Optimize binary get_number implementation by reading multiple bytes at once (#4391)

* multibyte binary reader

* wide_string_input_adapter fallback to get_character

Update input_adapters.hpp

* Update json.hpp

* Add from msgpack test

* Test for broken msgpack with stream, address some warnings

* Reading binary number from wchar as an error, address warnings

* Not casting float to int, it violates strict aliasing rule

Author: TianyiChen

include/nlohmann/detail/input/binary_reader.hpp

@@ -20,6 +20,9 @@
 #include <string> // char_traits, string
 #include <utility> // make_pair, move
 #include <vector> // vector
+#ifdef __cpp_lib_byteswap
+    #include <bit>  //byteswap
+#endif
 
 #include <nlohmann/detail/exceptions.hpp>
 #include <nlohmann/detail/input/input_adapters.hpp>
@@ -2754,6 +2757,29 @@ class binary_reader
         return current = ia.get_character();
     }
 
+    /*!
+    @brief get_to read into a primitive type
+
+    This function provides the interface to the used input adapter. It does
+    not throw in case the input reached EOF, but returns false instead
+
+    @return bool, whether the read was successful
+    */
+    template<class T>
+    bool get_to(T& dest, const input_format_t format, const char* context)
+    {
+        auto new_chars_read = ia.get_elements(&dest);
+        chars_read += new_chars_read;
+        if (JSON_HEDLEY_UNLIKELY(new_chars_read < sizeof(T)))
+        {
+            // in case of failure, advance position by 1 to report failing location
+            ++chars_read;
+            sax->parse_error(chars_read, "<end of file>", parse_error::create(110, chars_read, exception_message(format, "unexpected end of input", context), nullptr));
+            return false;
+        }
+        return true;
+    }
+
     /*!
     @return character read from the input after ignoring all 'N' entries
     */
@@ -2768,6 +2794,28 @@ class binary_reader
         return current;
     }
 
+    template<class NumberType>
+    static void byte_swap(NumberType& number)
+    {
+        constexpr std::size_t sz = sizeof(number);
+#ifdef __cpp_lib_byteswap
+        if constexpr (sz == 1)
+        {
+            return;
+        }
+        if constexpr(std::is_integral_v<NumberType>)
+        {
+            number = std::byteswap(number);
+            return;
+        }
+#endif
+        auto* ptr = reinterpret_cast<std::uint8_t*>(&number);
+        for (std::size_t i = 0; i < sz / 2; ++i)
+        {
+            std::swap(ptr[i], ptr[sz - i - 1]);
+        }
+    }
+
     /*
     @brief read a number from the input
 
@@ -2786,29 +2834,16 @@ class binary_reader
     template<typename NumberType, bool InputIsLittleEndian = false>
     bool get_number(const input_format_t format, NumberType& result)
     {
-        // step 1: read input into array with system's byte order
-        std::array<std::uint8_t, sizeof(NumberType)> vec{};
-        for (std::size_t i = 0; i < sizeof(NumberType); ++i)
-        {
-            get();
-            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "number")))
-            {
-                return false;
-            }
+        // read in the original format
 
-            // reverse byte order prior to conversion if necessary
-            if (is_little_endian != (InputIsLittleEndian || format == input_format_t::bjdata))
-            {
-                vec[sizeof(NumberType) - i - 1] = static_cast<std::uint8_t>(current);
-            }
-            else
-            {
-                vec[i] = static_cast<std::uint8_t>(current); // LCOV_EXCL_LINE
-            }
+        if (JSON_HEDLEY_UNLIKELY(!get_to(result, format, "number")))
+        {
+            return false;
+        }
+        if (is_little_endian != (InputIsLittleEndian || format == input_format_t::bjdata))
+        {
+            byte_swap(result);
         }
-
-        // step 2: convert array into number of type T and return
-        std::memcpy(&result, vec.data(), sizeof(NumberType));
         return true;
     }
 

include/nlohmann/detail/input/input_adapters.hpp

@@ -68,6 +68,13 @@ class file_input_adapter
         return std::fgetc(m_file);
     }
 
+    // returns the number of characters successfully read
+    template<class T>
+    std::size_t get_elements(T* dest, std::size_t count = 1)
+    {
+        return fread(dest, 1, sizeof(T) * count, m_file);
+    }
+
   private:
     /// the file pointer to read from
     std::FILE* m_file;
@@ -127,6 +134,17 @@ class input_stream_adapter
         return res;
     }
 
+    template<class T>
+    std::size_t get_elements(T* dest, std::size_t count = 1)
+    {
+        auto res = static_cast<std::size_t>(sb->sgetn(reinterpret_cast<char*>(dest), static_cast<std::streamsize>(count * sizeof(T))));
+        if (JSON_HEDLEY_UNLIKELY(res < count * sizeof(T)))
+        {
+            is->clear(is->rdstate() | std::ios::eofbit);
+        }
+        return res;
+    }
+
   private:
     /// the associated input stream
     std::istream* is = nullptr;
@@ -158,6 +176,26 @@ class iterator_input_adapter
         return char_traits<char_type>::eof();
     }
 
+    // for general iterators, we cannot really do something better than falling back to processing the range one-by-one
+    template<class T>
+    std::size_t get_elements(T* dest, std::size_t count = 1)
+    {
+        auto* ptr = reinterpret_cast<char*>(dest);
+        for (std::size_t read_index = 0; read_index < count * sizeof(T); ++read_index)
+        {
+            if (JSON_HEDLEY_LIKELY(current != end))
+            {
+                ptr[read_index] = static_cast<char>(*current);
+                std::advance(current, 1);
+            }
+            else
+            {
+                return read_index;
+            }
+        }
+        return count * sizeof(T);
+    }
+
   private:
     IteratorType current;
     IteratorType end;
@@ -321,6 +359,13 @@ class wide_string_input_adapter
         return utf8_bytes[utf8_bytes_index++];
     }
 
+    // parsing binary with wchar doesn't make sense, but since the parsing mode can be runtime, we need something here
+    template<class T>
+    std::size_t get_elements(T* /*dest*/, std::size_t /*count*/ = 1)
+    {
+        JSON_THROW(parse_error::create(112, 1, "wide string type cannot be interpreted as binary data", nullptr));
+    }
+
   private:
     BaseInputAdapter base_adapter;
 

single_include/nlohmann/json.hpp

@@ -6220,6 +6220,9 @@ NLOHMANN_JSON_NAMESPACE_END
 #include <string> // char_traits, string
 #include <utility> // make_pair, move
 #include <vector> // vector
+#ifdef __cpp_lib_byteswap
+    #include <bit>  //byteswap
+#endif
 
 // #include <nlohmann/detail/exceptions.hpp>
 
@@ -6298,6 +6301,13 @@ class file_input_adapter
         return std::fgetc(m_file);
     }
 
+    // returns the number of characters successfully read
+    template<class T>
+    std::size_t get_elements(T* dest, std::size_t count = 1)
+    {
+        return fread(dest, 1, sizeof(T) * count, m_file);
+    }
+
   private:
     /// the file pointer to read from
     std::FILE* m_file;
@@ -6357,6 +6367,17 @@ class input_stream_adapter
         return res;
     }
 
+    template<class T>
+    std::size_t get_elements(T* dest, std::size_t count = 1)
+    {
+        auto res = static_cast<std::size_t>(sb->sgetn(reinterpret_cast<char*>(dest), static_cast<std::streamsize>(count * sizeof(T))));
+        if (JSON_HEDLEY_UNLIKELY(res < count * sizeof(T)))
+        {
+            is->clear(is->rdstate() | std::ios::eofbit);
+        }
+        return res;
+    }
+
   private:
     /// the associated input stream
     std::istream* is = nullptr;
@@ -6388,6 +6409,26 @@ class iterator_input_adapter
         return char_traits<char_type>::eof();
     }
 
+    // for general iterators, we cannot really do something better than falling back to processing the range one-by-one
+    template<class T>
+    std::size_t get_elements(T* dest, std::size_t count = 1)
+    {
+        auto* ptr = reinterpret_cast<char*>(dest);
+        for (std::size_t read_index = 0; read_index < count * sizeof(T); ++read_index)
+        {
+            if (JSON_HEDLEY_LIKELY(current != end))
+            {
+                ptr[read_index] = static_cast<char>(*current);
+                std::advance(current, 1);
+            }
+            else
+            {
+                return read_index;
+            }
+        }
+        return count * sizeof(T);
+    }
+
   private:
     IteratorType current;
     IteratorType end;
@@ -6551,6 +6592,13 @@ class wide_string_input_adapter
         return utf8_bytes[utf8_bytes_index++];
     }
 
+    // parsing binary with wchar doesn't make sense, but since the parsing mode can be runtime, we need something here
+    template<class T>
+    std::size_t get_elements(T* /*dest*/, std::size_t /*count*/ = 1)
+    {
+        JSON_THROW(parse_error::create(112, 1, "wide string type cannot be interpreted as binary data", nullptr));
+    }
+
   private:
     BaseInputAdapter base_adapter;
 
@@ -12007,6 +12055,29 @@ class binary_reader
         return current = ia.get_character();
     }
 
+    /*!
+    @brief get_to read into a primitive type
+
+    This function provides the interface to the used input adapter. It does
+    not throw in case the input reached EOF, but returns false instead
+
+    @return bool, whether the read was successful
+    */
+    template<class T>
+    bool get_to(T& dest, const input_format_t format, const char* context)
+    {
+        auto new_chars_read = ia.get_elements(&dest);
+        chars_read += new_chars_read;
+        if (JSON_HEDLEY_UNLIKELY(new_chars_read < sizeof(T)))
+        {
+            // in case of failure, advance position by 1 to report failing location
+            ++chars_read;
+            sax->parse_error(chars_read, "<end of file>", parse_error::create(110, chars_read, exception_message(format, "unexpected end of input", context), nullptr));
+            return false;
+        }
+        return true;
+    }
+
     /*!
     @return character read from the input after ignoring all 'N' entries
     */
@@ -12021,6 +12092,28 @@ class binary_reader
         return current;
     }
 
+    template<class NumberType>
+    static void byte_swap(NumberType& number)
+    {
+        constexpr std::size_t sz = sizeof(number);
+#ifdef __cpp_lib_byteswap
+        if constexpr (sz == 1)
+        {
+            return;
+        }
+        if constexpr(std::is_integral_v<NumberType>)
+        {
+            number = std::byteswap(number);
+            return;
+        }
+#endif
+        auto* ptr = reinterpret_cast<std::uint8_t*>(&number);
+        for (std::size_t i = 0; i < sz / 2; ++i)
+        {
+            std::swap(ptr[i], ptr[sz - i - 1]);
+        }
+    }
+
     /*
     @brief read a number from the input
 
@@ -12039,29 +12132,16 @@ class binary_reader
     template<typename NumberType, bool InputIsLittleEndian = false>
     bool get_number(const input_format_t format, NumberType& result)
     {
-        // step 1: read input into array with system's byte order
-        std::array<std::uint8_t, sizeof(NumberType)> vec{};
-        for (std::size_t i = 0; i < sizeof(NumberType); ++i)
-        {
-            get();
-            if (JSON_HEDLEY_UNLIKELY(!unexpect_eof(format, "number")))
-            {
-                return false;
-            }
+        // read in the original format
 
-            // reverse byte order prior to conversion if necessary
-            if (is_little_endian != (InputIsLittleEndian || format == input_format_t::bjdata))
-            {
-                vec[sizeof(NumberType) - i - 1] = static_cast<std::uint8_t>(current);
-            }
-            else
-            {
-                vec[i] = static_cast<std::uint8_t>(current); // LCOV_EXCL_LINE
-            }
+        if (JSON_HEDLEY_UNLIKELY(!get_to(result, format, "number")))
+        {
+            return false;
+        }
+        if (is_little_endian != (InputIsLittleEndian || format == input_format_t::bjdata))
+        {
+            byte_swap(result);
         }
-
-        // step 2: convert array into number of type T and return
-        std::memcpy(&result, vec.data(), sizeof(NumberType));
         return true;
     }