_bulk_scan / bulk_data / bulk_remaining / bulk_skip for provably-contiguous, same-type, 1-byte iterator ranges (raw pointers in every standard; std::string/std::vector/std::array and friends additionally in C++20 via std::contiguous_iterator). - the lexer gains a bulk_scan capability (gated on lazy_token_string so bypassing the per-character capture cannot lose error diagnostics) and a scan_string_bulk() fast path; streaming/wide/user adapters are unchanged and keep the byte-at-a-time scanner. The run contains no newline (all bytes < 0x20 are treated as special), so position bookkeeping stays exact, and error tokens are still reconstructed lazily from the consumed byte range. The SWAR special-byte test is pure uint64_t arithmetic - no intrinsics, no runtime dispatch, C++11-clean. Measured on representative data, pointer input, g++ 13 -O3 (string values discarded by accept() see the largest gains): long ASCII strings: DOM +4.5x, SAX +14x, accept +17x (to ~2 GB/s) short strings: DOM +15%, SAX +62%, accept +85% escape-heavy: DOM +31%, SAX +26%, accept +28% Same-input parity verified: 200k randomized documents (escapes, multibyte UTF-8, surrogate pairs) accept/parse identically via the contiguous SWAR path and the streaming byte path; unit lexer/parser/diagnostic-position/ deserialization/conversions suites pass unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Validate UTF-8 in the bulk string scanner (portable ~2-3x on non-ASCII text) The SWAR bulk string path stopped at the first non-ASCII byte and handed every multibyte character to the byte-at-a-time scanner, whose per-byte get()/next_byte_in_range()/add() machinery runs at roughly half the speed of validating straight from the buffer. As a result, dense non-ASCII text (CJK, emoji, accented Latin) parsed ~10-15x slower than ASCII. Fold well-formed UTF-8 into the bulk run: scan_string_bulk() now, on a non-ASCII lead byte, validates one sequence with validate_one_utf8() - which mirrors scan_string()'s per-byte switch ranges exactly (rejecting overlong forms, surrogates, and out-of-range code points) - and appends it in place, continuing until the closing quote, an escape, a control byte, or an ill-formed sequence. All error handling still defers to the byte path, so error messages and positions are byte-for-byte unchanged. Because only well-formed content is fast-pathed and every rejection falls through to the existing scanner, behavior is identical; the win is purely throughput. Measured on pointer input (accept, string values discarded): content g++ 13 clang 18 dense CJK 277 -> 648 ~605 MB/s (~2.3x) dense emoji 299 -> 857 ~702 MB/s (~2.6-2.9x) mixed 90% ASCII 246 -> 331 ~334 MB/s (~1.35x) pure ASCII unchanged (~3.2 / 4.1 GB/s) Verified: 2,000,000 randomized documents built from arbitrary bytes (overlong, surrogate, truncated, out-of-range sequences) accept/reject and parse identically via the contiguous path and the streaming byte path; lexer/parser/diagnostic-position/deserialization/conversions suites pass unchanged. Pure C++11, no intrinsics. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Route contiguous byte containers through the pointer adapter (fast paths in C++11) json::parse(std::string) - the most common entry point - did not benefit from the contiguous fast paths (bulk string scanning, UTF-8 bulk validation, memcpy for binary formats) in C++11..17: std::string::iterator is a library wrapper, not a raw pointer, and pre-C++20 there is no portable way to prove it contiguous, so supports_bulk_scan was false. Only raw pointers, string literals, and C-arrays (and, in C++20, anything modelling std::contiguous_iterator) took the fast path. Detect contiguous single-byte containers (std::string, std::vector<char>, std::vector<std::uint8_t>, std::string_view, ...) via is_contiguous_byte_ container and route them through an iterator_input_adapter built from data()/data()+size(). The generic iterator-based container overload is constrained to exclude these, so the two overloads are disjoint and there is no ambiguity (a plain competing overload loses to the greedy forwarding-reference container overload on reference binding, and a factory partial-specialization is ambiguous - both were tried and rejected). The pointer keeps the container's own element type, so char_type - and therefore all parsing behavior - is byte-for-byte identical to the iterator path (const char* for std::string, const std::uint8_t* for std::vector<std::uint8_t>); only the raw pointer additionally turns on the fast paths. Lifetimes are unchanged: the container outlives the adapter for the full parse expression, exactly as the iterators it replaces did. Measured, C++11, json::parse/accept(std::string), g++ 13: long ASCII strings: accept 201 -> 3200 MB/s (~16x), parse 174 -> 1444 dense CJK: accept 263 -> 697 MB/s (~2.6x) short strings: accept 163 -> 243 MB/s (~1.5x) Verified: char_type preserved for std::string (char) and std::vector<std::uint8_t> (uint8_t); CBOR/MsgPack round-trips from std::vector<std::uint8_t> unchanged; 1,000,000 randomized documents accept and parse identically via std::string and via std::istream; deserialization/user-defined-input/parser/lexer/conversions/diagnostic- position suites pass (20,480 assertions); warning-clean on g++ and clang in C++11/17/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add optional simdutf backend for bulk UTF-8 validation (JSON_USE_SIMDUTF) The bulk string scanner validates UTF-8 straight from a contiguous buffer. The scalar validator caps at ~0.3-0.7 GB/s on non-ASCII text; a SIMD validator reaches several GB/s. Rather than hand-rolling SIMD UTF-8 validation (easy to get subtly wrong - a from-scratch SSE attempt rejected valid CJK), wire in the vetted simdutf library behind an opt-in switch. simdutf is not header-only (it ships simdutf.cpp and uses runtime CPU dispatch), so it is not vendored: defining JSON_USE_SIMDUTF includes <simdutf.h> and routes the bulk validator through simdutf::validate_utf8; the project supplies and links simdutf. Undefined (the default), nothing external is included and the portable C++11 scalar path is used, so the library stays header-only and its baseline behavior is unchanged. Design keeps behavior identical either way: - scan_string_bulk() now finds the run up to the next quote/escape/control byte (non-ASCII allowed) and validates it in one shot; on the rare validation failure it recomputes the exact valid prefix with the scalar helper, so ill-formed input still falls through to the byte path and is reported at the same position with the same message. - the per-sequence scalar path is factored into scalar_string_bulk_run() and is the default backend; the refactor is behavior-preserving and does not change scalar throughput. Verified: default and JSON_USE_SIMDUTF builds accept/reject/parse identically across 2,000,000 arbitrary-byte documents and 1,000,000 mixed-escape/UTF-8 documents (differential fuzz vs the streaming byte path); lexer/parser/diagnostic-position/deserialization suites pass under both configurations (20,188 assertions with the backend enabled); warning-clean on g++ and clang, C++11 and C++20, both configurations. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add a contiguous fast path for scanning numbers scan_number() reads a number one character at a time through the input adapter (get()) and appends each byte to token_buffer (add()) before converting. For contiguous input, the per-character get()/add() overhead dominates: it is roughly two thirds of the time spent on number-heavy parsing, far more than the value conversion itself. Add scan_number_bulk_contiguous(), which parses the whole number token straight from the input buffer: it validates and classifies the extent with the same grammar as scan_number()'s state machine, materializes token_buffer in one copy (substituting the locale decimal point exactly as scan_number() does), advances the adapter, and reuses the shared convert_number() tail. On anything it does not 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 then produces the exact diagnostic. Errors and their positions are therefore unchanged. The conversion tail is factored out of scan_number() into convert_number() so both scanners share it; the fast path is selected by tag dispatch on the existing bulk_scan capability, so streaming/wide/user adapters are unaffected. Measured on pointer input, g++ 13 -O3: - integers: parse +65%, accept +98% - floats: parse +39%, accept +70% Verified: 2,000,000 randomized number documents (including overflow-range integers, long digit strings and %.17g doubles) parse identically via the contiguous path and the streaming byte path, matching value, type and round-trip text; the locale suite and existing parser/lexer/conversions/ deserialization tests pass; a new "lexer number fast path" test checks contiguous-vs-streaming parity, token classification, and that malformed numbers are rejected identically. Pure C++11, no intrinsics. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Move byte-level scan/parse helpers out of lexer.hpp lexer.hpp had grown by ~600 lines of byte-level helpers that have no dependency on the lexer's template parameters and clutter the state machine. Move them, unchanged, into two focused headers as free functions in namespace detail: - number_parse.hpp: parse_integer_unsigned/parse_integer_signed (now templated on the number type) and parse_float_fast (Clinger's exact double fast path, with the decimal point passed as an argument instead of read from a lexer member). - string_scan.hpp: the SWAR string helpers (is_string_special, swar_string_special, find_string_special, validate_one_utf8, scalar_string_bulk_run) and the backend-dispatched string_bulk_run, including the optional simdutf include and find_string_delimiter. lexer.hpp now includes these and calls the free functions; the methods that touch lexer state (scan_string, scan_number, scan_string_bulk, scan_number_bulk_contiguous, convert_number) stay put. This is a pure code move with no behavior change: lexer.hpp drops from 2357 to 1934 lines, the now-unused <cstdint>/<cstring>/<limits> includes are removed, and the free-function form makes the SWAR helpers reusable elsewhere (e.g. the serializer's string escaping). Verified: default and JSON_USE_SIMDUTF builds compile; 2,000,000 number and 2,000,000 arbitrary-byte-string differential-fuzz documents parse identically to before; lexer/parser/conversions/deserialization/locale/ diagnostic-position suites pass (20,576 assertions); warning-clean on g++ and clang in C++11/17/20; the amalgamation regenerates and passes check-amalgamation. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix clang-tidy findings and document JSON_USE_SIMDUTF in the nav - number_parse.hpp: use std::array for the powers-of-ten table (avoid-c-arrays) and `auto` for the cast-initialized result (modernize-use-auto), matching the codebase style (cf. the serializer's utf8d table). Indexing casts keep the -Wsign-conversion build clean. - add JSON_USE_SIMDUTF to the mkdocs navigation so the macro page is reachable. No behavior change; clang-tidy is clean on the new headers and the amalgamation is regenerated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix number fast path for custom string types without assign() The contiguous number fast path materialized token_buffer with token_buffer.assign(data, len), but string_t is only required to provide the minimal interface the rest of the lexer uses (push_back, append, clear, operator[], ...). Custom string types such as the test's alt_string do not implement assign(), so scan_number_bulk_contiguous() failed to compile for them (unit-alt-string), breaking the gcc/clang standards and old-compiler CI jobs. reset() already clears token_buffer, so fill it with append() - which alt_string and std::string both provide and which the string fast path already relies on - instead of assign(). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Satisfy clang-tidy: parenthesize math and drop unused forwarding reference The CI clang-tidy (newer than the locally available version) reported two additional checks on the new code: - readability-math-missing-parentheses: parenthesize the (a * b) + c digit accumulations in number_parse.hpp. - cppcoreguidelines-missing-std-forward: the contiguous-byte-container input_adapter overload took a forwarding reference but only reads data()/size() and never forwards it. It is already disjoint from the generic container overload via SFINAE, so a plain const& is correct and clearer (and keeps the container alive for the whole parse just as before). No behavior change; char_type and routing are unchanged (std::string and std::vector<std::uint8_t> still take the pointer adapter with char/uint8_t char_type), CBOR/MsgPack round-trips and the 2M number fuzz still pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use std::from_chars (Eisel-Lemire) for float conversion when available The Clinger fast path is exact only for the "easy" subset (<=19 significant digits, |exp10| <= 22); high-precision and scientific floats fall through to strtod, where the failed Clinger attempt actually makes parsing a net loss. std::from_chars implements the Eisel-Lemire algorithm in modern standard libraries: locale-independent, correctly rounded, and fast over the whole value range. convert_number() now tries parse_float_from_chars() first (guarded by __cpp_lib_to_chars, so C++11 and libc++-without-float-support keep the Clinger + strtod path unchanged), then Clinger, then strtof. from_chars is used only when it consumes the entire token; a partial parse means a non-'.' locale decimal point, and an under-/overflow (result_out_of_range) also declines - in both cases the existing strtod fallback supplies the exact value and the well-defined +/-inf/0 the parser expects, side-stepping the P4168 divergence between implementations. float and long double now get the fast path too (Clinger was double-only). Measured, C++17, g++ 13 -O3, json::parse/accept: - canada-style floats: ~unchanged (Clinger already covered them) - high-precision (17 digits): parse 2.1x, accept 2.5x - scientific (17 digits + exp): parse 3.6x, accept 4.1x Verified: C++11 (Clinger/strtod) and C++17 (from_chars) parse every value - including subnormals, boundary values, and 1e9999/1e-9999 over-/underflow - to bit-identical results; 2M number-fuzz clean; conversions/deserialization/ locale/number-fast-path suites pass in both C++11 and C++17; clang-tidy clean; warning-clean on g++ and clang in C++11/17/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard from_chars use on JSON_HAS_CPP_17, not just __cpp_lib_to_chars libstdc++ 15 defines __cpp_lib_to_chars even in C++14 mode (via bits/version.h pulled in by other headers), but <charconv> is only included under JSON_HAS_CPP_17. That made parse_float_from_chars() reference std::from_chars without the header in C++14 builds, breaking gcc-latest, icpx, and the offline-testdata jobs. Gate the use on JSON_HAS_CPP_17 && __cpp_lib_to_chars so it matches the include condition exactly; C++11/14 always take the scalar fallback. Verified by forcing __cpp_lib_to_chars in a C++14 build: the guard suppresses std::from_chars and it compiles. C++17 behavior is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Disable Clinger float fast path under extended FP precision (x87) The contiguous number fast path uses a Clinger-style exact algorithm (significand * 10^scale in double arithmetic), which is only correctly rounded when double operations are evaluated in true 53-bit precision. On the x87 FPU used by 32-bit x86 (FLT_EVAL_METHOD == 2) the single multiply/divide is computed in 80-bit and then double-rounded to double, so a small fraction of values land 1 ULP off. This surfaced as test-cbor_cpp11 and test-msgpack_cpp11 failing on the mingw (x86) job for regression/floats.json: the C++17 builds pass because they take the correctly-rounded std::from_chars path, while C++11 falls back to parse_float_fast(). A 5M-sample check over shortest round-trip decimals reproduces it: 0 divergences with 53-bit doubles, ~1 in 25 000 with 80-bit intermediates; declining to std::strtod fixes all of them. Guard parse_float_fast() on FLT_EVAL_METHOD so it declines whenever the platform evaluates doubles in extended precision, letting the caller use the correctly-rounded std::from_chars / std::strtod path instead. On mainstream x86-64/ARM64 (FLT_EVAL_METHOD == 0) the fast path is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard <charconv> include with __has_include for GCC 7 GCC 7 sets __cplusplus to the C++17 value under -std=gnu++1z, so JSON_HAS_CPP_17 is defined, but its libstdc++ ships no <charconv> header (added in GCC 8; floating-point from_chars in GCC 11). The unconditional "#if defined(JSON_HAS_CPP_17) #include <charconv>" therefore failed to compile there: "fatal error: charconv: No such file or directory" in the ci_test_compilers_gcc (7) job. Wrap the include in __has_include(<charconv>), mirroring the library's existing handling of <version> and <filesystem> in macro_scope.hpp. When the header is absent, __cpp_lib_to_chars stays undefined and parse_float_from_chars() takes its scalar fallback, so the from_chars use site (already gated on __cpp_lib_to_chars) is never reached. GCC 8-10, which have <charconv> but no floating-point from_chars, are unaffected: they include the header but still take the fallback. GCC 11+ is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Restore the column when ungetting a newline The contiguous number fast path never reads the character that terminates a number token, while scan_number() reads it and then ungets it. When that character is a newline, get() has already cleared chars_read_current_line, and unget() could only restore lines_read - leaving the column at 0. The two paths therefore reported different columns for the same document: json::parse("[01\n]") -> line 1, column 3 json::parse(stringstream) -> line 1, column 0 Remember the column the newline was read at so unget() can restore it. Both paths now report the position the offending token actually starts at, which also fixes the pre-existing column-0 artifact for streaming input. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Extend the bulk scan fast paths to sized sentinels supports_bulk_scan required IteratorType and SentinelType to be the same type, which excluded std::counted_iterator paired with std::default_sentinel_t - the combination #5268 had already enabled for the memcpy fast path. Such input fell back to the byte-at-a-time scanner even though it is contiguous and its remaining length is computable in O(1). Factor the "distance is computable in O(1)" test into sentinel_is_sized and use it for iterator_is_contiguous, supports_seek, and supports_bulk_scan alike, and share the std::ranges::distance/std::distance dispatch through a remaining_count() helper. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Document JSON_USE_SIMDUTF on the macro overview page The macro was only listed in the API macro index; add it to the supported macros overview alongside the other JSON_USE_* macros, and note that it selects between two definitions of the same inline function and so must be defined identically in every translation unit. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cover the counted-iterator bulk scan paths Sized sentinels newly reach the bulk string/number scanners and the seek-based token reconstruction, so exercise both: - diagnostics that quote the offending token, which are rebuilt from the consumed input via copy_consumed_range() - inputs whose count ends before the underlying buffer does, including a closing quote that exists only behind the count, a cut inside an 8-byte SWAR stride, and a cut inside a UTF-8 sequence Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix the SentinelType example on iterator_input_adapter The comment offered "a C++20 sentinel or counted_iterator" as examples of a SentinelType, but std::counted_iterator is the IteratorType - the sentinel it pairs with is std::default_sentinel_t. #5268 corrected the same wording in the API documentation and left the code comment behind. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not discard the parse result in the error position check json::parse is declared warn_unused_result, and CHECK_THROWS_WITH_AS evaluates its expression as a discarded statement, so the assertion broke the -Werror builds (GCC -Werror=unused-result, MSVC C4834 under /WX). Compare against the helper that already captures the message instead. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Lock the two number grammars together with a parity test The JSON number grammar is encoded twice: as the scan_number() state machine and as the contiguous fast path. The fast path declining on anything it does not recognize keeps most divergence harmless, but if it ever accepted something the state machine rejects the result would be a silent correctness bug, and the existing test only pinned a hand-written list of numbers. Enumerate every string of length 1..4 over "01.eE+-" (2800 tokens) and require both paths to agree on the parsed value and on the exact error message. Verified to fail if the fast path's grammar is perturbed. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Skip token_buffer for integers on the contiguous path An integer token does not need token_buffer: the number_integer and number_unsigned SAX callbacks take only the value, and the overflow diagnostic rebuilds the text from the input via get_token_string(). Convert straight from the input buffer and materialize the token only for the floating-point tail, which still needs a NUL-terminated buffer for strtod. JSON_DIAGNOSTIC_POSITIONS derives a number's start position from get_string().size(), so the copy is kept when that is enabled. The integer dispatch is factored into convert_integer() and shared with convert_number(), so both scanners keep using one implementation. Integer-heavy input, 400k values, -O3: parse accept gcc 16 +14% +23% clang +15% +22% Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cover the bulk string and UTF-8 scanners These paths had no dedicated tests and rested on differential fuzzing only. Add three sections, all comparing the contiguous scanner against the byte-at-a-time one on the parsed value and on the exact error message: - every string of length 1..3 over an alphabet of ordinary ASCII, both specials, a control byte, escape characters, UTF-8 lead and continuation bytes, and a byte that is never valid - each at offset 0 and offset 9, so the bulk scanner sees them with and without a run behind them - every kind of run-ending byte at each offset across two 8-byte SWAR words, so multibyte sequences also straddle the word boundary - the boundaries of every range validate_one_utf8() recognizes: shortest and longest encodings, overlongs, both ends of the surrogate block, U+10FFFF and just past it, and truncated sequences Verified to fail if the bulk validator accepts surrogates, and if the SWAR word test stops detecting control characters. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Format the new string fast path test with astyle The pinned astyle expands a braced-init-list used as a range-for range onto several lines; hoist the two offsets into a named vector instead, which reads better and leaves nothing for astyle to reformat. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix shadowed locals and guard the exception-dependent tests Two problems in the tests added for the bulk scanners, both found by CI: - the inner `const json j` in the counted-iterator diagnostics shadowed the one declared at test-case scope, which -Wshadow rejects on GCC and clang and C4456 rejects on MSVC under /WX; rename them - the new parity checks parse deliberately invalid input, which calls std::abort() rather than throwing when JSON_NOEXCEPTION is defined, so they would have crashed the no-exception build; guard them the way the other tests do json::accept() does not abort, so the UTF-8 range assertions stay compiled without exceptions and keep covering validate_one_utf8() there. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Gate the C++20 iterator classification on JSON_HAS_RANGES Making supports_bulk_scan depend on iterator_is_contiguous meant the trait is now instantiated for every adapter, not only when get_elements() is called. On standard libraries with an incomplete <ranges> that is fatal: libstdc++ 10 evaluates std::contiguous_iterator<std::counted_iterator<T*>> by calling std::to_address, which needs an operator-> its counted_iterator does not have, so satisfaction checking is a hard error rather than false. Reported by clang 14 + libstdc++ 10. JSON_HAS_RANGES already encodes exactly this ("libstdc++ < 11 has incomplete C++20 ranges", #4440), so require it for the C++20 branch. Affected toolchains fall back to the pointer-only test and the byte-at-a-time scanner, which parses identically, just without the bulk fast paths. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Satisfy clang-tidy in the new bulk scanner tests - give the helper lambdas an explicit std::string return type and return braced initializer lists (modernize-return-braced-init-list) - replace the C-style array of test cases with a std::vector (modernize-avoid-c-arrays) - silence pro-type-member-init on the two brace-initialized aggregates; default member initializers would stop them being aggregates in C++11 Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Avoid escaped literals in the counted-iterator diagnostics list clang-tidy reads "[\"\\ud834\"]" as a literal better written raw, and the two literals written next to each other in "[\"a\x01""b\"]" as a missing comma. The concatenation was there to stop the hex escape swallowing the following character; build those documents from explicit bytes instead and use raw strings elsewhere. The byte sequences are unchanged. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Reattach convert_number's documentation @gregmarr spotted that convert_integer() was inserted between convert_number() and its doc block, leaving convert_integer() with two stacked blocks and convert_number() with none. Comment only; no code change. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Address review comments on the simdutf backend string_bulk_run() had the same `return scalar_string_bulk_run(...)` in both arms of the `#if`. The simdutf arm already falls through when validation fails, so a single return after the `#endif` says the same thing. The JSON_USE_SIMDUTF example showed `#include <simdutf.h>`, which string_scan.hpp already does under the same guard; users only have to put the header on the include path and link the library, not include it themselves. Set the version history entry to 3.13.0, matching the other macro pages documenting unreleased features. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Pin the number error position against a non-newline terminator The comment claimed the reported column is the one the offending token starts at. It is the column reached after the token's last character - which is the actual point of the unget() change: a number terminated by a newline now reports what the same number terminated by a space always did. Assert that equality directly, and add a multi-character token where the start and end columns differ, so the invariant cannot be read off a single-character example. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not repeat the integer conversion that just failed scan_number_bulk_contiguous() converts an integer token straight from the input buffer. When the value does not fit, it materializes token_buffer and calls convert_number(), which tried the very same integer conversion again before falling back to floating point. Recording the outcome in number_type skips the second attempt. The resulting token type and value are unchanged: convert_number() reached the float tail either way. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Require a container's value_type to match what data() points at is_contiguous_byte_container accepted any type with a data() returning a pointer to a single-byte integral plus a size(). That is duck typing: the two members say nothing about size() counting the units data() points at. A type where it does not - fixed-size records, say - was routed to the pointer-based adapter and parsed as [data(), data() + size()) bytes, silently truncating input the iterator-based adapter had read in full: struct record_buffer { using value_type = std::array<char, 4>; std::string bytes; const char* data() const; // raw bytes std::size_t size() const; // in records const char* begin() const; const char* end() const; }; json::parse(record_buffer{"[1,2,3,4,5]"}); // parse error at column 3 Requiring the container's own value_type to be that same element type ties the two together. Every contiguous standard container satisfies it, so std::string, std::vector<char>, std::array<char, N> and std::string_view keep the fast path; anything else falls back to the iterator-based adapter, which is always correct. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Only compile the simdutf backend from C++17 on simdutf.h rejects anything below C++17 with an #error, so defining JSON_USE_SIMDUTF in a C++11 or C++14 translation unit did not fail with a message about simdutf being unavailable - it failed to compile at all, taking the library's C++11 support with it. Nothing caught this because no build ever compiled that path. Gate the include and both uses on JSON_HAS_CPP_17, the same way number_parse.hpp gates std::from_chars. Below C++17 the macro now has no effect and the scalar validator runs; it accepts and rejects exactly the same input, so the macro is safe to set project-wide even when some translation units use an older standard. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Test the simdutf backend in CI JSON_USE_SIMDUTF was documented and shipped but never compiled by anything in the repository, so nothing held the backend to the behavior the docs promise. Add JSON_TestSimdutf (OFF by default), which fetches simdutf and defines JSON_USE_SIMDUTF for every test target, and a ci_test_simdutf target that runs the whole suite in that configuration. Because simdutf needs C++17, the suite is built at C++11 as well, so one job covers both the scalar fallback with the macro defined and simdutf itself. The dependency hangs off test_main, whose usage requirements every test target inherits. The library target and the installed CMake package are deliberately untouched: making nlohmann_json link simdutf would put a find_dependency() in the exported package, which is a separate decision. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Warn when JSON_TestSimdutf cannot reach the simdutf backend simdutf needs C++17: without it the dependency does not even compile, and with a C++17 compiler but no C++17-or-later standard under test it builds and then goes unused. Either way the option silently did nothing useful, or broke the configure step outright. Resolve the tested standards first, then check them: when none of them can reach simdutf, skip the dependency and say so, naming which of the two reasons applies and how to fix it. The tests then run against the scalar validator, which is what would have happened anyway. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use record_buffer::data() so clang does not flag it unneeded The record_buffer test type declares data() and size() so the is_contiguous_byte_container trait can see both and still reject the type on its value_type. data() was never called, so clang's -Wunneeded-member-function (under -Weverything -Werror) failed the C++20 build. Assert that data() points at the underlying bytes: it ODR-uses the member and documents the property the type is meant to demonstrate. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Skip the float fast path when it cannot succeed parse_float_fast() (Clinger) needs a significand below 2^53, so it always declines once the mantissa has 17 or more significant digits. convert_number() called it unconditionally, so those numbers were walked an extra time before strtod had to run anyway. On streaming input, where scanning is byte-at-a-time and there is no compensating win, that made canada.json about 6% slower than develop. Derive the significant-digit count from token_buffer indices - the digits are not scanned again - and skip the call when it is guaranteed to decline. Both scanners pass the offset where the mantissa ends; the count only has to be corrected for a leading "0", which the JSON grammar admits nowhere else. The integer path returns before the check, so integer-heavy input is unaffected. Values are unchanged: this only avoids an attempt that would have failed. Verified bit-exact against develop over every number in canada.json, floats.json, signed_ints.json, unsigned_ints.json, small_signed_ints.json, citm_catalog.json and twitter.json, for both the contiguous and the streaming scanner. parse, streaming develop before after canada.json 19.4ms 20.5ms 19.3ms floats.json 135.9ms 131.8ms 128.0ms parse, contiguous develop before after canada.json 15.5ms 12.9ms 11.7ms floats.json 98.6ms 69.8ms 66.7ms Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> | 6 天前 | |
Only reserve array capacity if the array type supports it (#5522) | 2 天前 | |
Speed up parsing of contiguous input (numbers, strings, UTF-8) (#5283) * Speed up number parsing in the lexer (fast paths from the fast_float/simdjson world) The number scanner converted its already-validated digit buffer with std::strtoull/std::strtoll/std::strtod. Those pull in locale and errno machinery and dominate number-heavy parsing (strtod runs at ~6 M/s). Replace them with dedicated parsers over the validated buffer: - parse_integer_unsigned / parse_integer_signed: accumulate digits with overflow detection, falling back to the float path on overflow exactly as the strtoull/strtoll round-trip check did. Overflow behavior is unchanged for narrower or wider custom number types. - parse_float_fast: Clinger's exact fast path for `double` (<=19 significant digits, |exp10| <= 22, significand < 2^53), where significand * 10^exp is exact under IEEE round-to-nearest. This is the same fast path used by fast_float/simdjson. It is bit-identical to strtod on this subset and declines (falling back to strtod) otherwise. Only `double` uses it; float and long double keep std::strtof/std::strtold via a templated overload. Measured on representative data (g++ 13, -O3): - integers: DOM parse +11%, SAX +25-34% - floats: DOM parse +37%, SAX +70% (clang: float DOM ~1.9x) No dependencies added; header-only and C++11-clean. Existing parser, lexer, conversion and deserialization unit tests pass unchanged; a 3M-value random-double fuzz matches strtod bit-for-bit. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add SWAR bulk string scanning for contiguous input (simdjson-style) scan_string() read the input one character at a time through the input adapter and classified every byte with a large switch. For contiguous byte buffers we can instead scan 8 bytes at a time with a SWAR word test that finds the first byte needing individual handling (the closing quote, an escape, a control character, or a non-ASCII UTF-8 byte) and bulk-append the ordinary run in one go. - input adapters expose supports_bulk_scan / bulk_data / bulk_remaining / bulk_skip for provably-contiguous, same-type, 1-byte iterator ranges (raw pointers in every standard; std::string/std::vector/std::array and friends additionally in C++20 via std::contiguous_iterator). - the lexer gains a bulk_scan capability (gated on lazy_token_string so bypassing the per-character capture cannot lose error diagnostics) and a scan_string_bulk() fast path; streaming/wide/user adapters are unchanged and keep the byte-at-a-time scanner. The run contains no newline (all bytes < 0x20 are treated as special), so position bookkeeping stays exact, and error tokens are still reconstructed lazily from the consumed byte range. The SWAR special-byte test is pure uint64_t arithmetic - no intrinsics, no runtime dispatch, C++11-clean. Measured on representative data, pointer input, g++ 13 -O3 (string values discarded by accept() see the largest gains): long ASCII strings: DOM +4.5x, SAX +14x, accept +17x (to ~2 GB/s) short strings: DOM +15%, SAX +62%, accept +85% escape-heavy: DOM +31%, SAX +26%, accept +28% Same-input parity verified: 200k randomized documents (escapes, multibyte UTF-8, surrogate pairs) accept/parse identically via the contiguous SWAR path and the streaming byte path; unit lexer/parser/diagnostic-position/ deserialization/conversions suites pass unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Validate UTF-8 in the bulk string scanner (portable ~2-3x on non-ASCII text) The SWAR bulk string path stopped at the first non-ASCII byte and handed every multibyte character to the byte-at-a-time scanner, whose per-byte get()/next_byte_in_range()/add() machinery runs at roughly half the speed of validating straight from the buffer. As a result, dense non-ASCII text (CJK, emoji, accented Latin) parsed ~10-15x slower than ASCII. Fold well-formed UTF-8 into the bulk run: scan_string_bulk() now, on a non-ASCII lead byte, validates one sequence with validate_one_utf8() - which mirrors scan_string()'s per-byte switch ranges exactly (rejecting overlong forms, surrogates, and out-of-range code points) - and appends it in place, continuing until the closing quote, an escape, a control byte, or an ill-formed sequence. All error handling still defers to the byte path, so error messages and positions are byte-for-byte unchanged. Because only well-formed content is fast-pathed and every rejection falls through to the existing scanner, behavior is identical; the win is purely throughput. Measured on pointer input (accept, string values discarded): content g++ 13 clang 18 dense CJK 277 -> 648 ~605 MB/s (~2.3x) dense emoji 299 -> 857 ~702 MB/s (~2.6-2.9x) mixed 90% ASCII 246 -> 331 ~334 MB/s (~1.35x) pure ASCII unchanged (~3.2 / 4.1 GB/s) Verified: 2,000,000 randomized documents built from arbitrary bytes (overlong, surrogate, truncated, out-of-range sequences) accept/reject and parse identically via the contiguous path and the streaming byte path; lexer/parser/diagnostic-position/deserialization/conversions suites pass unchanged. Pure C++11, no intrinsics. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Route contiguous byte containers through the pointer adapter (fast paths in C++11) json::parse(std::string) - the most common entry point - did not benefit from the contiguous fast paths (bulk string scanning, UTF-8 bulk validation, memcpy for binary formats) in C++11..17: std::string::iterator is a library wrapper, not a raw pointer, and pre-C++20 there is no portable way to prove it contiguous, so supports_bulk_scan was false. Only raw pointers, string literals, and C-arrays (and, in C++20, anything modelling std::contiguous_iterator) took the fast path. Detect contiguous single-byte containers (std::string, std::vector<char>, std::vector<std::uint8_t>, std::string_view, ...) via is_contiguous_byte_ container and route them through an iterator_input_adapter built from data()/data()+size(). The generic iterator-based container overload is constrained to exclude these, so the two overloads are disjoint and there is no ambiguity (a plain competing overload loses to the greedy forwarding-reference container overload on reference binding, and a factory partial-specialization is ambiguous - both were tried and rejected). The pointer keeps the container's own element type, so char_type - and therefore all parsing behavior - is byte-for-byte identical to the iterator path (const char* for std::string, const std::uint8_t* for std::vector<std::uint8_t>); only the raw pointer additionally turns on the fast paths. Lifetimes are unchanged: the container outlives the adapter for the full parse expression, exactly as the iterators it replaces did. Measured, C++11, json::parse/accept(std::string), g++ 13: long ASCII strings: accept 201 -> 3200 MB/s (~16x), parse 174 -> 1444 dense CJK: accept 263 -> 697 MB/s (~2.6x) short strings: accept 163 -> 243 MB/s (~1.5x) Verified: char_type preserved for std::string (char) and std::vector<std::uint8_t> (uint8_t); CBOR/MsgPack round-trips from std::vector<std::uint8_t> unchanged; 1,000,000 randomized documents accept and parse identically via std::string and via std::istream; deserialization/user-defined-input/parser/lexer/conversions/diagnostic- position suites pass (20,480 assertions); warning-clean on g++ and clang in C++11/17/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add optional simdutf backend for bulk UTF-8 validation (JSON_USE_SIMDUTF) The bulk string scanner validates UTF-8 straight from a contiguous buffer. The scalar validator caps at ~0.3-0.7 GB/s on non-ASCII text; a SIMD validator reaches several GB/s. Rather than hand-rolling SIMD UTF-8 validation (easy to get subtly wrong - a from-scratch SSE attempt rejected valid CJK), wire in the vetted simdutf library behind an opt-in switch. simdutf is not header-only (it ships simdutf.cpp and uses runtime CPU dispatch), so it is not vendored: defining JSON_USE_SIMDUTF includes <simdutf.h> and routes the bulk validator through simdutf::validate_utf8; the project supplies and links simdutf. Undefined (the default), nothing external is included and the portable C++11 scalar path is used, so the library stays header-only and its baseline behavior is unchanged. Design keeps behavior identical either way: - scan_string_bulk() now finds the run up to the next quote/escape/control byte (non-ASCII allowed) and validates it in one shot; on the rare validation failure it recomputes the exact valid prefix with the scalar helper, so ill-formed input still falls through to the byte path and is reported at the same position with the same message. - the per-sequence scalar path is factored into scalar_string_bulk_run() and is the default backend; the refactor is behavior-preserving and does not change scalar throughput. Verified: default and JSON_USE_SIMDUTF builds accept/reject/parse identically across 2,000,000 arbitrary-byte documents and 1,000,000 mixed-escape/UTF-8 documents (differential fuzz vs the streaming byte path); lexer/parser/diagnostic-position/deserialization suites pass under both configurations (20,188 assertions with the backend enabled); warning-clean on g++ and clang, C++11 and C++20, both configurations. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add a contiguous fast path for scanning numbers scan_number() reads a number one character at a time through the input adapter (get()) and appends each byte to token_buffer (add()) before converting. For contiguous input, the per-character get()/add() overhead dominates: it is roughly two thirds of the time spent on number-heavy parsing, far more than the value conversion itself. Add scan_number_bulk_contiguous(), which parses the whole number token straight from the input buffer: it validates and classifies the extent with the same grammar as scan_number()'s state machine, materializes token_buffer in one copy (substituting the locale decimal point exactly as scan_number() does), advances the adapter, and reuses the shared convert_number() tail. On anything it does not 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 then produces the exact diagnostic. Errors and their positions are therefore unchanged. The conversion tail is factored out of scan_number() into convert_number() so both scanners share it; the fast path is selected by tag dispatch on the existing bulk_scan capability, so streaming/wide/user adapters are unaffected. Measured on pointer input, g++ 13 -O3: - integers: parse +65%, accept +98% - floats: parse +39%, accept +70% Verified: 2,000,000 randomized number documents (including overflow-range integers, long digit strings and %.17g doubles) parse identically via the contiguous path and the streaming byte path, matching value, type and round-trip text; the locale suite and existing parser/lexer/conversions/ deserialization tests pass; a new "lexer number fast path" test checks contiguous-vs-streaming parity, token classification, and that malformed numbers are rejected identically. Pure C++11, no intrinsics. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Move byte-level scan/parse helpers out of lexer.hpp lexer.hpp had grown by ~600 lines of byte-level helpers that have no dependency on the lexer's template parameters and clutter the state machine. Move them, unchanged, into two focused headers as free functions in namespace detail: - number_parse.hpp: parse_integer_unsigned/parse_integer_signed (now templated on the number type) and parse_float_fast (Clinger's exact double fast path, with the decimal point passed as an argument instead of read from a lexer member). - string_scan.hpp: the SWAR string helpers (is_string_special, swar_string_special, find_string_special, validate_one_utf8, scalar_string_bulk_run) and the backend-dispatched string_bulk_run, including the optional simdutf include and find_string_delimiter. lexer.hpp now includes these and calls the free functions; the methods that touch lexer state (scan_string, scan_number, scan_string_bulk, scan_number_bulk_contiguous, convert_number) stay put. This is a pure code move with no behavior change: lexer.hpp drops from 2357 to 1934 lines, the now-unused <cstdint>/<cstring>/<limits> includes are removed, and the free-function form makes the SWAR helpers reusable elsewhere (e.g. the serializer's string escaping). Verified: default and JSON_USE_SIMDUTF builds compile; 2,000,000 number and 2,000,000 arbitrary-byte-string differential-fuzz documents parse identically to before; lexer/parser/conversions/deserialization/locale/ diagnostic-position suites pass (20,576 assertions); warning-clean on g++ and clang in C++11/17/20; the amalgamation regenerates and passes check-amalgamation. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix clang-tidy findings and document JSON_USE_SIMDUTF in the nav - number_parse.hpp: use std::array for the powers-of-ten table (avoid-c-arrays) and `auto` for the cast-initialized result (modernize-use-auto), matching the codebase style (cf. the serializer's utf8d table). Indexing casts keep the -Wsign-conversion build clean. - add JSON_USE_SIMDUTF to the mkdocs navigation so the macro page is reachable. No behavior change; clang-tidy is clean on the new headers and the amalgamation is regenerated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix number fast path for custom string types without assign() The contiguous number fast path materialized token_buffer with token_buffer.assign(data, len), but string_t is only required to provide the minimal interface the rest of the lexer uses (push_back, append, clear, operator[], ...). Custom string types such as the test's alt_string do not implement assign(), so scan_number_bulk_contiguous() failed to compile for them (unit-alt-string), breaking the gcc/clang standards and old-compiler CI jobs. reset() already clears token_buffer, so fill it with append() - which alt_string and std::string both provide and which the string fast path already relies on - instead of assign(). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Satisfy clang-tidy: parenthesize math and drop unused forwarding reference The CI clang-tidy (newer than the locally available version) reported two additional checks on the new code: - readability-math-missing-parentheses: parenthesize the (a * b) + c digit accumulations in number_parse.hpp. - cppcoreguidelines-missing-std-forward: the contiguous-byte-container input_adapter overload took a forwarding reference but only reads data()/size() and never forwards it. It is already disjoint from the generic container overload via SFINAE, so a plain const& is correct and clearer (and keeps the container alive for the whole parse just as before). No behavior change; char_type and routing are unchanged (std::string and std::vector<std::uint8_t> still take the pointer adapter with char/uint8_t char_type), CBOR/MsgPack round-trips and the 2M number fuzz still pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use std::from_chars (Eisel-Lemire) for float conversion when available The Clinger fast path is exact only for the "easy" subset (<=19 significant digits, |exp10| <= 22); high-precision and scientific floats fall through to strtod, where the failed Clinger attempt actually makes parsing a net loss. std::from_chars implements the Eisel-Lemire algorithm in modern standard libraries: locale-independent, correctly rounded, and fast over the whole value range. convert_number() now tries parse_float_from_chars() first (guarded by __cpp_lib_to_chars, so C++11 and libc++-without-float-support keep the Clinger + strtod path unchanged), then Clinger, then strtof. from_chars is used only when it consumes the entire token; a partial parse means a non-'.' locale decimal point, and an under-/overflow (result_out_of_range) also declines - in both cases the existing strtod fallback supplies the exact value and the well-defined +/-inf/0 the parser expects, side-stepping the P4168 divergence between implementations. float and long double now get the fast path too (Clinger was double-only). Measured, C++17, g++ 13 -O3, json::parse/accept: - canada-style floats: ~unchanged (Clinger already covered them) - high-precision (17 digits): parse 2.1x, accept 2.5x - scientific (17 digits + exp): parse 3.6x, accept 4.1x Verified: C++11 (Clinger/strtod) and C++17 (from_chars) parse every value - including subnormals, boundary values, and 1e9999/1e-9999 over-/underflow - to bit-identical results; 2M number-fuzz clean; conversions/deserialization/ locale/number-fast-path suites pass in both C++11 and C++17; clang-tidy clean; warning-clean on g++ and clang in C++11/17/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard from_chars use on JSON_HAS_CPP_17, not just __cpp_lib_to_chars libstdc++ 15 defines __cpp_lib_to_chars even in C++14 mode (via bits/version.h pulled in by other headers), but <charconv> is only included under JSON_HAS_CPP_17. That made parse_float_from_chars() reference std::from_chars without the header in C++14 builds, breaking gcc-latest, icpx, and the offline-testdata jobs. Gate the use on JSON_HAS_CPP_17 && __cpp_lib_to_chars so it matches the include condition exactly; C++11/14 always take the scalar fallback. Verified by forcing __cpp_lib_to_chars in a C++14 build: the guard suppresses std::from_chars and it compiles. C++17 behavior is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Disable Clinger float fast path under extended FP precision (x87) The contiguous number fast path uses a Clinger-style exact algorithm (significand * 10^scale in double arithmetic), which is only correctly rounded when double operations are evaluated in true 53-bit precision. On the x87 FPU used by 32-bit x86 (FLT_EVAL_METHOD == 2) the single multiply/divide is computed in 80-bit and then double-rounded to double, so a small fraction of values land 1 ULP off. This surfaced as test-cbor_cpp11 and test-msgpack_cpp11 failing on the mingw (x86) job for regression/floats.json: the C++17 builds pass because they take the correctly-rounded std::from_chars path, while C++11 falls back to parse_float_fast(). A 5M-sample check over shortest round-trip decimals reproduces it: 0 divergences with 53-bit doubles, ~1 in 25 000 with 80-bit intermediates; declining to std::strtod fixes all of them. Guard parse_float_fast() on FLT_EVAL_METHOD so it declines whenever the platform evaluates doubles in extended precision, letting the caller use the correctly-rounded std::from_chars / std::strtod path instead. On mainstream x86-64/ARM64 (FLT_EVAL_METHOD == 0) the fast path is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard <charconv> include with __has_include for GCC 7 GCC 7 sets __cplusplus to the C++17 value under -std=gnu++1z, so JSON_HAS_CPP_17 is defined, but its libstdc++ ships no <charconv> header (added in GCC 8; floating-point from_chars in GCC 11). The unconditional "#if defined(JSON_HAS_CPP_17) #include <charconv>" therefore failed to compile there: "fatal error: charconv: No such file or directory" in the ci_test_compilers_gcc (7) job. Wrap the include in __has_include(<charconv>), mirroring the library's existing handling of <version> and <filesystem> in macro_scope.hpp. When the header is absent, __cpp_lib_to_chars stays undefined and parse_float_from_chars() takes its scalar fallback, so the from_chars use site (already gated on __cpp_lib_to_chars) is never reached. GCC 8-10, which have <charconv> but no floating-point from_chars, are unaffected: they include the header but still take the fallback. GCC 11+ is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Restore the column when ungetting a newline The contiguous number fast path never reads the character that terminates a number token, while scan_number() reads it and then ungets it. When that character is a newline, get() has already cleared chars_read_current_line, and unget() could only restore lines_read - leaving the column at 0. The two paths therefore reported different columns for the same document: json::parse("[01\n]") -> line 1, column 3 json::parse(stringstream) -> line 1, column 0 Remember the column the newline was read at so unget() can restore it. Both paths now report the position the offending token actually starts at, which also fixes the pre-existing column-0 artifact for streaming input. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Extend the bulk scan fast paths to sized sentinels supports_bulk_scan required IteratorType and SentinelType to be the same type, which excluded std::counted_iterator paired with std::default_sentinel_t - the combination #5268 had already enabled for the memcpy fast path. Such input fell back to the byte-at-a-time scanner even though it is contiguous and its remaining length is computable in O(1). Factor the "distance is computable in O(1)" test into sentinel_is_sized and use it for iterator_is_contiguous, supports_seek, and supports_bulk_scan alike, and share the std::ranges::distance/std::distance dispatch through a remaining_count() helper. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Document JSON_USE_SIMDUTF on the macro overview page The macro was only listed in the API macro index; add it to the supported macros overview alongside the other JSON_USE_* macros, and note that it selects between two definitions of the same inline function and so must be defined identically in every translation unit. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cover the counted-iterator bulk scan paths Sized sentinels newly reach the bulk string/number scanners and the seek-based token reconstruction, so exercise both: - diagnostics that quote the offending token, which are rebuilt from the consumed input via copy_consumed_range() - inputs whose count ends before the underlying buffer does, including a closing quote that exists only behind the count, a cut inside an 8-byte SWAR stride, and a cut inside a UTF-8 sequence Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix the SentinelType example on iterator_input_adapter The comment offered "a C++20 sentinel or counted_iterator" as examples of a SentinelType, but std::counted_iterator is the IteratorType - the sentinel it pairs with is std::default_sentinel_t. #5268 corrected the same wording in the API documentation and left the code comment behind. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not discard the parse result in the error position check json::parse is declared warn_unused_result, and CHECK_THROWS_WITH_AS evaluates its expression as a discarded statement, so the assertion broke the -Werror builds (GCC -Werror=unused-result, MSVC C4834 under /WX). Compare against the helper that already captures the message instead. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Lock the two number grammars together with a parity test The JSON number grammar is encoded twice: as the scan_number() state machine and as the contiguous fast path. The fast path declining on anything it does not recognize keeps most divergence harmless, but if it ever accepted something the state machine rejects the result would be a silent correctness bug, and the existing test only pinned a hand-written list of numbers. Enumerate every string of length 1..4 over "01.eE+-" (2800 tokens) and require both paths to agree on the parsed value and on the exact error message. Verified to fail if the fast path's grammar is perturbed. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Skip token_buffer for integers on the contiguous path An integer token does not need token_buffer: the number_integer and number_unsigned SAX callbacks take only the value, and the overflow diagnostic rebuilds the text from the input via get_token_string(). Convert straight from the input buffer and materialize the token only for the floating-point tail, which still needs a NUL-terminated buffer for strtod. JSON_DIAGNOSTIC_POSITIONS derives a number's start position from get_string().size(), so the copy is kept when that is enabled. The integer dispatch is factored into convert_integer() and shared with convert_number(), so both scanners keep using one implementation. Integer-heavy input, 400k values, -O3: parse accept gcc 16 +14% +23% clang +15% +22% Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cover the bulk string and UTF-8 scanners These paths had no dedicated tests and rested on differential fuzzing only. Add three sections, all comparing the contiguous scanner against the byte-at-a-time one on the parsed value and on the exact error message: - every string of length 1..3 over an alphabet of ordinary ASCII, both specials, a control byte, escape characters, UTF-8 lead and continuation bytes, and a byte that is never valid - each at offset 0 and offset 9, so the bulk scanner sees them with and without a run behind them - every kind of run-ending byte at each offset across two 8-byte SWAR words, so multibyte sequences also straddle the word boundary - the boundaries of every range validate_one_utf8() recognizes: shortest and longest encodings, overlongs, both ends of the surrogate block, U+10FFFF and just past it, and truncated sequences Verified to fail if the bulk validator accepts surrogates, and if the SWAR word test stops detecting control characters. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Format the new string fast path test with astyle The pinned astyle expands a braced-init-list used as a range-for range onto several lines; hoist the two offsets into a named vector instead, which reads better and leaves nothing for astyle to reformat. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix shadowed locals and guard the exception-dependent tests Two problems in the tests added for the bulk scanners, both found by CI: - the inner `const json j` in the counted-iterator diagnostics shadowed the one declared at test-case scope, which -Wshadow rejects on GCC and clang and C4456 rejects on MSVC under /WX; rename them - the new parity checks parse deliberately invalid input, which calls std::abort() rather than throwing when JSON_NOEXCEPTION is defined, so they would have crashed the no-exception build; guard them the way the other tests do json::accept() does not abort, so the UTF-8 range assertions stay compiled without exceptions and keep covering validate_one_utf8() there. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Gate the C++20 iterator classification on JSON_HAS_RANGES Making supports_bulk_scan depend on iterator_is_contiguous meant the trait is now instantiated for every adapter, not only when get_elements() is called. On standard libraries with an incomplete <ranges> that is fatal: libstdc++ 10 evaluates std::contiguous_iterator<std::counted_iterator<T*>> by calling std::to_address, which needs an operator-> its counted_iterator does not have, so satisfaction checking is a hard error rather than false. Reported by clang 14 + libstdc++ 10. JSON_HAS_RANGES already encodes exactly this ("libstdc++ < 11 has incomplete C++20 ranges", #4440), so require it for the C++20 branch. Affected toolchains fall back to the pointer-only test and the byte-at-a-time scanner, which parses identically, just without the bulk fast paths. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Satisfy clang-tidy in the new bulk scanner tests - give the helper lambdas an explicit std::string return type and return braced initializer lists (modernize-return-braced-init-list) - replace the C-style array of test cases with a std::vector (modernize-avoid-c-arrays) - silence pro-type-member-init on the two brace-initialized aggregates; default member initializers would stop them being aggregates in C++11 Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Avoid escaped literals in the counted-iterator diagnostics list clang-tidy reads "[\"\\ud834\"]" as a literal better written raw, and the two literals written next to each other in "[\"a\x01""b\"]" as a missing comma. The concatenation was there to stop the hex escape swallowing the following character; build those documents from explicit bytes instead and use raw strings elsewhere. The byte sequences are unchanged. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Reattach convert_number's documentation @gregmarr spotted that convert_integer() was inserted between convert_number() and its doc block, leaving convert_integer() with two stacked blocks and convert_number() with none. Comment only; no code change. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Address review comments on the simdutf backend string_bulk_run() had the same `return scalar_string_bulk_run(...)` in both arms of the `#if`. The simdutf arm already falls through when validation fails, so a single return after the `#endif` says the same thing. The JSON_USE_SIMDUTF example showed `#include <simdutf.h>`, which string_scan.hpp already does under the same guard; users only have to put the header on the include path and link the library, not include it themselves. Set the version history entry to 3.13.0, matching the other macro pages documenting unreleased features. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Pin the number error position against a non-newline terminator The comment claimed the reported column is the one the offending token starts at. It is the column reached after the token's last character - which is the actual point of the unget() change: a number terminated by a newline now reports what the same number terminated by a space always did. Assert that equality directly, and add a multi-character token where the start and end columns differ, so the invariant cannot be read off a single-character example. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not repeat the integer conversion that just failed scan_number_bulk_contiguous() converts an integer token straight from the input buffer. When the value does not fit, it materializes token_buffer and calls convert_number(), which tried the very same integer conversion again before falling back to floating point. Recording the outcome in number_type skips the second attempt. The resulting token type and value are unchanged: convert_number() reached the float tail either way. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Require a container's value_type to match what data() points at is_contiguous_byte_container accepted any type with a data() returning a pointer to a single-byte integral plus a size(). That is duck typing: the two members say nothing about size() counting the units data() points at. A type where it does not - fixed-size records, say - was routed to the pointer-based adapter and parsed as [data(), data() + size()) bytes, silently truncating input the iterator-based adapter had read in full: struct record_buffer { using value_type = std::array<char, 4>; std::string bytes; const char* data() const; // raw bytes std::size_t size() const; // in records const char* begin() const; const char* end() const; }; json::parse(record_buffer{"[1,2,3,4,5]"}); // parse error at column 3 Requiring the container's own value_type to be that same element type ties the two together. Every contiguous standard container satisfies it, so std::string, std::vector<char>, std::array<char, N> and std::string_view keep the fast path; anything else falls back to the iterator-based adapter, which is always correct. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Only compile the simdutf backend from C++17 on simdutf.h rejects anything below C++17 with an #error, so defining JSON_USE_SIMDUTF in a C++11 or C++14 translation unit did not fail with a message about simdutf being unavailable - it failed to compile at all, taking the library's C++11 support with it. Nothing caught this because no build ever compiled that path. Gate the include and both uses on JSON_HAS_CPP_17, the same way number_parse.hpp gates std::from_chars. Below C++17 the macro now has no effect and the scalar validator runs; it accepts and rejects exactly the same input, so the macro is safe to set project-wide even when some translation units use an older standard. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Test the simdutf backend in CI JSON_USE_SIMDUTF was documented and shipped but never compiled by anything in the repository, so nothing held the backend to the behavior the docs promise. Add JSON_TestSimdutf (OFF by default), which fetches simdutf and defines JSON_USE_SIMDUTF for every test target, and a ci_test_simdutf target that runs the whole suite in that configuration. Because simdutf needs C++17, the suite is built at C++11 as well, so one job covers both the scalar fallback with the macro defined and simdutf itself. The dependency hangs off test_main, whose usage requirements every test target inherits. The library target and the installed CMake package are deliberately untouched: making nlohmann_json link simdutf would put a find_dependency() in the exported package, which is a separate decision. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Warn when JSON_TestSimdutf cannot reach the simdutf backend simdutf needs C++17: without it the dependency does not even compile, and with a C++17 compiler but no C++17-or-later standard under test it builds and then goes unused. Either way the option silently did nothing useful, or broke the configure step outright. Resolve the tested standards first, then check them: when none of them can reach simdutf, skip the dependency and say so, naming which of the two reasons applies and how to fix it. The tests then run against the scalar validator, which is what would have happened anyway. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use record_buffer::data() so clang does not flag it unneeded The record_buffer test type declares data() and size() so the is_contiguous_byte_container trait can see both and still reject the type on its value_type. data() was never called, so clang's -Wunneeded-member-function (under -Weverything -Werror) failed the C++20 build. Assert that data() points at the underlying bytes: it ODR-uses the member and documents the property the type is meant to demonstrate. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Skip the float fast path when it cannot succeed parse_float_fast() (Clinger) needs a significand below 2^53, so it always declines once the mantissa has 17 or more significant digits. convert_number() called it unconditionally, so those numbers were walked an extra time before strtod had to run anyway. On streaming input, where scanning is byte-at-a-time and there is no compensating win, that made canada.json about 6% slower than develop. Derive the significant-digit count from token_buffer indices - the digits are not scanned again - and skip the call when it is guaranteed to decline. Both scanners pass the offset where the mantissa ends; the count only has to be corrected for a leading "0", which the JSON grammar admits nowhere else. The integer path returns before the check, so integer-heavy input is unaffected. Values are unchanged: this only avoids an attempt that would have failed. Verified bit-exact against develop over every number in canada.json, floats.json, signed_ints.json, unsigned_ints.json, small_signed_ints.json, citm_catalog.json and twitter.json, for both the contiguous and the streaming scanner. parse, streaming develop before after canada.json 19.4ms 20.5ms 19.3ms floats.json 135.9ms 131.8ms 128.0ms parse, contiguous develop before after canada.json 15.5ms 12.9ms 11.7ms floats.json 98.6ms 69.8ms 66.7ms Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> | 6 天前 | |
Speed up parsing of contiguous input (numbers, strings, UTF-8) (#5283) * Speed up number parsing in the lexer (fast paths from the fast_float/simdjson world) The number scanner converted its already-validated digit buffer with std::strtoull/std::strtoll/std::strtod. Those pull in locale and errno machinery and dominate number-heavy parsing (strtod runs at ~6 M/s). Replace them with dedicated parsers over the validated buffer: - parse_integer_unsigned / parse_integer_signed: accumulate digits with overflow detection, falling back to the float path on overflow exactly as the strtoull/strtoll round-trip check did. Overflow behavior is unchanged for narrower or wider custom number types. - parse_float_fast: Clinger's exact fast path for `double` (<=19 significant digits, |exp10| <= 22, significand < 2^53), where significand * 10^exp is exact under IEEE round-to-nearest. This is the same fast path used by fast_float/simdjson. It is bit-identical to strtod on this subset and declines (falling back to strtod) otherwise. Only `double` uses it; float and long double keep std::strtof/std::strtold via a templated overload. Measured on representative data (g++ 13, -O3): - integers: DOM parse +11%, SAX +25-34% - floats: DOM parse +37%, SAX +70% (clang: float DOM ~1.9x) No dependencies added; header-only and C++11-clean. Existing parser, lexer, conversion and deserialization unit tests pass unchanged; a 3M-value random-double fuzz matches strtod bit-for-bit. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add SWAR bulk string scanning for contiguous input (simdjson-style) scan_string() read the input one character at a time through the input adapter and classified every byte with a large switch. For contiguous byte buffers we can instead scan 8 bytes at a time with a SWAR word test that finds the first byte needing individual handling (the closing quote, an escape, a control character, or a non-ASCII UTF-8 byte) and bulk-append the ordinary run in one go. - input adapters expose supports_bulk_scan / bulk_data / bulk_remaining / bulk_skip for provably-contiguous, same-type, 1-byte iterator ranges (raw pointers in every standard; std::string/std::vector/std::array and friends additionally in C++20 via std::contiguous_iterator). - the lexer gains a bulk_scan capability (gated on lazy_token_string so bypassing the per-character capture cannot lose error diagnostics) and a scan_string_bulk() fast path; streaming/wide/user adapters are unchanged and keep the byte-at-a-time scanner. The run contains no newline (all bytes < 0x20 are treated as special), so position bookkeeping stays exact, and error tokens are still reconstructed lazily from the consumed byte range. The SWAR special-byte test is pure uint64_t arithmetic - no intrinsics, no runtime dispatch, C++11-clean. Measured on representative data, pointer input, g++ 13 -O3 (string values discarded by accept() see the largest gains): long ASCII strings: DOM +4.5x, SAX +14x, accept +17x (to ~2 GB/s) short strings: DOM +15%, SAX +62%, accept +85% escape-heavy: DOM +31%, SAX +26%, accept +28% Same-input parity verified: 200k randomized documents (escapes, multibyte UTF-8, surrogate pairs) accept/parse identically via the contiguous SWAR path and the streaming byte path; unit lexer/parser/diagnostic-position/ deserialization/conversions suites pass unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Validate UTF-8 in the bulk string scanner (portable ~2-3x on non-ASCII text) The SWAR bulk string path stopped at the first non-ASCII byte and handed every multibyte character to the byte-at-a-time scanner, whose per-byte get()/next_byte_in_range()/add() machinery runs at roughly half the speed of validating straight from the buffer. As a result, dense non-ASCII text (CJK, emoji, accented Latin) parsed ~10-15x slower than ASCII. Fold well-formed UTF-8 into the bulk run: scan_string_bulk() now, on a non-ASCII lead byte, validates one sequence with validate_one_utf8() - which mirrors scan_string()'s per-byte switch ranges exactly (rejecting overlong forms, surrogates, and out-of-range code points) - and appends it in place, continuing until the closing quote, an escape, a control byte, or an ill-formed sequence. All error handling still defers to the byte path, so error messages and positions are byte-for-byte unchanged. Because only well-formed content is fast-pathed and every rejection falls through to the existing scanner, behavior is identical; the win is purely throughput. Measured on pointer input (accept, string values discarded): content g++ 13 clang 18 dense CJK 277 -> 648 ~605 MB/s (~2.3x) dense emoji 299 -> 857 ~702 MB/s (~2.6-2.9x) mixed 90% ASCII 246 -> 331 ~334 MB/s (~1.35x) pure ASCII unchanged (~3.2 / 4.1 GB/s) Verified: 2,000,000 randomized documents built from arbitrary bytes (overlong, surrogate, truncated, out-of-range sequences) accept/reject and parse identically via the contiguous path and the streaming byte path; lexer/parser/diagnostic-position/deserialization/conversions suites pass unchanged. Pure C++11, no intrinsics. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Route contiguous byte containers through the pointer adapter (fast paths in C++11) json::parse(std::string) - the most common entry point - did not benefit from the contiguous fast paths (bulk string scanning, UTF-8 bulk validation, memcpy for binary formats) in C++11..17: std::string::iterator is a library wrapper, not a raw pointer, and pre-C++20 there is no portable way to prove it contiguous, so supports_bulk_scan was false. Only raw pointers, string literals, and C-arrays (and, in C++20, anything modelling std::contiguous_iterator) took the fast path. Detect contiguous single-byte containers (std::string, std::vector<char>, std::vector<std::uint8_t>, std::string_view, ...) via is_contiguous_byte_ container and route them through an iterator_input_adapter built from data()/data()+size(). The generic iterator-based container overload is constrained to exclude these, so the two overloads are disjoint and there is no ambiguity (a plain competing overload loses to the greedy forwarding-reference container overload on reference binding, and a factory partial-specialization is ambiguous - both were tried and rejected). The pointer keeps the container's own element type, so char_type - and therefore all parsing behavior - is byte-for-byte identical to the iterator path (const char* for std::string, const std::uint8_t* for std::vector<std::uint8_t>); only the raw pointer additionally turns on the fast paths. Lifetimes are unchanged: the container outlives the adapter for the full parse expression, exactly as the iterators it replaces did. Measured, C++11, json::parse/accept(std::string), g++ 13: long ASCII strings: accept 201 -> 3200 MB/s (~16x), parse 174 -> 1444 dense CJK: accept 263 -> 697 MB/s (~2.6x) short strings: accept 163 -> 243 MB/s (~1.5x) Verified: char_type preserved for std::string (char) and std::vector<std::uint8_t> (uint8_t); CBOR/MsgPack round-trips from std::vector<std::uint8_t> unchanged; 1,000,000 randomized documents accept and parse identically via std::string and via std::istream; deserialization/user-defined-input/parser/lexer/conversions/diagnostic- position suites pass (20,480 assertions); warning-clean on g++ and clang in C++11/17/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add optional simdutf backend for bulk UTF-8 validation (JSON_USE_SIMDUTF) The bulk string scanner validates UTF-8 straight from a contiguous buffer. The scalar validator caps at ~0.3-0.7 GB/s on non-ASCII text; a SIMD validator reaches several GB/s. Rather than hand-rolling SIMD UTF-8 validation (easy to get subtly wrong - a from-scratch SSE attempt rejected valid CJK), wire in the vetted simdutf library behind an opt-in switch. simdutf is not header-only (it ships simdutf.cpp and uses runtime CPU dispatch), so it is not vendored: defining JSON_USE_SIMDUTF includes <simdutf.h> and routes the bulk validator through simdutf::validate_utf8; the project supplies and links simdutf. Undefined (the default), nothing external is included and the portable C++11 scalar path is used, so the library stays header-only and its baseline behavior is unchanged. Design keeps behavior identical either way: - scan_string_bulk() now finds the run up to the next quote/escape/control byte (non-ASCII allowed) and validates it in one shot; on the rare validation failure it recomputes the exact valid prefix with the scalar helper, so ill-formed input still falls through to the byte path and is reported at the same position with the same message. - the per-sequence scalar path is factored into scalar_string_bulk_run() and is the default backend; the refactor is behavior-preserving and does not change scalar throughput. Verified: default and JSON_USE_SIMDUTF builds accept/reject/parse identically across 2,000,000 arbitrary-byte documents and 1,000,000 mixed-escape/UTF-8 documents (differential fuzz vs the streaming byte path); lexer/parser/diagnostic-position/deserialization suites pass under both configurations (20,188 assertions with the backend enabled); warning-clean on g++ and clang, C++11 and C++20, both configurations. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Add a contiguous fast path for scanning numbers scan_number() reads a number one character at a time through the input adapter (get()) and appends each byte to token_buffer (add()) before converting. For contiguous input, the per-character get()/add() overhead dominates: it is roughly two thirds of the time spent on number-heavy parsing, far more than the value conversion itself. Add scan_number_bulk_contiguous(), which parses the whole number token straight from the input buffer: it validates and classifies the extent with the same grammar as scan_number()'s state machine, materializes token_buffer in one copy (substituting the locale decimal point exactly as scan_number() does), advances the adapter, and reuses the shared convert_number() tail. On anything it does not 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 then produces the exact diagnostic. Errors and their positions are therefore unchanged. The conversion tail is factored out of scan_number() into convert_number() so both scanners share it; the fast path is selected by tag dispatch on the existing bulk_scan capability, so streaming/wide/user adapters are unaffected. Measured on pointer input, g++ 13 -O3: - integers: parse +65%, accept +98% - floats: parse +39%, accept +70% Verified: 2,000,000 randomized number documents (including overflow-range integers, long digit strings and %.17g doubles) parse identically via the contiguous path and the streaming byte path, matching value, type and round-trip text; the locale suite and existing parser/lexer/conversions/ deserialization tests pass; a new "lexer number fast path" test checks contiguous-vs-streaming parity, token classification, and that malformed numbers are rejected identically. Pure C++11, no intrinsics. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Move byte-level scan/parse helpers out of lexer.hpp lexer.hpp had grown by ~600 lines of byte-level helpers that have no dependency on the lexer's template parameters and clutter the state machine. Move them, unchanged, into two focused headers as free functions in namespace detail: - number_parse.hpp: parse_integer_unsigned/parse_integer_signed (now templated on the number type) and parse_float_fast (Clinger's exact double fast path, with the decimal point passed as an argument instead of read from a lexer member). - string_scan.hpp: the SWAR string helpers (is_string_special, swar_string_special, find_string_special, validate_one_utf8, scalar_string_bulk_run) and the backend-dispatched string_bulk_run, including the optional simdutf include and find_string_delimiter. lexer.hpp now includes these and calls the free functions; the methods that touch lexer state (scan_string, scan_number, scan_string_bulk, scan_number_bulk_contiguous, convert_number) stay put. This is a pure code move with no behavior change: lexer.hpp drops from 2357 to 1934 lines, the now-unused <cstdint>/<cstring>/<limits> includes are removed, and the free-function form makes the SWAR helpers reusable elsewhere (e.g. the serializer's string escaping). Verified: default and JSON_USE_SIMDUTF builds compile; 2,000,000 number and 2,000,000 arbitrary-byte-string differential-fuzz documents parse identically to before; lexer/parser/conversions/deserialization/locale/ diagnostic-position suites pass (20,576 assertions); warning-clean on g++ and clang in C++11/17/20; the amalgamation regenerates and passes check-amalgamation. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix clang-tidy findings and document JSON_USE_SIMDUTF in the nav - number_parse.hpp: use std::array for the powers-of-ten table (avoid-c-arrays) and `auto` for the cast-initialized result (modernize-use-auto), matching the codebase style (cf. the serializer's utf8d table). Indexing casts keep the -Wsign-conversion build clean. - add JSON_USE_SIMDUTF to the mkdocs navigation so the macro page is reachable. No behavior change; clang-tidy is clean on the new headers and the amalgamation is regenerated. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix number fast path for custom string types without assign() The contiguous number fast path materialized token_buffer with token_buffer.assign(data, len), but string_t is only required to provide the minimal interface the rest of the lexer uses (push_back, append, clear, operator[], ...). Custom string types such as the test's alt_string do not implement assign(), so scan_number_bulk_contiguous() failed to compile for them (unit-alt-string), breaking the gcc/clang standards and old-compiler CI jobs. reset() already clears token_buffer, so fill it with append() - which alt_string and std::string both provide and which the string fast path already relies on - instead of assign(). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Satisfy clang-tidy: parenthesize math and drop unused forwarding reference The CI clang-tidy (newer than the locally available version) reported two additional checks on the new code: - readability-math-missing-parentheses: parenthesize the (a * b) + c digit accumulations in number_parse.hpp. - cppcoreguidelines-missing-std-forward: the contiguous-byte-container input_adapter overload took a forwarding reference but only reads data()/size() and never forwards it. It is already disjoint from the generic container overload via SFINAE, so a plain const& is correct and clearer (and keeps the container alive for the whole parse just as before). No behavior change; char_type and routing are unchanged (std::string and std::vector<std::uint8_t> still take the pointer adapter with char/uint8_t char_type), CBOR/MsgPack round-trips and the 2M number fuzz still pass. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use std::from_chars (Eisel-Lemire) for float conversion when available The Clinger fast path is exact only for the "easy" subset (<=19 significant digits, |exp10| <= 22); high-precision and scientific floats fall through to strtod, where the failed Clinger attempt actually makes parsing a net loss. std::from_chars implements the Eisel-Lemire algorithm in modern standard libraries: locale-independent, correctly rounded, and fast over the whole value range. convert_number() now tries parse_float_from_chars() first (guarded by __cpp_lib_to_chars, so C++11 and libc++-without-float-support keep the Clinger + strtod path unchanged), then Clinger, then strtof. from_chars is used only when it consumes the entire token; a partial parse means a non-'.' locale decimal point, and an under-/overflow (result_out_of_range) also declines - in both cases the existing strtod fallback supplies the exact value and the well-defined +/-inf/0 the parser expects, side-stepping the P4168 divergence between implementations. float and long double now get the fast path too (Clinger was double-only). Measured, C++17, g++ 13 -O3, json::parse/accept: - canada-style floats: ~unchanged (Clinger already covered them) - high-precision (17 digits): parse 2.1x, accept 2.5x - scientific (17 digits + exp): parse 3.6x, accept 4.1x Verified: C++11 (Clinger/strtod) and C++17 (from_chars) parse every value - including subnormals, boundary values, and 1e9999/1e-9999 over-/underflow - to bit-identical results; 2M number-fuzz clean; conversions/deserialization/ locale/number-fast-path suites pass in both C++11 and C++17; clang-tidy clean; warning-clean on g++ and clang in C++11/17/20. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard from_chars use on JSON_HAS_CPP_17, not just __cpp_lib_to_chars libstdc++ 15 defines __cpp_lib_to_chars even in C++14 mode (via bits/version.h pulled in by other headers), but <charconv> is only included under JSON_HAS_CPP_17. That made parse_float_from_chars() reference std::from_chars without the header in C++14 builds, breaking gcc-latest, icpx, and the offline-testdata jobs. Gate the use on JSON_HAS_CPP_17 && __cpp_lib_to_chars so it matches the include condition exactly; C++11/14 always take the scalar fallback. Verified by forcing __cpp_lib_to_chars in a C++14 build: the guard suppresses std::from_chars and it compiles. C++17 behavior is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Disable Clinger float fast path under extended FP precision (x87) The contiguous number fast path uses a Clinger-style exact algorithm (significand * 10^scale in double arithmetic), which is only correctly rounded when double operations are evaluated in true 53-bit precision. On the x87 FPU used by 32-bit x86 (FLT_EVAL_METHOD == 2) the single multiply/divide is computed in 80-bit and then double-rounded to double, so a small fraction of values land 1 ULP off. This surfaced as test-cbor_cpp11 and test-msgpack_cpp11 failing on the mingw (x86) job for regression/floats.json: the C++17 builds pass because they take the correctly-rounded std::from_chars path, while C++11 falls back to parse_float_fast(). A 5M-sample check over shortest round-trip decimals reproduces it: 0 divergences with 53-bit doubles, ~1 in 25 000 with 80-bit intermediates; declining to std::strtod fixes all of them. Guard parse_float_fast() on FLT_EVAL_METHOD so it declines whenever the platform evaluates doubles in extended precision, letting the caller use the correctly-rounded std::from_chars / std::strtod path instead. On mainstream x86-64/ARM64 (FLT_EVAL_METHOD == 0) the fast path is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Guard <charconv> include with __has_include for GCC 7 GCC 7 sets __cplusplus to the C++17 value under -std=gnu++1z, so JSON_HAS_CPP_17 is defined, but its libstdc++ ships no <charconv> header (added in GCC 8; floating-point from_chars in GCC 11). The unconditional "#if defined(JSON_HAS_CPP_17) #include <charconv>" therefore failed to compile there: "fatal error: charconv: No such file or directory" in the ci_test_compilers_gcc (7) job. Wrap the include in __has_include(<charconv>), mirroring the library's existing handling of <version> and <filesystem> in macro_scope.hpp. When the header is absent, __cpp_lib_to_chars stays undefined and parse_float_from_chars() takes its scalar fallback, so the from_chars use site (already gated on __cpp_lib_to_chars) is never reached. GCC 8-10, which have <charconv> but no floating-point from_chars, are unaffected: they include the header but still take the fallback. GCC 11+ is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AXcDtEma2PjxgmPS9cQGzA Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Restore the column when ungetting a newline The contiguous number fast path never reads the character that terminates a number token, while scan_number() reads it and then ungets it. When that character is a newline, get() has already cleared chars_read_current_line, and unget() could only restore lines_read - leaving the column at 0. The two paths therefore reported different columns for the same document: json::parse("[01\n]") -> line 1, column 3 json::parse(stringstream) -> line 1, column 0 Remember the column the newline was read at so unget() can restore it. Both paths now report the position the offending token actually starts at, which also fixes the pre-existing column-0 artifact for streaming input. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Extend the bulk scan fast paths to sized sentinels supports_bulk_scan required IteratorType and SentinelType to be the same type, which excluded std::counted_iterator paired with std::default_sentinel_t - the combination #5268 had already enabled for the memcpy fast path. Such input fell back to the byte-at-a-time scanner even though it is contiguous and its remaining length is computable in O(1). Factor the "distance is computable in O(1)" test into sentinel_is_sized and use it for iterator_is_contiguous, supports_seek, and supports_bulk_scan alike, and share the std::ranges::distance/std::distance dispatch through a remaining_count() helper. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Document JSON_USE_SIMDUTF on the macro overview page The macro was only listed in the API macro index; add it to the supported macros overview alongside the other JSON_USE_* macros, and note that it selects between two definitions of the same inline function and so must be defined identically in every translation unit. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cover the counted-iterator bulk scan paths Sized sentinels newly reach the bulk string/number scanners and the seek-based token reconstruction, so exercise both: - diagnostics that quote the offending token, which are rebuilt from the consumed input via copy_consumed_range() - inputs whose count ends before the underlying buffer does, including a closing quote that exists only behind the count, a cut inside an 8-byte SWAR stride, and a cut inside a UTF-8 sequence Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix the SentinelType example on iterator_input_adapter The comment offered "a C++20 sentinel or counted_iterator" as examples of a SentinelType, but std::counted_iterator is the IteratorType - the sentinel it pairs with is std::default_sentinel_t. #5268 corrected the same wording in the API documentation and left the code comment behind. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not discard the parse result in the error position check json::parse is declared warn_unused_result, and CHECK_THROWS_WITH_AS evaluates its expression as a discarded statement, so the assertion broke the -Werror builds (GCC -Werror=unused-result, MSVC C4834 under /WX). Compare against the helper that already captures the message instead. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Lock the two number grammars together with a parity test The JSON number grammar is encoded twice: as the scan_number() state machine and as the contiguous fast path. The fast path declining on anything it does not recognize keeps most divergence harmless, but if it ever accepted something the state machine rejects the result would be a silent correctness bug, and the existing test only pinned a hand-written list of numbers. Enumerate every string of length 1..4 over "01.eE+-" (2800 tokens) and require both paths to agree on the parsed value and on the exact error message. Verified to fail if the fast path's grammar is perturbed. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Skip token_buffer for integers on the contiguous path An integer token does not need token_buffer: the number_integer and number_unsigned SAX callbacks take only the value, and the overflow diagnostic rebuilds the text from the input via get_token_string(). Convert straight from the input buffer and materialize the token only for the floating-point tail, which still needs a NUL-terminated buffer for strtod. JSON_DIAGNOSTIC_POSITIONS derives a number's start position from get_string().size(), so the copy is kept when that is enabled. The integer dispatch is factored into convert_integer() and shared with convert_number(), so both scanners keep using one implementation. Integer-heavy input, 400k values, -O3: parse accept gcc 16 +14% +23% clang +15% +22% Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Cover the bulk string and UTF-8 scanners These paths had no dedicated tests and rested on differential fuzzing only. Add three sections, all comparing the contiguous scanner against the byte-at-a-time one on the parsed value and on the exact error message: - every string of length 1..3 over an alphabet of ordinary ASCII, both specials, a control byte, escape characters, UTF-8 lead and continuation bytes, and a byte that is never valid - each at offset 0 and offset 9, so the bulk scanner sees them with and without a run behind them - every kind of run-ending byte at each offset across two 8-byte SWAR words, so multibyte sequences also straddle the word boundary - the boundaries of every range validate_one_utf8() recognizes: shortest and longest encodings, overlongs, both ends of the surrogate block, U+10FFFF and just past it, and truncated sequences Verified to fail if the bulk validator accepts surrogates, and if the SWAR word test stops detecting control characters. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Format the new string fast path test with astyle The pinned astyle expands a braced-init-list used as a range-for range onto several lines; hoist the two offsets into a named vector instead, which reads better and leaves nothing for astyle to reformat. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fix shadowed locals and guard the exception-dependent tests Two problems in the tests added for the bulk scanners, both found by CI: - the inner `const json j` in the counted-iterator diagnostics shadowed the one declared at test-case scope, which -Wshadow rejects on GCC and clang and C4456 rejects on MSVC under /WX; rename them - the new parity checks parse deliberately invalid input, which calls std::abort() rather than throwing when JSON_NOEXCEPTION is defined, so they would have crashed the no-exception build; guard them the way the other tests do json::accept() does not abort, so the UTF-8 range assertions stay compiled without exceptions and keep covering validate_one_utf8() there. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Gate the C++20 iterator classification on JSON_HAS_RANGES Making supports_bulk_scan depend on iterator_is_contiguous meant the trait is now instantiated for every adapter, not only when get_elements() is called. On standard libraries with an incomplete <ranges> that is fatal: libstdc++ 10 evaluates std::contiguous_iterator<std::counted_iterator<T*>> by calling std::to_address, which needs an operator-> its counted_iterator does not have, so satisfaction checking is a hard error rather than false. Reported by clang 14 + libstdc++ 10. JSON_HAS_RANGES already encodes exactly this ("libstdc++ < 11 has incomplete C++20 ranges", #4440), so require it for the C++20 branch. Affected toolchains fall back to the pointer-only test and the byte-at-a-time scanner, which parses identically, just without the bulk fast paths. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Amalgamate source code Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Satisfy clang-tidy in the new bulk scanner tests - give the helper lambdas an explicit std::string return type and return braced initializer lists (modernize-return-braced-init-list) - replace the C-style array of test cases with a std::vector (modernize-avoid-c-arrays) - silence pro-type-member-init on the two brace-initialized aggregates; default member initializers would stop them being aggregates in C++11 Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Avoid escaped literals in the counted-iterator diagnostics list clang-tidy reads "[\"\\ud834\"]" as a literal better written raw, and the two literals written next to each other in "[\"a\x01""b\"]" as a missing comma. The concatenation was there to stop the hex escape swallowing the following character; build those documents from explicit bytes instead and use raw strings elsewhere. The byte sequences are unchanged. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Reattach convert_number's documentation @gregmarr spotted that convert_integer() was inserted between convert_number() and its doc block, leaving convert_integer() with two stacked blocks and convert_number() with none. Comment only; no code change. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Address review comments on the simdutf backend string_bulk_run() had the same `return scalar_string_bulk_run(...)` in both arms of the `#if`. The simdutf arm already falls through when validation fails, so a single return after the `#endif` says the same thing. The JSON_USE_SIMDUTF example showed `#include <simdutf.h>`, which string_scan.hpp already does under the same guard; users only have to put the header on the include path and link the library, not include it themselves. Set the version history entry to 3.13.0, matching the other macro pages documenting unreleased features. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Pin the number error position against a non-newline terminator The comment claimed the reported column is the one the offending token starts at. It is the column reached after the token's last character - which is the actual point of the unget() change: a number terminated by a newline now reports what the same number terminated by a space always did. Assert that equality directly, and add a multi-character token where the start and end columns differ, so the invariant cannot be read off a single-character example. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not repeat the integer conversion that just failed scan_number_bulk_contiguous() converts an integer token straight from the input buffer. When the value does not fit, it materializes token_buffer and calls convert_number(), which tried the very same integer conversion again before falling back to floating point. Recording the outcome in number_type skips the second attempt. The resulting token type and value are unchanged: convert_number() reached the float tail either way. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Require a container's value_type to match what data() points at is_contiguous_byte_container accepted any type with a data() returning a pointer to a single-byte integral plus a size(). That is duck typing: the two members say nothing about size() counting the units data() points at. A type where it does not - fixed-size records, say - was routed to the pointer-based adapter and parsed as [data(), data() + size()) bytes, silently truncating input the iterator-based adapter had read in full: struct record_buffer { using value_type = std::array<char, 4>; std::string bytes; const char* data() const; // raw bytes std::size_t size() const; // in records const char* begin() const; const char* end() const; }; json::parse(record_buffer{"[1,2,3,4,5]"}); // parse error at column 3 Requiring the container's own value_type to be that same element type ties the two together. Every contiguous standard container satisfies it, so std::string, std::vector<char>, std::array<char, N> and std::string_view keep the fast path; anything else falls back to the iterator-based adapter, which is always correct. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Only compile the simdutf backend from C++17 on simdutf.h rejects anything below C++17 with an #error, so defining JSON_USE_SIMDUTF in a C++11 or C++14 translation unit did not fail with a message about simdutf being unavailable - it failed to compile at all, taking the library's C++11 support with it. Nothing caught this because no build ever compiled that path. Gate the include and both uses on JSON_HAS_CPP_17, the same way number_parse.hpp gates std::from_chars. Below C++17 the macro now has no effect and the scalar validator runs; it accepts and rejects exactly the same input, so the macro is safe to set project-wide even when some translation units use an older standard. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Test the simdutf backend in CI JSON_USE_SIMDUTF was documented and shipped but never compiled by anything in the repository, so nothing held the backend to the behavior the docs promise. Add JSON_TestSimdutf (OFF by default), which fetches simdutf and defines JSON_USE_SIMDUTF for every test target, and a ci_test_simdutf target that runs the whole suite in that configuration. Because simdutf needs C++17, the suite is built at C++11 as well, so one job covers both the scalar fallback with the macro defined and simdutf itself. The dependency hangs off test_main, whose usage requirements every test target inherits. The library target and the installed CMake package are deliberately untouched: making nlohmann_json link simdutf would put a find_dependency() in the exported package, which is a separate decision. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Warn when JSON_TestSimdutf cannot reach the simdutf backend simdutf needs C++17: without it the dependency does not even compile, and with a C++17 compiler but no C++17-or-later standard under test it builds and then goes unused. Either way the option silently did nothing useful, or broke the configure step outright. Resolve the tested standards first, then check them: when none of them can reach simdutf, skip the dependency and say so, naming which of the two reasons applies and how to fix it. The tests then run against the scalar validator, which is what would have happened anyway. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01MXDi7NTMKAmoArUZSKMc4T Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Use record_buffer::data() so clang does not flag it unneeded The record_buffer test type declares data() and size() so the is_contiguous_byte_container trait can see both and still reject the type on its value_type. data() was never called, so clang's -Wunneeded-member-function (under -Weverything -Werror) failed the C++20 build. Assert that data() points at the underlying bytes: it ODR-uses the member and documents the property the type is meant to demonstrate. Co-Authored-By: Claude <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Skip the float fast path when it cannot succeed parse_float_fast() (Clinger) needs a significand below 2^53, so it always declines once the mantissa has 17 or more significant digits. convert_number() called it unconditionally, so those numbers were walked an extra time before strtod had to run anyway. On streaming input, where scanning is byte-at-a-time and there is no compensating win, that made canada.json about 6% slower than develop. Derive the significant-digit count from token_buffer indices - the digits are not scanned again - and skip the call when it is guaranteed to decline. Both scanners pass the offset where the mantissa ends; the count only has to be corrected for a leading "0", which the JSON grammar admits nowhere else. The integer path returns before the check, so integer-heavy input is unaffected. Values are unchanged: this only avoids an attempt that would have failed. Verified bit-exact against develop over every number in canada.json, floats.json, signed_ints.json, unsigned_ints.json, small_signed_ints.json, citm_catalog.json and twitter.json, for both the contiguous and the streaming scanner. parse, streaming develop before after canada.json 19.4ms 20.5ms 19.3ms floats.json 135.9ms 131.8ms 128.0ms parse, contiguous develop before after canada.json 15.5ms 12.9ms 11.7ms floats.json 98.6ms 69.8ms 66.7ms Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> | 6 天前 | |
Skip integer conversion in accept()/SAX validation when the value is unused (#5484) * Skip integer conversion in accept()/SAX validation when the value is unused lexer::scan_number() always converted every numeric token with strtoull()/strtoll() before returning, even though accept() (and any consumer using json_sax_acceptor) immediately discards the converted value. For value_unsigned/value_integer tokens whose digit count already guarantees the value fits into 64 bits, the conversion cannot change the accept/reject decision (such tokens are always finite and unconditionally accepted), so scan_number() can skip strtoull()/ strtoll() entirely in that case when the caller signals it does not need the value. Numbers with more digits keep using the exact, unmodified conversion path, so overflow reclassification to value_float (and the finiteness check on it) is unaffected. parse() and value_float handling are completely unchanged. Fixes #5411 Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Document why the digit-count fast path is safe regardless of number_unsigned_t/number_integer_t width Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Avoid temporary-string concatenation flagged by clang-tidy in the differential test Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> | 7 天前 | |
:page_facing_up: adjust year (#5044) Signed-off-by: Niels Lohmann <mail@nlohmann.me> | 8 个月前 | |
Speed up dump(), and keep it from overflowing the stack (#5285) * Add SWAR bulk fast path to string serialization (dump_escaped) When ensure_ascii is false, dump_escaped previously ran every byte of every string and object key through the UTF-8 DFA decoder, even for the common case of ordinary text with nothing to escape. This mirrors the per-byte cost the parser had before the contiguous fast paths. At a character boundary, bulk-copy the longest run of bytes that need no escaping using string_bulk_run() - the same SWAR scanner and UTF-8 bulk validator the lexer's contiguous path uses - and only fall back to the byte-at-a-time DFA loop for the first byte that needs individual handling (a quote, backslash, control character, or ill-formed/truncated UTF-8). Because every "hard" or invalid byte is still processed by the unchanged byte path, escaping output and error handling (including strict-mode error 316 position and message) are byte-identical to before. The ensure_ascii=true path is unchanged: it must escape non-ASCII and 0x7F, which string_bulk_run does not stop on, so a separate predicate would be needed for it. Verified byte-for-byte identical dump output against the pre-change implementation across ~20k randomized byte strings plus curated edge cases (all escapes, control chars, valid multibyte, surrogates, overlong, truncated sequences) for both ensure_ascii settings and all three error handlers, in C++11/17/20 at -O2/-O3. Throughput (g++ -O3, ensure_ascii=false, vs pre-change): long ASCII strings 4.2x twitter-like objects 2.3x dense CJK 1.4x (further headroom with JSON_USE_SIMDUTF) Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Buffer serializer output and add ensure_ascii string fast path Two further serialization speedups on top of the ensure_ascii=false bulk copy, both reusing the SWAR primitives in detail/input/string_scan.hpp. 1. Internal write buffer (devirtualization). Every structural character ('{', '"', ',', ...) previously went straight to the output adapter through a virtual call. Route all writes through put_char/put_chars into a 1 KiB buffer that flushes in bulk; the public dump() flushes once the top-level value is done (the recursive worker is split out as dump_internal). Runs larger than the buffer are written straight through, so large payloads are not copied twice. This is the dominant cost for object/array-heavy values. 2. ensure_ascii fast path. dump_escaped previously ran the UTF-8 DFA over every byte when escaping non-ASCII. Add find_ascii_copyable_run() (a SWAR scan stopping at '"', '\\', < 0x20, 0x7F, and >= 0x80) so runs of printable ASCII are bulk-copied, with the byte path handling each escape/non-ASCII byte exactly as before. Behavior is unchanged: dump output is byte-for-byte identical to the previous implementation across ~20k randomized byte strings plus curated edge cases (all escapes, control chars, 0x7F, valid multibyte, surrogates, overlong, truncated), for object/array/pretty output, both ensure_ascii settings, and all three error handlers, in C++11/17/20 at -O2/-O3. New unit tests cover the buffer flush boundaries, the escape and 0x7F handling, multibyte under both settings, and invalid-UTF-8 handling. Throughput (g++ -O3, vs the ensure_ascii=false-only baseline): long ASCII, ensure_ascii=0 4.2x long ASCII, ensure_ascii=1 4.1x twitter-like objects 2.7x dense CJK 1.8x Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Flush serializer buffer in dump_escaped unit test test-convenience failed (macOS finished first; the failure is platform-independent) because check_escaped() calls the internal serializer::dump_escaped() directly and then reads the output stream. Since dump_escaped() now writes into the serializer's internal write buffer, the bytes were still buffered and the stream was empty. Expose flush() under JSON_PRIVATE_UNLESS_TESTED (same visibility as dump_escaped) and flush in check_escaped() before inspecting the output. Per-string flushing inside dump_escaped() was rejected on purpose: it would defeat the buffering that makes object/array-heavy dumps faster. Library behavior is unchanged (flush()'s body is identical; only its access label moved). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Avoid deep recursion in serialization write-buffer test The "many small structural writes exceed the write buffer" subcase built a 1100-deep nested array and dumped it to force >1024 consecutive single-character writes through put_char (exercising the write buffer's flush-when-full branch). dump() recurses per nesting level, so on MSVC debug builds (smaller default stack, larger frames) this overflowed the stack and crashed test-serialization; Linux/macOS have enough headroom to hide it. Replace the nesting with a flat array of 500 empty strings. Each element emits '"', '"', ',' via put_char, so the dump is a long run of single-character writes (1501 bytes > the 1024-byte buffer) at nesting depth two, hitting the same flush branch without deep recursion. Library code is unchanged. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAYM1qhSA2FDaDcGfPW3fG Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Split the write-buffer helpers and write indentation directly Follow-up to @gregmarr's review: put_chars() was doing four unrelated jobs, so give the two that can be made safe their own entry points. - put_literal(): takes the literal by reference and deduces the length from the array bound, so the 27 hand-counted lengths at the call sites can no longer drift from the literals they describe. A literal is checked at compile time to fit the buffer, so this path needs no write-through branch. - put_buffer(): takes the fixed-size buffer itself rather than a bare pointer, so the length can be checked against the buffer's own bound. - put_indent(): memsets the indentation into the write buffer, filling and flushing it as needed. This removes indent_string entirely, and with it both bugs of #5186: the indentation string was grown by doubling, which is not enough when indent_step more than doubles it (a heap over-read - dump(2000) read 2000 bytes out of a 1024-byte string), and the grown part was filled with a space instead of the configured indent_char. next_indent() keeps that PR's assertion against the unsigned indentation accumulation wrapping on deep nesting. put_chars() keeps the two cases that are genuinely a pointer and a count: the run-length copies out of the string being escaped, and to_chars() output. Tests cover an indent_step wider than the write buffer, a non-space indentation character past the old growth point, and nesting whose accumulated indentation spans several buffer-fulls. All three fail against develop. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fill the indentation buffer once instead of once per flush @gregmarr's point on the fill-and-flush loop: flushing does not disturb what the write buffer holds, so an indentation spanning several buffer-fulls only has to be written into the buffer once and can then be handed to the adapter as many times as needed. The loop re-filled it every time, doing work it already knew was there. put_indent() now fills the room left in the buffer, and if anything remains, flushes, fills the buffer once, and re-flushes that same content. It also returns early for a zero-width indentation, which is what the closing brace of every outermost value asks for. Measured over a dump(), counting memset calls and bytes inside put_indent: indent before after 4 1 call / 4 B 1 call / 4 B 2000 2 calls / 2000 B 2 calls / 2046 B 100000 98 calls / 100000 B 2 calls / 2046 B The wide case is now constant work rather than proportional to the indentation width; ordinary widths are unchanged. Tests extended to cover several whole buffer-fulls and an exact multiple of the buffer size. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Tighten the write-buffer helpers after review More of @gregmarr's review on the put_* split: - Reattach the put_chars() doc comment, which the new helpers had been inserted in front of, leaving it describing put_indent(). - Compute the literal length once in put_literal() instead of spelling N - 1 at each use. - Add put_string(str, start, end), which keeps the pointer arithmetic and the bounds assertions inside the function instead of at the call site. With dump_float()'s to_chars() output moved onto put_buffer() as well, put_chars() now has no callers outside put_string()/put_buffer(): nothing passes a bare pointer and a count any more. - Carry the indentation as std::size_t rather than unsigned int. It is a size, it is compared and combined with buffer sizes throughout, and the casts in put_indent() disappear. next_indent() keeps its assertion, which is far harder to trip on a 64-bit size_t but still reachable where that is 32 bits. No output change: pretty and compact dumps, binary values included, are byte-identical to develop. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Silence avoid-c-arrays on put_literal's array reference clang-tidy flags the reference-to-array parameter under cppcoreguidelines/hicpp/modernize-avoid-c-arrays, and the CI treats warnings as errors. Binding to the array is the whole point here - it is what lets the length be deduced from the literal instead of hand-written at the call site - so suppress it the same way from_json(), to_json() and get_to() already suppress it for their own T (&arr)[N] parameters. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Bound the descent of dump() Serializing a container serializes its elements, so dump() descended into one call per nesting level. A value nested deeply enough exhausted the call stack and terminated the process with a segmentation fault - no exception, nothing the caller could catch. Parsing such a value works, as the parser is iterative, and so does destroying one, as #1436 made destruction iterative. Bound how far the descent goes rather than take the call stack away from it. The first 128 levels are written by exactly the code that always wrote them, and only below that does dump_iteratively write out what is left, keeping the containers it has entered on an explicit stack. Serializing can therefore no longer exhaust the stack, however deeply a value is nested, while a value nested less deeply than the bound pays only for one comparison per container. Writing every value that way instead measured between 2% and 20% slower - 20% on object-heavy documents - which is why the descent is kept for all but the values that cannot afford it. The bound costs nothing measurable: between -1.4% and +1.2% across compact and pretty output of number, integer, string, object-heavy, wide-object and deeply nested documents. The output is unchanged for every value. Both ways of writing a container emit the separator in front of every element but the first, rather than after every element but the last, which puts exactly one between each pair and none at the end. This fixes #5387 for dump(). The copy constructor is fixed in #5389. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Fold ensure_ascii into the escaper and write bytes without dump_integer Two hot spots that the write buffer and the bulk scanner left behind. dump_escaped took ensure_ascii as a runtime flag and tested it inside the loop, once per character run, although it cannot change while a string is written. It is now a template parameter, dispatched once per string, which folds the choice of scanner and lets each of the two be inlined into a loop of its own. This is the hottest loop in the serializer: it runs over every string and every object key. A binary value's bytes went through dump_integer, which counts digits and does 64-bit arithmetic for a number that is always in [0, 255]. dump_byte writes the three digits it takes at most straight into the write buffer instead. Any byte type that is not a plain unsigned byte is still left to dump_integer, whose representation of it may differ. Measured against the previous commit (medians of 9 interleaved runs, clang -O3): binary values -33.8%, dense CJK with ensure_ascii -20.6%, key-heavy objects -17.8%, deeply nested pretty output -17.9%, dense CJK without ensure_ascii -11.8%, object-heavy documents -9.3% compact and -9.5% pretty, a small value dumped in a loop -21.4%, wide objects -2.3%. Arrays of plain ASCII strings measured 3.5% to 4.2% slower, the one shape that loses; number and integer arrays are unchanged. Also tried and dropped: leaving the write and string buffers uninitialized rather than zeroing 1.5 KB per dump() call. It is worth -30% on small values, but two nearly identical string workloads moved 18% apart in opposite directions, so the measurements did not support it. The output is unchanged for every value: the differential now also covers every one of the 256 byte values, alone and together, in both binary layouts. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Write a byte without walking a pointer over the buffer clang-tidy's misc-const-correctness reads the pointer dump_byte advanced over the write buffer as one whose pointee could be const. Index the buffer instead, which says the same thing without a raw pointer at all. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Parenthesize the reserve arithmetic in the deep-nesting test clang-tidy's readability-math-missing-parentheses wants the multiplication spelled out in reserve(6 * depth + 1), and CI treats its warnings as errors. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Do not scan for a copyable run that cannot exist Under ensure_ascii, dump_escaped() calls find_ascii_copyable_run() at every character boundary. When the text is dense non-ASCII - CJK, where every byte is >= 0x80 - the scanner stops on its first byte and returns zero, so its SWAR block runs once per character and buys nothing, on top of the escaping that still has to happen afterwards. A run can only be non-empty when the first byte is one the scanner may copy, so test that single byte before calling it. Runs that do exist are found exactly as before, so the bulk-copy win is unchanged; only the calls that were always going to return zero are skipped. Output is unchanged: the dump digest over canada/citm/twitter, in compact, pretty and ensure_ascii form, matches develop byte for byte. dump(ensure_ascii=true) develop before after CJK text 3.54ms 4.25ms 3.36ms CJK, no ASCII at all 3.09ms 4.02ms 3.02ms Latin-1-ish text 4.39ms 3.04ms 2.93ms plain ASCII 3.92ms 0.80ms 0.79ms Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Address review of the write-buffer helpers Three points from @gregmarr's review: put_chars() is gone. It was the only entry point taking a bare pointer and a count, and it existed only so put_string() and put_buffer() had something to delegate to. Its body now lives in put_string(), and put_buffer() is put_string(buffer, 0, length) - std::array already carries data() and size(), so it satisfies the same interface a string does. Nothing appends characters without a bound any more. dump_escaped()'s documentation block was duplicated. The dispatcher was inserted between the original comment and the function it described, and the comment was copied rather than split. The worker now has its own short comment saying why ensure_ascii is a template parameter. The local in dump_byte() is deliberate, and is now documented as such: writing through write_buffer[] is a char write, which may alias any object, so with write_buffer_pos updated in place the compiler must reload and store it around every digit. Measured on a dump of a 4 MiB binary value, 18.0 ms without the local against 7.4 ms with it. Output is unchanged: byte-identical dumps across 77 files in compact, pretty, ensure_ascii, pretty+ascii, indent 600 and tab-indent form. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Address review: drop unneeded backslash-escapes and duplicate scan loop '"' does not need escaping in a char literal, unlike in a string literal. find_ascii_copyable_run() also duplicated the byte-at-a-time search that already exists as the loop's own scalar tail; break into it instead of re-deriving the offset in a second, near-identical loop. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Move pretty_print, ensure_ascii and indent_step into the serializer None of these change over the life of a serializer, unlike current_indent and depth, which do change on every recursive call. They are now captured once in the constructor - matching indent_char and error_handler - instead of being threaded through dump(), dump_internal(), dump_iteratively(), dump_value() and dump_escaped() on every call. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Stop the serializer from holding onto std::localeconv()'s pointer loc was only ever read twice, immediately, to seed thousands_sep and decimal_point; nothing else in the class used it. A local in the constructor body serves the same purpose without keeping the pointer around for the serializer's lifetime. Signed-off-by: Niels Lohmann <mail@nlohmann.me> * Keep thousands_sep/decimal_point const via a small locale_chars struct const members can't be assigned in a constructor body, so seeding them from std::localeconv() meant either dropping const or holding onto the lconv* for longer than needed. A sub-object computes both from the pointer in its own constructor and is itself initialized in serializer's mem-initializer-list, so the two chars stay const, std::localeconv() is still called exactly once, and nothing outlives the constructor. Signed-off-by: Niels Lohmann <mail@nlohmann.me> --------- Signed-off-by: Niels Lohmann <mail@nlohmann.me> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com> | 5 天前 |
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
| 1 天前 | ||
| 6 天前 | ||
| 2 天前 | ||
| 6 天前 | ||
| 6 天前 | ||
| 7 天前 | ||
| 8 个月前 | ||
| 5 天前 |