3.8.0

* In some cases, clang might try to cast an ondemand::document to an
ondemand::document, instead of calling the move constructor. So we
can disable the template cast.

* In some cases, clang might get confused when constructing an ondemand
document in a constructor: instead of calling the move constructor, it
somehow ends up trying to cast a document to a document. We can easily
disallow this behavior with std::enable_if.

* making compatible with C++11

---------

Co-authored-by: Daniel Lemire <dlemire@lemire.me>

.vscode/settings.json

@@ -103,6 +103,12 @@
         "charconv": "cpp",
         "source_location": "cpp",
         "strstream": "cpp",
-        "typeindex": "cpp"
+        "typeindex": "cpp",
+        "*.tcc": "cpp",
+        "memory_resource": "cpp",
+        "numbers": "cpp",
+        "semaphore": "cpp",
+        "stop_token": "cpp",
+        "cfenv": "cpp"
     }
 }

CMakeLists.txt

@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 3.14)
 project(
     simdjson
     # The version number is modified by tools/release.py
-    VERSION 3.7.0
+    VERSION 3.8.0
     DESCRIPTION "Parsing gigabytes of JSON per second"
     HOMEPAGE_URL "https://simdjson.org/"
     LANGUAGES CXX C
@@ -20,8 +20,8 @@ string(
 # ---- Options, variables ----
 
 # These version numbers are modified by tools/release.py
-set(SIMDJSON_LIB_VERSION "20.0.0" CACHE STRING "simdjson library version")
-set(SIMDJSON_LIB_SOVERSION "20" CACHE STRING "simdjson library soversion")
+set(SIMDJSON_LIB_VERSION "21.0.0" CACHE STRING "simdjson library version")
+set(SIMDJSON_LIB_SOVERSION "21" CACHE STRING "simdjson library soversion")
 
 option(SIMDJSON_BUILD_STATIC_LIB "Build simdjson_static library along with simdjson" OFF)
 

Doxyfile

@@ -38,7 +38,7 @@ PROJECT_NAME           = simdjson
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = "3.7.0"
+PROJECT_NUMBER         = "3.8.0"
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a

HACKING.md

@@ -283,8 +283,6 @@ If your compiler does not default on C++11 support or better you may get failing
 
 Note that the name of directory (`build`) is arbitrary, you can name it as you want (e.g., `buildgcc`) and you can have as many different such directories as you would like (one per configuration).
 
-
-
 ## Usage (CMake on 64-bit Windows using Visual Studio 2019 or better)
 
 Recent versions of Visual Studio support CMake natively, [please refer to the Visual Studio documentation](https://learn.microsoft.com/en-us/cpp/build/cmake-projects-in-visual-studio?view=msvc-170).
@@ -309,13 +307,11 @@ Though having Visual Studio installed is necessary, one can build simdjson using
 
 Furthermore, if you have installed LLVM clang on Windows, for example as a component of Visual Studio 2019, you can configure and build simdjson using LLVM clang on Windows using cmake:
 
-
 - `mkdir build`
 - `cd build`
 - `cmake -T ClangCL ..`
 - `cmake --build . --config Release`
 
-
 ## Various References
 
 - [How to implement atoi using SIMD?](https://stackoverflow.com/questions/35127060/how-to-implement-atoi-using-simd)

README.md

@@ -169,6 +169,7 @@ We distinguish between "bindings" (which just wrap the C++ code) and a port to a
 - [JSON-Simd](https://github.com/rawleyfowler/JSON-simd): Raku bindings.
 - [JSON::SIMD](https://metacpan.org/pod/JSON::SIMD): Perl bindings; fully-featured JSON module that uses simdjson for decoding.
 - [gemmaJSON](https://github.com/sainttttt/gemmaJSON): Nim JSON parser based on simdjson bindings.
+- [simdjson-java](https://github.com/simdjson/simdjson-java): Java port.
 
 About simdjson
 --------------

benchmark/apple/apple_arm_events.h

@@ -993,7 +993,6 @@ struct AppleEvents {
     // check permission
     int force_ctrs = 0;
     if (kpc_force_all_ctrs_get(&force_ctrs)) {
-      printf("Permission denied, xnu/kpc requires root privileges.\n");
       return (worked = false);
     }
     int ret;

dependencies/CMakeLists.txt

@@ -22,7 +22,7 @@ endif()
 # The bulk of our benchmarking and testing data has been moved simdjson/simdjson-data
 import_dependency(simdjson-data simdjson/simdjson-data a5b13babe65c1bba7186b41b43d4cbdc20a5c470)
 add_dependency(simdjson-data)
-
+option(SIMDJSON_USE_BOOST_JSON "Try to include BOOST_JSON, this may break your binaries under some systems." OFF)
 # This prevents variables declared with set() from unnecessarily escaping and
 # should not be called more than once
 function(competition_scope_)
@@ -36,8 +36,8 @@ function(competition_scope_)
 #endif
 
 int main() {}
-]] USE_BOOST_JSON)
-  if(USE_BOOST_JSON)
+]] SIMDJSON_FOUND_STRING_VIEW)
+  if(SIMDJSON_FOUND_STRING_VIEW AND SIMDJSON_USE_BOOST_JSON)
     import_dependency(boostjson boostorg/json ee8d72d)
     add_library(boostjson STATIC "${boostjson_SOURCE_DIR}/src/src.cpp")
     target_compile_definitions(boostjson PUBLIC BOOST_JSON_STANDALONE)
@@ -134,7 +134,7 @@ int main() {}
   add_library(competition-core INTERFACE)
   target_link_libraries(competition-core INTERFACE nlohmann_json rapidjson sajson cjson jsmn yyjson)
 
-  if(USE_BOOST_JSON)
+  if(TARGET boostjson)
      target_compile_definitions(boostjson INTERFACE HAS_BOOST_JSON)
      target_link_libraries(competition-core INTERFACE boostjson)
   endif()

doc/basics.md

@@ -1910,8 +1910,10 @@ An `ondemand::number` instance may contain an integer value or a floating-point
 Thus it is a dynamically typed number. Before accessing the value, you must determine the detected type:
 
 * `number.get_number_type()` has value `number_type::signed_integer` if we have a integer in [-9223372036854775808,9223372036854775808). You can recover the value by the `get_int64()` method applied on the `ondemand::number` instance. When `number.get_number_type()` has value `number_type::signed_integer`, you also have that `number.is_int64()` is true. Calling `get_int64()` on the `ondemand::number` instance when `number.get_number_type()` is not `number_type::signed_integer` is unsafe. You may replace `get_int64()` by a cast to a `int64_t` value.
-* `number.get_number_type()` has value `number_type::unsigned_integer` if we have a integer in [9223372036854775808,18446744073709551616). You can recover the value by the `get_uint64()` method applied on the `ondemand::number` instance.  When `number.get_number_type()` has value `number_type::unsigned_integer`, you also have that `number.is_uint64()` is true.  Calling `get_uint64()` on the `ondemand::number` instance when `number.get_number_type()` is not `number_type::unsigned_integer` is unsafe. You may replace `get_uint64()` by a cast to a `uint64_t` value.
+* `number.get_number_type()` has value `number_type::unsigned_integer` if we have a integer in `[9223372036854775808,18446744073709551616)`. You can recover the value by the `get_uint64()` method applied on the `ondemand::number` instance.  When `number.get_number_type()` has value `number_type::unsigned_integer`, you also have that `number.is_uint64()` is true.  Calling `get_uint64()` on the `ondemand::number` instance when `number.get_number_type()` is not `number_type::unsigned_integer` is unsafe. You may replace `get_uint64()` by a cast to a `uint64_t` value.
 * `number.get_number_type()` has value `number_type::floating_point_number` if we have and we have a floating-point (binary64) number. You can recover the value by the `get_double()` method applied on the `ondemand::number` instance.  When `number.get_number_type()` has value `number_type::floating_point_number`, you also have that `number.is_double()` is true.  Calling `get_double()` on the `ondemand::number` instance when `number.get_number_type()` is not `number_type::floating_point_number` is unsafe. You may replace `get_double()` by a cast to a `double` value.
+* When the value is an integer outside of the valid ranges for a 64-bit integers, e.g., when it is smaller than -9223372036854775808 or larger than 18446744073709551615, then `number.get_number_type()` has value `number_type::big_integer`. If you try to parse
+such a number of `get_number()`, you get the error `BIGINT_ERROR`. You can access the underlying string of digits with the function `raw_json_token()` which returns an `std::string_view` instance starting at the beginning of the digit. You can also call `get_double()` to get a floating-point approximation.
 
 
 You must check the type before accessing the value: it is an error to call `get_int64()` when `number.get_number_type()` is not `number_type::signed_integer` and when `number.is_int64()` is false. You are responsible for this check as the user of the library.
@@ -1944,6 +1946,9 @@ Consider the following example:
           std::cout  << "float: " << double(num) << " ";
           std::cout << "float: " << num.get_double() << std::endl;
           break;
+        case ondemand::number_type::big_integer:
+          std::cout  << "big-integer: " << val.raw_json_token() << std::endl;
+          break;
       }
     }
 ```
@@ -1959,6 +1964,55 @@ It will output:
 9999999999999999999 negative: 0 is_integer: 1 large 64-bit integer: 9999999999999999999 large 64-bit integer: 9999999999999999999
 ```
 
+In the following example, we have an array of integers that are outside the valid range of 64-bit signed or
+unsigned integers. Calling `get_number_type()` on the values returns `ondemand::number_type::big_integer`.
+You can try to represent these big integers as 64-bit floating-point numbers, though you typically lose
+precision in the process (as illustrated in the example).
+
+```C++
+  ondemand::parser parser;
+  padded_string docdata = R"([-9223372036854775809, 18446744073709551617, 99999999999999999999999 ])"_padded;
+  double dexpected[] = {-9223372036854775808.0, 18446744073709551616.0, 1e23};
+  ondemand::document doc = parser.iterate(docdata);
+  ondemand::array arr = doc.get_array();
+  for(ondemand::value val : arr) {
+    if(val.get_number_type() == ondemand::number_type::big_integer) {
+      std::cout << val.get_double() << std::endl;
+      // might print -9.22337e+18, 1.84467e+19, 1e+23
+    }
+  }
+```
+This program might print:
+```
+-9.22337e+18
+1.84467e+19
+1e+23
+```
+
+You may get access to the underlying string representing the big integer with
+`raw_json_token()` and you may parse the resulting number strings using your own parser.
+
+```c++
+  ondemand::parser parser;
+  padded_string docdata = R"([-9223372036854775809, 18446744073709551617, 99999999999999999999999 ])"_padded;
+  ondemand::document doc = parser.iterate(docdata);
+  ondemand::array arr = doc.get_array();
+  for(ondemand::value val : arr) {
+    // val.get_number_type() == ondemand::number_type::big_integer
+    if(val.get_number_type() == ondemand::number_type::big_integer) {
+      std::string_view token = val.raw_json_token();
+      // token = "-9223372036854775809", "18446744073709551617", "99999999999999999999999 "
+      std::cout << "'" << token << "'" << std::endl;
+    }
+  }
+```
+This code prints the following:
+```
+'-9223372036854775809'
+'18446744073709551617'
+'99999999999999999999999 '
+```
+
 Raw Strings From Keys
 -----------
 

include/simdjson/error.h

@@ -27,6 +27,7 @@ enum error_code {
   F_ATOM_ERROR,               ///< Problem while parsing an atom starting with the letter 'f'
   N_ATOM_ERROR,               ///< Problem while parsing an atom starting with the letter 'n'
   NUMBER_ERROR,               ///< Problem while parsing a number
+  BIGINT_ERROR,               ///< The integer value exceeds 64 bits
   UTF8_ERROR,                 ///< the input is not valid UTF-8
   UNINITIALIZED,              ///< unknown error, or uninitialized document
   EMPTY,                      ///< no structural element found

include/simdjson/generic/base.h

@@ -37,7 +37,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace SIMDJSON_IMPLEMENTATION

include/simdjson/generic/numberparsing.h

@@ -20,11 +20,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -355,6 +357,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -434,6 +440,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -508,6 +531,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -586,11 +621,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -1060,19 +1095,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -1250,6 +1298,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -1258,4 +1307,4 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
 } // namespace SIMDJSON_IMPLEMENTATION
 } // namespace simdjson
 
-#endif // SIMDJSON_GENERIC_NUMBERPARSING_H
+#endif // SIMDJSON_GENERIC_NUMBERPARSING_H

include/simdjson/generic/ondemand/document-inl.h

@@ -295,7 +295,7 @@ simdjson_inline simdjson_result<number> document::get_number() noexcept {
 
 simdjson_inline simdjson_result<std::string_view> document::raw_json_token() noexcept {
   auto _iter = get_root_value_iterator();
-  return std::string_view(reinterpret_cast<const char*>(_iter.peek_start()), _iter.peek_start_length());
+  return std::string_view(reinterpret_cast<const char*>(_iter.peek_start()), _iter.peek_root_length());
 }
 
 simdjson_inline simdjson_result<value> document::at_pointer(std::string_view json_pointer) noexcept {
@@ -535,6 +535,11 @@ simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::number> simdj
 
 
 #if SIMDJSON_EXCEPTIONS
+template <class T, typename std::enable_if<std::is_same<T, SIMDJSON_IMPLEMENTATION::ondemand::document>::value == false>::type>
+simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::document>::operator T() noexcept(false) {
+  if (error()) { throw simdjson_error(error()); }
+  return first;
+}
 simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::document>::operator SIMDJSON_IMPLEMENTATION::ondemand::array() & noexcept(false) {
   if (error()) { throw simdjson_error(error()); }
   return first;
@@ -842,6 +847,11 @@ simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::number> simdj
   return first.get_number();
 }
 #if SIMDJSON_EXCEPTIONS
+template <class T, typename std::enable_if<std::is_same<T, SIMDJSON_IMPLEMENTATION::ondemand::document_reference>::value == false>::type>
+simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::document_reference>::operator T() noexcept(false) {
+  if (error()) { throw simdjson_error(error()); }
+  return first;
+}
 simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::document_reference>::operator SIMDJSON_IMPLEMENTATION::ondemand::array() & noexcept(false) {
   if (error()) { throw simdjson_error(error()); }
   return first;

include/simdjson/generic/ondemand/document.h

@@ -179,21 +179,23 @@ public:
    * @returns INCORRECT_TYPE If the JSON value is not the given type.
    */
   template<typename T> simdjson_inline simdjson_result<T> get() & noexcept {
-    // Unless the simdjson library provides an inline implementation, calling this method should
+    // Unless the simdjson library or the user provides an inline implementation, calling this method should
     // immediately fail.
     static_assert(!sizeof(T), "The get method with given type is not implemented by the simdjson library. "
       "The supported types are ondemand::object, ondemand::array, raw_json_string, std::string_view, uint64_t, "
       "int64_t, double, and bool. We recommend you use get_double(), get_bool(), get_uint64(), get_int64(), "
-      " get_object(), get_array(), get_raw_json_string(), or get_string() instead of the get template.");
+      " get_object(), get_array(), get_raw_json_string(), or get_string() instead of the get template."
+      " You may also add support for custom types, see our documentation.");
   }
   /** @overload template<typename T> simdjson_result<T> get() & noexcept */
   template<typename T> simdjson_inline simdjson_result<T> get() && noexcept {
-    // Unless the simdjson library provides an inline implementation, calling this method should
+    // Unless the simdjson library or the user provides an inline implementation, calling this method should
     // immediately fail.
     static_assert(!sizeof(T), "The get method with given type is not implemented by the simdjson library. "
       "The supported types are ondemand::object, ondemand::array, raw_json_string, std::string_view, uint64_t, "
       "int64_t, double, and bool. We recommend you use get_double(), get_bool(), get_uint64(), get_int64(), "
-      " get_object(), get_array(), get_raw_json_string(), or get_string() instead of the get template.");
+      " get_object(), get_array(), get_raw_json_string(), or get_string() instead of the get template."
+      " You may also add support for custom types, see our documentation.");
   }
 
   /**
@@ -212,6 +214,15 @@ public:
   template<typename T> simdjson_inline error_code get(T &out) && noexcept;
 
 #if SIMDJSON_EXCEPTIONS
+  /**
+   * Cast this JSON value to an instance of type T. The programmer is responsible for
+   * providing an implementation of get<T> for the type T, if T is not one of the types
+   * supported by the library (object, array, raw_json_string, string_view, uint64_t, etc.)
+   *
+   * See https://github.com/simdjson/simdjson/blob/master/doc/basics.md#adding-support-for-custom-types
+   *
+   * @returns An instance of type T
+   */
   template <class T>
   explicit simdjson_inline operator T() noexcept(false);
   /**
@@ -386,7 +397,7 @@ public:
    * APIs assume this. Therefore, you must be explicit if you want to treat objects as out of order.
    *
    * Use find_field() if you are sure fields will be in order (or are willing to treat it as if the
-   * field wasn't there when they aren't).
+   * field was not there when they are not in order).
    *
    * You must consume the fields on an object one at a time. A request for a new key
    * invalidates previous field values: it makes them unsafe. E.g., the array
@@ -467,9 +478,11 @@ public:
    * get_number().get_number_type().
    *
    * get_number_type() is number_type::unsigned_integer if we have
-   * an integer greater or equal to 9223372036854775808
+   * an integer greater or equal to 9223372036854775808 and no larger than 18446744073709551615.
    * get_number_type() is number_type::signed_integer if we have an
-   * integer that is less than 9223372036854775808
+   * integer that is less than 9223372036854775808 and greater or equal to -9223372036854775808.
+   * get_number_type() is number_type::big_integer if we have an integer outside
+   * of those ranges (either larger than 18446744073709551615 or smaller than -9223372036854775808).
    * Otherwise, get_number_type() has value number_type::floating_point_number
    *
    * This function requires processing the number string, but it is expected
@@ -782,7 +795,7 @@ public:
   template<typename T> simdjson_inline error_code get(T &out) & noexcept;
   template<typename T> simdjson_inline error_code get(T &out) && noexcept;
 #if SIMDJSON_EXCEPTIONS
-  template <class T>
+  template <class T, typename std::enable_if<std::is_same<T, SIMDJSON_IMPLEMENTATION::ondemand::document>::value == false>::type>
   explicit simdjson_inline operator T() noexcept(false);
   simdjson_inline operator SIMDJSON_IMPLEMENTATION::ondemand::array() & noexcept(false);
   simdjson_inline operator SIMDJSON_IMPLEMENTATION::ondemand::object() & noexcept(false);
@@ -853,7 +866,7 @@ public:
   simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::ondemand::value> get_value() noexcept;
   simdjson_inline simdjson_result<bool> is_null() noexcept;
 #if SIMDJSON_EXCEPTIONS
-  template <class T>
+  template <class T, typename std::enable_if<std::is_same<T, SIMDJSON_IMPLEMENTATION::ondemand::document_reference>::value == false>::type>
   explicit simdjson_inline operator T() noexcept(false);
   simdjson_inline operator SIMDJSON_IMPLEMENTATION::ondemand::array() & noexcept(false);
   simdjson_inline operator SIMDJSON_IMPLEMENTATION::ondemand::object() & noexcept(false);

include/simdjson/generic/ondemand/json_iterator-inl.h

@@ -286,6 +286,12 @@ simdjson_inline uint32_t json_iterator::peek_length(token_position position) con
 #endif // SIMDJSON_CHECK_EOF
   return token.peek_length(position);
 }
+simdjson_inline uint32_t json_iterator::peek_root_length(token_position position) const noexcept {
+#if SIMDJSON_CHECK_EOF
+  assert_valid_position(position);
+#endif // SIMDJSON_CHECK_EOF
+  return token.peek_root_length(position);
+}
 
 simdjson_inline token_position json_iterator::last_position() const noexcept {
   // The following line fails under some compilers...

include/simdjson/generic/ondemand/json_iterator.h

@@ -185,6 +185,14 @@ public:
    * @param position The position of the token to retrieve.
    */
   simdjson_inline uint32_t peek_length(token_position position) const noexcept;
+  /**
+   * Get the maximum length of the JSON text for the current root token.
+   *
+   * The length will include any whitespace at the end of the token.
+   *
+   * @param position The position of the token to retrieve.
+   */
+  simdjson_inline uint32_t peek_root_length(token_position position) const noexcept;
   /**
    * Get the JSON text for the last token in the document.
    *

include/simdjson/generic/ondemand/json_type-inl.h

@@ -48,7 +48,6 @@ simdjson_inline number::operator uint64_t() const noexcept {
   return get_uint64();
 }
 
-
 simdjson_inline bool number::is_int64() const noexcept {
   return get_number_type() == number_type::signed_integer;
 }

include/simdjson/generic/ondemand/object.h

@@ -77,7 +77,7 @@ public:
    * APIs assume this. Therefore, you must be explicit if you want to treat objects as out of order.
    *
    * Use find_field() if you are sure fields will be in order (or are willing to treat it as if the
-   * field wasn't there when they aren't).
+   * field was not there when they are not in order).
    *
    * If you have multiple fields with a matching key ({"x": 1,  "x": 1}) be mindful
    * that only one field is returned.

include/simdjson/generic/ondemand/token_iterator-inl.h

@@ -37,6 +37,11 @@ simdjson_inline uint32_t token_iterator::peek_length(token_position position) co
   return *(position+1) - *position;
 }
 
+simdjson_inline uint32_t token_iterator::peek_root_length(token_position position) const noexcept {
+  return *(position+2) - *(position) > *(position+1) - *(position) ?
+      *(position+1) - *(position)
+      : *(position+2) - *(position);
+}
 simdjson_inline const uint8_t *token_iterator::peek(int32_t delta) const noexcept {
   return &buf[*(_position+delta)];
 }

include/simdjson/generic/ondemand/token_iterator.h

@@ -77,7 +77,14 @@ public:
    * @param position The position of the token.
    */
   simdjson_inline uint32_t peek_length(token_position position) const noexcept;
-
+  /**
+   * Get the maximum length of the JSON text for a root token.
+   *
+   * The length will include any whitespace at the end of the token.
+   *
+   * @param position The position of the token (start of the document).
+   */
+  simdjson_inline uint32_t peek_root_length(token_position position) const noexcept;
   /**
    * Return the current index.
    */

include/simdjson/generic/ondemand/value.h

@@ -36,12 +36,13 @@ public:
    * @returns INCORRECT_TYPE If the JSON value is not the given type.
    */
   template<typename T> simdjson_inline simdjson_result<T> get() noexcept {
-    // Unless the simdjson library provides an inline implementation, calling this method should
+    // Unless the simdjson library or the user provides an inline implementation, calling this method should
     // immediately fail.
     static_assert(!sizeof(T), "The get method with given type is not implemented by the simdjson library. "
       "The supported types are ondemand::object, ondemand::array, raw_json_string, std::string_view, uint64_t, "
       "int64_t, double, and bool. We recommend you use get_double(), get_bool(), get_uint64(), get_int64(), "
-      " get_object(), get_array(), get_raw_json_string(), or get_string() instead of the get template.");
+      " get_object(), get_array(), get_raw_json_string(), or get_string() instead of the get template."
+      " You may also add support for custom types, see our documentation.");
   }
 
   /**
@@ -196,7 +197,10 @@ public:
 #if SIMDJSON_EXCEPTIONS
   /**
    * Cast this JSON value to an instance of type T. The programmer is responsible for
-   * providing an implementation of get<T> for the type T.
+   * providing an implementation of get<T> for the type T, if T is not one of the types
+   * supported by the library (object, array, raw_json_string, string_view, uint64_t, etc.).
+   *
+   * See https://github.com/simdjson/simdjson/blob/master/doc/basics.md#adding-support-for-custom-types
    *
    * @returns An instance of type T
    */
@@ -365,7 +369,7 @@ public:
    * that only one field is returned.
    *
    * Use find_field() if you are sure fields will be in order (or are willing to treat it as if the
-   * field wasn't there when they aren't).
+   * field as not there when they are not in order).
    *
    * @param key The key to look up.
    * @returns The value of the field, or NO_SUCH_FIELD if the field is not in the object.
@@ -440,10 +444,12 @@ public:
    * get_number().get_number_type().
    *
    * get_number_type() is number_type::unsigned_integer if we have
-   * an integer greater or equal to 9223372036854775808
+   * an integer greater or equal to 9223372036854775808.
    * get_number_type() is number_type::signed_integer if we have an
-   * integer that is less than 9223372036854775808
-   * Otherwise, get_number_type() has value number_type::floating_point_number
+   * integer that is less than 9223372036854775808.
+   * get_number_type() is number_type::big_integer for integers that do not fit in 64 bits,
+   * in which case the digit_count is set to the length of the big integer string.
+   * Otherwise, get_number_type() has value number_type::floating_point_number.
    *
    * This function requires processing the number string, but it is expected
    * to be faster than get_number().get_number_type() because it is does not
@@ -470,6 +476,8 @@ public:
    * You can recover the value by calling number.get_uint64() and you
    * have that number.is_uint64() is true.
    *
+   * For integers that do not fit in 64 bits, the function returns BIGINT_ERROR error code.
+   *
    * Otherwise, number.get_number_type() has value number_type::floating_point_number
    * and we have a binary64 number.
    * You can recover the value by calling number.get_double() and you
@@ -485,7 +493,6 @@ public:
    */
   simdjson_warn_unused simdjson_inline simdjson_result<number> get_number() noexcept;
 
-
   /**
    * Get the raw JSON for this token.
    *
@@ -729,7 +736,7 @@ public:
    * APIs assume this. Therefore, you must be explicit if you want to treat objects as out of order.
    *
    * Use find_field() if you are sure fields will be in order (or are willing to treat it as if the
-   * field wasn't there when they aren't).
+   * field as not there when they are not in order).
    *
    * @param key The key to look up.
    * @returns The value of the field, or NO_SUCH_FIELD if the field is not in the object.

include/simdjson/generic/ondemand/value_iterator-inl.h

@@ -137,7 +137,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<bool> value_iterator::find_
   } else if (!is_open()) {
 #if SIMDJSON_DEVELOPMENT_CHECKS
     // If we're past the end of the object, we're being iterated out of order.
-    // Note: this isn't perfect detection. It's possible the user is inside some other object; if so,
+    // Note: this is not perfect detection. It's possible the user is inside some other object; if so,
     // this object iterator will blithely scan that object for fields.
     if (_json_iter->depth() < depth() - 1) { return OUT_OF_ORDER_ITERATION; }
 #endif
@@ -249,7 +249,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<bool> value_iterator::find_
 
 #if SIMDJSON_DEVELOPMENT_CHECKS
     // If we're past the end of the object, we're being iterated out of order.
-    // Note: this isn't perfect detection. It's possible the user is inside some other object; if so,
+    // Note: this is not perfect detection. It's possible the user is inside some other object; if so,
     // this object iterator will blithely scan that object for fields.
     if (_json_iter->depth() < depth() - 1) { return OUT_OF_ORDER_ITERATION; }
 #endif
@@ -591,7 +591,7 @@ simdjson_inline simdjson_result<number> value_iterator::get_number() noexcept {
 }
 
 simdjson_inline simdjson_result<bool> value_iterator::is_root_integer(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("is_root_integer");
   uint8_t tmpbuf[20+1+1]{}; // <20 digits> is the longest possible unsigned integer
   tmpbuf[20+1] = '\0'; // make sure that buffer is always null terminated.
@@ -607,7 +607,7 @@ simdjson_inline simdjson_result<bool> value_iterator::is_root_integer(bool check
 }
 
 simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::number_type> value_iterator::get_root_number_type(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("number");
   // Per https://www.exploringbinary.com/maximum-number-of-decimal-digits-in-binary-floating-point-numbers/,
   // 1074 is the maximum number of significant fractional digits. Add 8 more digits for the biggest
@@ -615,7 +615,12 @@ simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::number_type> value_iter
   uint8_t tmpbuf[1074+8+1+1];
   tmpbuf[1074+8+1] = '\0'; // make sure that buffer is always null terminated.
   if (!_json_iter->copy_to_buffer(json, max_len, tmpbuf, 1074+8+1)) {
-    logger::log_error(*_json_iter, start_position(), depth(), "Root number more than 1082 characters");
+    if(numberparsing::check_if_integer(json, max_len)) {
+      if (check_trailing && !_json_iter->is_single_token()) { return TRAILING_CONTENT; }
+      logger::log_error(*_json_iter, start_position(), depth(), "Found big integer");
+      return number_type::big_integer;
+    }
+    logger::log_error(*_json_iter, start_position(), depth(), "Root number more than 1082 characters and not a big integer");
     return NUMBER_ERROR;
   }
   auto answer = numberparsing::get_number_type(tmpbuf);
@@ -623,15 +628,21 @@ simdjson_inline simdjson_result<SIMDJSON_IMPLEMENTATION::number_type> value_iter
   return answer;
 }
 simdjson_inline simdjson_result<number> value_iterator::get_root_number(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("number");
   // Per https://www.exploringbinary.com/maximum-number-of-decimal-digits-in-binary-floating-point-numbers/,
   // 1074 is the maximum number of significant fractional digits. Add 8 more digits for the biggest
   // number: -0.<fraction>e-308.
+  // NOTE: the current approach doesn't work for very big integer numbers containing more than 1074 digits.
   uint8_t tmpbuf[1074+8+1+1];
   tmpbuf[1074+8+1] = '\0'; // make sure that buffer is always null terminated.
   if (!_json_iter->copy_to_buffer(json, max_len, tmpbuf, 1074+8+1)) {
-    logger::log_error(*_json_iter, start_position(), depth(), "Root number more than 1082 characters");
+    if(numberparsing::check_if_integer(json, max_len)) {
+      if (check_trailing && !_json_iter->is_single_token()) { return TRAILING_CONTENT; }
+      logger::log_error(*_json_iter, start_position(), depth(), "Found big integer");
+      return BIGINT_ERROR;
+    }
+    logger::log_error(*_json_iter, start_position(), depth(), "Root number more than 1082 characters and not a big integer");
     return NUMBER_ERROR;
   }
   number num;
@@ -663,7 +674,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<raw_json_string> value_iter
   return raw_json_string(json+1);
 }
 simdjson_warn_unused simdjson_inline simdjson_result<uint64_t> value_iterator::get_root_uint64(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("uint64");
   uint8_t tmpbuf[20+1+1]{}; // <20 digits> is the longest possible unsigned integer
   tmpbuf[20+1] = '\0'; // make sure that buffer is always null terminated.
@@ -679,7 +690,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<uint64_t> value_iterator::g
   return result;
 }
 simdjson_warn_unused simdjson_inline simdjson_result<uint64_t> value_iterator::get_root_uint64_in_string(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("uint64");
   uint8_t tmpbuf[20+1+1]{}; // <20 digits> is the longest possible unsigned integer
   tmpbuf[20+1] = '\0'; // make sure that buffer is always null terminated.
@@ -695,7 +706,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<uint64_t> value_iterator::g
   return result;
 }
 simdjson_warn_unused simdjson_inline simdjson_result<int64_t> value_iterator::get_root_int64(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("int64");
   uint8_t tmpbuf[20+1+1]; // -<19 digits> is the longest possible integer
   tmpbuf[20+1] = '\0'; // make sure that buffer is always null terminated.
@@ -712,7 +723,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<int64_t> value_iterator::ge
   return result;
 }
 simdjson_warn_unused simdjson_inline simdjson_result<int64_t> value_iterator::get_root_int64_in_string(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("int64");
   uint8_t tmpbuf[20+1+1]; // -<19 digits> is the longest possible integer
   tmpbuf[20+1] = '\0'; // make sure that buffer is always null terminated.
@@ -729,7 +740,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<int64_t> value_iterator::ge
   return result;
 }
 simdjson_warn_unused simdjson_inline simdjson_result<double> value_iterator::get_root_double(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("double");
   // Per https://www.exploringbinary.com/maximum-number-of-decimal-digits-in-binary-floating-point-numbers/,
   // 1074 is the maximum number of significant fractional digits. Add 8 more digits for the biggest
@@ -749,7 +760,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<double> value_iterator::get
 }
 
 simdjson_warn_unused simdjson_inline simdjson_result<double> value_iterator::get_root_double_in_string(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("double");
   // Per https://www.exploringbinary.com/maximum-number-of-decimal-digits-in-binary-floating-point-numbers/,
   // 1074 is the maximum number of significant fractional digits. Add 8 more digits for the biggest
@@ -768,7 +779,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<double> value_iterator::get
   return result;
 }
 simdjson_warn_unused simdjson_inline simdjson_result<bool> value_iterator::get_root_bool(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("bool");
   uint8_t tmpbuf[5+1+1]; // +1 for null termination
   tmpbuf[5+1] = '\0'; // make sure that buffer is always null terminated.
@@ -781,7 +792,7 @@ simdjson_warn_unused simdjson_inline simdjson_result<bool> value_iterator::get_r
   return result;
 }
 simdjson_inline simdjson_result<bool> value_iterator::is_root_null(bool check_trailing) noexcept {
-  auto max_len = peek_start_length();
+  auto max_len = peek_root_length();
   auto json = peek_root_scalar("null");
   bool result = (max_len >= 4 && !atomparsing::str4ncmp(json, "null") &&
          (max_len == 4 || jsoncharutils::is_structural_or_whitespace(json[4])));
@@ -853,6 +864,9 @@ simdjson_inline const uint8_t *value_iterator::peek_start() const noexcept {
 simdjson_inline uint32_t value_iterator::peek_start_length() const noexcept {
   return _json_iter->peek_length(start_position());
 }
+simdjson_inline uint32_t value_iterator::peek_root_length() const noexcept {
+  return _json_iter->peek_root_length(start_position());
+}
 
 simdjson_inline const uint8_t *value_iterator::peek_scalar(const char *type) noexcept {
   logger::log_value(*_json_iter, start_position(), depth(), type);

include/simdjson/generic/ondemand/value_iterator.h

@@ -375,6 +375,7 @@ protected:
   simdjson_inline simdjson_result<bool> parse_bool(const uint8_t *json) const noexcept;
   simdjson_inline const uint8_t *peek_start() const noexcept;
   simdjson_inline uint32_t peek_start_length() const noexcept;
+  simdjson_inline uint32_t peek_root_length() const noexcept;
 
   /**
    * The general idea of the advance_... methods and the peek_* methods

include/simdjson/portability.h

@@ -38,6 +38,8 @@
 #define SIMDJSON_IS_ARM64 1
 #elif defined(__riscv) && __riscv_xlen == 64
 #define SIMDJSON_IS_RISCV64 1
+#elif defined(__loongarch_lp64)
+#define SIMDJSON_IS_LOONGARCH64 1
 #elif defined(__PPC64__) || defined(_M_PPC64)
 #if defined(__ALTIVEC__)
 #define SIMDJSON_IS_PPC64_VMX 1

include/simdjson/simdjson_version.h

@@ -4,7 +4,7 @@
 #define SIMDJSON_SIMDJSON_VERSION_H
 
 /** The version of simdjson being used (major.minor.revision) */
-#define SIMDJSON_VERSION "3.7.0"
+#define SIMDJSON_VERSION "3.8.0"
 
 namespace simdjson {
 enum {
@@ -15,7 +15,7 @@ enum {
   /**
    * The minor version (major.MINOR.revision) of simdjson being used.
    */
-  SIMDJSON_VERSION_MINOR = 7,
+  SIMDJSON_VERSION_MINOR = 8,
   /**
    * The revision (major.minor.REVISION) of simdjson being used.
    */

singleheader/simdjson.cpp

@@ -1,4 +1,4 @@
-/* auto-generated on 2024-02-18 11:50:12 -0500. Do not edit! */
+/* auto-generated on 2024-03-10 13:24:07 -0400. Do not edit! */
 /* including simdjson.cpp:  */
 /* begin file simdjson.cpp */
 #define SIMDJSON_SRC_SIMDJSON_CPP
@@ -111,6 +111,8 @@
 #define SIMDJSON_IS_ARM64 1
 #elif defined(__riscv) && __riscv_xlen == 64
 #define SIMDJSON_IS_RISCV64 1
+#elif defined(__loongarch_lp64)
+#define SIMDJSON_IS_LOONGARCH64 1
 #elif defined(__PPC64__) || defined(_M_PPC64)
 #if defined(__ALTIVEC__)
 #define SIMDJSON_IS_PPC64_VMX 1
@@ -2345,6 +2347,7 @@ enum error_code {
   F_ATOM_ERROR,               ///< Problem while parsing an atom starting with the letter 'f'
   N_ATOM_ERROR,               ///< Problem while parsing an atom starting with the letter 'n'
   NUMBER_ERROR,               ///< Problem while parsing a number
+  BIGINT_ERROR,               ///< The integer value exceeds 64 bits
   UTF8_ERROR,                 ///< the input is not valid UTF-8
   UNINITIALIZED,              ///< unknown error, or uninitialized document
   EMPTY,                      ///< no structural element found
@@ -4493,6 +4496,7 @@ namespace internal {
     { F_ATOM_ERROR, "F_ATOM_ERROR: Problem while parsing an atom starting with the letter 'f'" },
     { N_ATOM_ERROR, "N_ATOM_ERROR: Problem while parsing an atom starting with the letter 'n'" },
     { NUMBER_ERROR, "NUMBER_ERROR: Problem while parsing a number" },
+    { BIGINT_ERROR, "BIGINT_ERROR: Big integer value that cannot be represented using 64 bits" },
     { UTF8_ERROR, "UTF8_ERROR: The input is not valid UTF-8" },
     { UNINITIALIZED, "UNINITIALIZED: Uninitialized" },
     { EMPTY, "EMPTY: no JSON found" },
@@ -8213,7 +8217,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace arm64
@@ -8661,11 +8666,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -8996,6 +9003,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -9075,6 +9086,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -9149,6 +9177,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -9227,11 +9267,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -9701,19 +9741,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -9891,6 +9944,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -13019,8 +13073,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -13077,8 +13131,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must
@@ -14501,7 +14555,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace haswell
@@ -14949,11 +15004,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -15284,6 +15341,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -15363,6 +15424,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -15437,6 +15515,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -15515,11 +15605,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -15989,19 +16079,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -16179,6 +16282,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -19184,8 +19288,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -19242,8 +19346,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must
@@ -20661,7 +20765,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace icelake
@@ -21109,11 +21214,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -21444,6 +21551,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -21523,6 +21634,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -21597,6 +21725,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -21675,11 +21815,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -22149,19 +22289,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -22339,6 +22492,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -25342,8 +25496,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -25400,8 +25554,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must
@@ -26977,7 +27131,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace ppc64
@@ -27425,11 +27580,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -27760,6 +27917,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -27839,6 +28000,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -27913,6 +28091,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -27991,11 +28181,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -28465,19 +28655,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -28655,6 +28858,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -31771,8 +31975,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -31829,8 +32033,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must
@@ -33659,7 +33863,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace westmere
@@ -34107,11 +34312,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -34442,6 +34649,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -34521,6 +34732,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -34595,6 +34823,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -34673,11 +34913,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -35147,19 +35387,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -35337,6 +35590,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -38774,8 +39028,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -38832,8 +39086,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must
@@ -39768,7 +40022,8 @@ class dom_parser_implementation;
 enum class number_type {
     floating_point_number=1, /// a binary64 number
     signed_integer,          /// a signed integer that fits in a 64-bit word using two's complement
-    unsigned_integer         /// a positive integer larger or equal to 1<<63
+    unsigned_integer,        /// a positive integer larger or equal to 1<<63
+    big_integer              /// a big integer that does not fit in a 64-bit word
 };
 
 } // namespace fallback
@@ -40216,11 +40471,13 @@ namespace numberparsing {
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (found_integer((VALUE), (SRC)), (WRITER).append_s64((VALUE)))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (found_unsigned_integer((VALUE), (SRC)), (WRITER).append_u64((VALUE)))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (found_float((VALUE), (SRC)), (WRITER).append_double((VALUE)))
+#define BIGINT_NUMBER(SRC) (found_invalid_number((SRC)), BIGINT_ERROR)
 #else
 #define INVALID_NUMBER(SRC) (NUMBER_ERROR)
 #define WRITE_INTEGER(VALUE, SRC, WRITER) (WRITER).append_s64((VALUE))
 #define WRITE_UNSIGNED(VALUE, SRC, WRITER) (WRITER).append_u64((VALUE))
 #define WRITE_DOUBLE(VALUE, SRC, WRITER) (WRITER).append_double((VALUE))
+#define BIGINT_NUMBER(SRC) (BIGINT_ERROR)
 #endif
 
 namespace {
@@ -40551,6 +40808,10 @@ simdjson_inline bool parse_digit(const uint8_t c, I &i) {
   return true;
 }
 
+simdjson_inline bool is_digit(const uint8_t c) {
+  return static_cast<uint8_t>(c - '0') <= 9;
+}
+
 simdjson_inline error_code parse_decimal_after_separator(simdjson_unused const uint8_t *const src, const uint8_t *&p, uint64_t &i, int64_t &exponent) {
   // we continue with the fiction that we have an integer. If the
   // floating point number is representable as x * 10^z for some integer
@@ -40630,6 +40891,23 @@ simdjson_inline error_code parse_exponent(simdjson_unused const uint8_t *const s
   return SUCCESS;
 }
 
+simdjson_inline bool check_if_integer(const uint8_t *const src, size_t max_length) {
+  const uint8_t *const srcend = src + max_length;
+  bool negative = (*src == '-'); // we can always read at least one character after the '-'
+  const uint8_t *p = src + uint8_t(negative);
+  if(p == srcend) { return false; }
+  if(*p == '0') {
+    ++p;
+    if(p == srcend) { return true; }
+    if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+    return true;
+  }
+  while(p != srcend && is_digit(*p)) { ++p; }
+  if(p == srcend) { return true; }
+  if(jsoncharutils::is_not_structural_or_whitespace(*p)) { return false; }
+  return true;
+}
+
 simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t digit_count) {
   // It is possible that the integer had an overflow.
   // We have to handle the case where we have 0.0000somenumber.
@@ -40704,6 +40982,18 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
   return SUCCESS;
 }
 
+// parse the number at src
+// define JSON_TEST_NUMBERS for unit testing
+//
+// It is assumed that the number is followed by a structural ({,},],[) character
+// or a white space character. If that is not the case (e.g., when the JSON
+// document is made of a single number), then it is necessary to copy the
+// content and append a space before calling this function.
+//
+// Our objective is accurate parsing (ULP of 0) at high speed.
+template<typename W>
+simdjson_inline error_code parse_number(const uint8_t *const src, W &writer);
+
 // for performance analysis, it is sometimes  useful to skip parsing
 #ifdef SIMDJSON_SKIPNUMBERPARSING
 
@@ -40782,11 +41072,11 @@ simdjson_inline error_code parse_number(const uint8_t *const src, W &writer) {
   // The longest positive 64-bit number is 20 digits.
   // We do it this way so we don't trigger this branch unless we must.
   size_t longest_digit_count = negative ? 19 : 20;
-  if (digit_count > longest_digit_count) { return INVALID_NUMBER(src); }
+  if (digit_count > longest_digit_count) { return BIGINT_NUMBER(src); }
   if (digit_count == longest_digit_count) {
     if (negative) {
       // Anything negative above INT64_MAX+1 is invalid
-      if (i > uint64_t(INT64_MAX)+1) { return INVALID_NUMBER(src);  }
+      if (i > uint64_t(INT64_MAX)+1) { return BIGINT_NUMBER(src);  }
       WRITE_INTEGER(~i+1, src, writer);
       if (jsoncharutils::is_not_structural_or_whitespace(*p)) { return INVALID_NUMBER(src); }
       return SUCCESS;
@@ -41256,19 +41546,32 @@ simdjson_unused simdjson_inline simdjson_result<number_type> get_number_type(con
   src += uint8_t(negative);
   const uint8_t *p = src;
   while(static_cast<uint8_t>(*p - '0') <= 9) { p++; }
+  size_t digit_count = size_t(p - src);
   if ( p == src ) { return NUMBER_ERROR; }
   if (jsoncharutils::is_structural_or_whitespace(*p)) {
+    static const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
     // We have an integer.
+    if(simdjson_unlikely(digit_count > 20)) {
+      return number_type::big_integer;
+    }
     // If the number is negative and valid, it must be a signed integer.
-    if(negative) { return number_type::signed_integer; }
+    if(negative) {
+      if (simdjson_unlikely(digit_count > 19)) return number_type::big_integer;
+      if (simdjson_unlikely(digit_count == 19 && memcmp(src, smaller_big_integer, 19) > 0)) {
+        return number_type::big_integer;
+      }
+      return number_type::signed_integer;
+    }
+    // Let us check if we have a big integer (>=2**64).
+    static const uint8_t * two_to_sixtyfour = reinterpret_cast<const uint8_t *>("18446744073709551616");
+    if((digit_count > 20) || (digit_count == 20 && memcmp(src, two_to_sixtyfour, 20) >= 0)) {
+      return number_type::big_integer;
+    }
+    // The number is positive and smaller than 18446744073709551616 (or 2**64).
     // We want values larger or equal to 9223372036854775808 to be unsigned
     // integers, and the other values to be signed integers.
-    int digit_count = int(p - src);
-    if(digit_count >= 19) {
-      const uint8_t * smaller_big_integer = reinterpret_cast<const uint8_t *>("9223372036854775808");
-      if((digit_count >= 20) || (memcmp(src, smaller_big_integer, 19) >= 0)) {
-        return number_type::unsigned_integer;
-      }
+    if((digit_count == 20) || (digit_count >= 19 && memcmp(src, smaller_big_integer, 19) >= 0)) {
+      return number_type::unsigned_integer;
     }
     return number_type::signed_integer;
   }
@@ -41446,6 +41749,7 @@ inline std::ostream& operator<<(std::ostream& out, number_type type) noexcept {
         case number_type::signed_integer: out << "integer in [-9223372036854775808,9223372036854775808)"; break;
         case number_type::unsigned_integer: out << "unsigned integer in [9223372036854775808,18446744073709551616)"; break;
         case number_type::floating_point_number: out << "floating-point number (binary64)"; break;
+        case number_type::big_integer: out << "big integer"; break;
         default: SIMDJSON_UNREACHABLE();
     }
     return out;
@@ -41966,8 +42270,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -42024,8 +42328,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must

src/generic/stage2/stringparsing.h

@@ -53,8 +53,8 @@ simdjson_inline bool handle_unicode_codepoint(const uint8_t **src_ptr,
   // Use the default Unicode Character 'REPLACEMENT CHARACTER' (U+FFFD)
   constexpr uint32_t substitution_code_point = 0xfffd;
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
 
@@ -111,8 +111,8 @@ simdjson_inline bool handle_unicode_codepoint_wobbly(const uint8_t **src_ptr,
   // It is not ideal that this function is nearly identical to handle_unicode_codepoint.
   //
   // jsoncharutils::hex_to_u32_nocheck fills high 16 bits of the return value with 1s if the
-  // conversion isn't valid; we defer the check for this to inside the
-  // multilingual plane check
+  // conversion is not valid; we defer the check for this to inside the
+  // multilingual plane check.
   uint32_t code_point = jsoncharutils::hex_to_u32_nocheck(*src_ptr + 2);
   *src_ptr += 6;
   // If we found a high surrogate, we must

src/internal/error_tables.cpp

@@ -18,6 +18,7 @@ namespace internal {
     { F_ATOM_ERROR, "F_ATOM_ERROR: Problem while parsing an atom starting with the letter 'f'" },
     { N_ATOM_ERROR, "N_ATOM_ERROR: Problem while parsing an atom starting with the letter 'n'" },
     { NUMBER_ERROR, "NUMBER_ERROR: Problem while parsing a number" },
+    { BIGINT_ERROR, "BIGINT_ERROR: Big integer value that cannot be represented using 64 bits" },
     { UTF8_ERROR, "UTF8_ERROR: The input is not valid UTF-8" },
     { UNINITIALIZED, "UNINITIALIZED: Uninitialized" },
     { EMPTY, "EMPTY: no JSON found" },

tests/ondemand/ondemand_array_error_tests.cpp

@@ -102,7 +102,7 @@ namespace array_error_tests {
     ONDEMAND_SUBTEST("unclosed extra comma", R"({ "a": [,)",  assert_iterate(doc["a"],                 { INCOMPLETE_ARRAY_OR_OBJECT }));
     ONDEMAND_SUBTEST("unclosed extra comma", R"({ "a": [,,)", assert_iterate(doc["a"],                 { INCOMPLETE_ARRAY_OR_OBJECT }));
     ONDEMAND_SUBTEST("unclosed     ", R"({ "a": [1 )",        assert_iterate(doc["a"],                 { INCOMPLETE_ARRAY_OR_OBJECT }));
-    // TODO These pass the user values that may run past the end of the buffer if they aren't careful
+    // TODO These pass the user values that may run past the end of the buffer if they are not careful
     // In particular, if the padding is decorated with the wrong values, we could cause overrun!
     ONDEMAND_SUBTEST("unclosed     ", R"({ "a": [1,)",        assert_iterate(doc["a"],                 { INCOMPLETE_ARRAY_OR_OBJECT }));
     ONDEMAND_SUBTEST("unclosed     ", R"({ "a": [1)",         assert_iterate(doc["a"],                 { INCOMPLETE_ARRAY_OR_OBJECT }));

tests/ondemand/ondemand_document_stream_tests.cpp

@@ -651,6 +651,26 @@ namespace document_stream_tests {
     }
 
 
+    bool issue2137() {
+        TEST_START();
+        auto json = "true  {  "_padded;
+        ondemand::parser parser;
+        ondemand::document_stream stream;
+        ASSERT_SUCCESS(parser.iterate_many(json).get(stream));
+        for (auto doc: stream) {
+            bool val{};
+            ASSERT_SUCCESS(doc.get_bool().get(val));
+            std::string_view raw_json;
+            ASSERT_SUCCESS(doc.raw_json_token().get(raw_json));
+            ASSERT_EQUAL(val, 1);
+            std::string s(raw_json);
+            ASSERT_EQUAL(s, "true  ");
+        }
+        size_t t = stream.truncated_bytes();
+        ASSERT_EQUAL(3, t);
+        TEST_SUCCEED();
+    }
+
     bool issue1668() {
         TEST_START();
         auto json = R"([1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100])"_padded;
@@ -861,6 +881,7 @@ namespace document_stream_tests {
 
     bool run() {
         return
+            issue2137() &&
             skipbom() &&
             issue1977() &&
             string_with_trailing() &&

tests/ondemand/ondemand_error_tests.cpp

@@ -106,6 +106,29 @@ namespace error_tests {
     TEST_SUCCEED();
   }
 #if SIMDJSON_EXCEPTIONS
+  // This is a compile-only test
+  struct Class {
+      Class(std::string text)
+          : parser()
+          , doc(parser.iterate(text))
+      {
+      }
+      simdjson::ondemand::parser parser;
+      simdjson::ondemand::document doc;
+  };
+  bool document_in_class() {
+    TEST_START();
+    std::string text = "{}";
+    Class c(text);
+    TEST_SUCCEED();
+  }
+  bool direct_document() {
+    TEST_START();
+    std::string text = "{}";
+    simdjson::ondemand::parser parser;
+    simdjson::ondemand::document doc(parser.iterate(text));
+    TEST_SUCCEED();
+  }
   bool raw_json_string_except() {
     TEST_START();
     ondemand::parser parser;
@@ -381,6 +404,8 @@ namespace error_tests {
            issue1834() &&
            issue1834_2() &&
 #if SIMDJSON_EXCEPTIONS
+           document_in_class() &&
+           direct_document() &&
            raw_json_string_except() &&
            raw_json_string_except_with_io() &&
 #endif

tests/ondemand/ondemand_number_tests.cpp

@@ -211,7 +211,7 @@ namespace number_tests {
   bool get_number_tests() {
     TEST_START();
     ondemand::parser parser;
-    padded_string docdata = R"([1.0, 3, 1, 3.1415,-13231232,9999999999999999999,-9223372036854775807,-9223372036854775808])"_padded;
+    padded_string docdata = R"([1.0, 3, 1, 3.1415,-13231232,9999999999999999999,-9223372036854775807,-9223372036854775808,12345678901234567890123])"_padded;
     ondemand::number_type expectedtypes[] = {ondemand::number_type::floating_point_number,
                                  ondemand::number_type::signed_integer,
                                  ondemand::number_type::signed_integer,
@@ -219,10 +219,11 @@ namespace number_tests {
                                  ondemand::number_type::signed_integer,
                                  ondemand::number_type::unsigned_integer,
                                  ondemand::number_type::signed_integer,
-                                 ondemand::number_type::signed_integer
+                                 ondemand::number_type::signed_integer,
+                                 ondemand::number_type::big_integer
                                  };
-    bool is_negative[] = {false, false, false, false, true, false, true, true};
-    bool is_integer[] = {false, true, true, false, true, true, true, true};
+    bool is_negative[] = {false, false, false, false, true, false, true, true, false};
+    bool is_integer[] = {false, true, true, false, true, true, true, true, true};
 
     ondemand::document doc;
     ASSERT_SUCCESS(parser.iterate(docdata).get(doc));
@@ -233,17 +234,25 @@ namespace number_tests {
       ondemand::value val;
       ASSERT_SUCCESS(valr.get(val));
       ondemand::number_type nt{};
-      ASSERT_SUCCESS(val.get_number_type().get(nt));
+      auto r = val.get_number_type().get(nt);
+      ASSERT_SUCCESS(r);
       ASSERT_EQUAL(expectedtypes[counter], nt);
-      ondemand::number num;
-      ASSERT_SUCCESS(val.get_number().get(num));
-      ASSERT_EQUAL(is_negative[counter], val.is_negative());
       bool intvalue{};
-      ASSERT_SUCCESS(val.is_integer().get(intvalue));
+      auto res = val.is_integer().get(intvalue);
+      ASSERT_SUCCESS(res);
+      ondemand::number num;
+      auto rr = val.get_number().get(num);
+      if(counter == 8) {
+        ASSERT_EQUAL(rr, BIGINT_ERROR);
+        ASSERT_EQUAL(nt, ondemand::number_type::big_integer);
+      } else {
+        ASSERT_SUCCESS(rr);
+        ondemand::number_type t = num.get_number_type();
+        ASSERT_EQUAL(expectedtypes[counter], t);
+      }
       ASSERT_EQUAL(is_integer[counter], intvalue);
-      ondemand::number_type t = num.get_number_type();
-      ASSERT_EQUAL(expectedtypes[counter], t);
-      switch(t) {
+      ASSERT_EQUAL(is_negative[counter], val.is_negative());
+      switch(nt) {
         case ondemand::number_type::signed_integer:
           ASSERT_TRUE(num.is_int64());
           break;
@@ -253,25 +262,39 @@ namespace number_tests {
         case ondemand::number_type::floating_point_number:
           ASSERT_TRUE(num.is_double());
           break;
+        case ondemand::number_type::big_integer:
+          break;
       }
-      if(counter == 0) {
-        ASSERT_EQUAL(num.get_double(), 1.0);
-        ASSERT_EQUAL((double)num, 1.0);
-      } else if(counter == 1) {
-        ASSERT_EQUAL(num.get_int64(), 3);
-        ASSERT_EQUAL((int64_t)num, 3);
-      } else if(counter == 2) {
-        ASSERT_EQUAL(num.get_int64(), 1);
-        ASSERT_EQUAL((int64_t)num, 1);
-      } else if(counter == 3) {
-        ASSERT_EQUAL(num.get_double(), 3.1415);
-        ASSERT_EQUAL((double)num, 3.1415);
-      } else if(counter == 4) {
-        ASSERT_EQUAL(num.get_int64(), -13231232);
-        ASSERT_EQUAL((int64_t)num, -13231232);
-      } else if(counter == 5) {
-        ASSERT_EQUAL(num.get_uint64(), UINT64_C(9999999999999999999));
-        ASSERT_EQUAL((uint64_t)num, UINT64_C(9999999999999999999));
+      switch(counter) {
+        case 0:
+          ASSERT_EQUAL(num.get_double(), 1.0);
+          ASSERT_EQUAL((double)num, 1.0);
+          break;
+        case 1:
+          ASSERT_EQUAL(num.get_int64(), 3);
+          ASSERT_EQUAL((int64_t)num, 3);
+          break;
+        case 2:
+          ASSERT_EQUAL(num.get_int64(), 1);
+          ASSERT_EQUAL((int64_t)num, 1);
+          break;
+        case 3:
+          ASSERT_EQUAL(num.get_double(), 3.1415);
+          ASSERT_EQUAL((double)num, 3.1415);
+          break;
+        case 4:
+          ASSERT_EQUAL(num.get_int64(), -13231232);
+          ASSERT_EQUAL((int64_t)num, -13231232);
+          break;
+        case 5:
+          ASSERT_EQUAL(num.get_uint64(), UINT64_C(9999999999999999999));
+          ASSERT_EQUAL((uint64_t)num, UINT64_C(9999999999999999999));
+          break;
+        case 8:
+          ASSERT_EQUAL(val.raw_json_token(), "12345678901234567890123");
+          break;
+        default:
+          break;
       }
       counter++;
     }
@@ -441,8 +464,50 @@ namespace number_tests {
     TEST_SUCCEED();
   }
 
+  bool big_int_not_zero() {
+    TEST_START();
+    ondemand::parser parser;
+    ondemand::document doc;
+    // This is not a big integer, it is a mistake
+    padded_string docdata = R"(09500000000000000000000000000000000000)"_padded;
+    ASSERT_SUCCESS(parser.iterate(docdata).get(doc));
+    ASSERT_ERROR(doc.get_number(), NUMBER_ERROR);
+    TEST_SUCCEED();
+  }
+
+  bool negative_big_int() {
+    TEST_START();
+    ondemand::parser parser;
+    ondemand::document doc;
+    // This is not a big integer, it is a mistake
+    padded_string docdata = R"(-18446744073709551616)"_padded;
+    ASSERT_SUCCESS(parser.iterate(docdata).get(doc));
+    ASSERT_ERROR(doc.get_number(), BIGINT_ERROR);
+    std::string_view my_big;
+    ASSERT_SUCCESS(doc.raw_json_token().get(my_big));
+    ASSERT_EQUAL(my_big, "-18446744073709551616");
+    TEST_SUCCEED();
+  }
+
+  bool gigantic_big_int() {
+    TEST_START();
+    ondemand::parser parser;
+    ondemand::document doc;
+    std::string number(2000, '1');
+    // This is not a big integer, it is a mistake
+    ASSERT_SUCCESS(parser.iterate(number).get(doc));
+    ASSERT_ERROR(doc.get_number(), BIGINT_ERROR);
+    std::string_view my_big;
+    ASSERT_SUCCESS(doc.raw_json_token().get(my_big));
+    ASSERT_EQUAL(my_big, number);
+    TEST_SUCCEED();
+  }
+
   bool run() {
-    return issue2099() &&
+    return gigantic_big_int() &&
+           big_int_not_zero() &&
+           negative_big_int() &&
+           issue2099() &&
            issue2093() &&
            issue2045() &&
            issue2017() &&

tests/ondemand/ondemand_object_error_tests.cpp

@@ -119,7 +119,7 @@ namespace object_error_tests {
   bool object_lookup_miss_unclosed_error() {
     TEST_START();
     ONDEMAND_SUBTEST("unclosed", R"({ "a": 1,         )",    assert_error(doc["b"], INCOMPLETE_ARRAY_OR_OBJECT));
-    // TODO These next two pass the user a value that may run past the end of the buffer if they aren't careful.
+    // TODO These next two pass the user a value that may run past the end of the buffer if they are not careful.
     // In particular, if the padding is decorated with the wrong values, we could cause overrun!
     ONDEMAND_SUBTEST("unclosed", R"({ "a": 1          )",    assert_error(doc["b"], INCOMPLETE_ARRAY_OR_OBJECT));
     ONDEMAND_SUBTEST("unclosed", R"({ "a":            )",    assert_error(doc["b"], INCOMPLETE_ARRAY_OR_OBJECT));

tests/ondemand/ondemand_readme_examples.cpp

@@ -3,6 +3,9 @@
 #if __cpp_lib_optional >= 201606L
 #include <optional>
 #endif
+#if SIMDJSON_CPLUSPLUS17
+#include <charconv>
+#endif
 using namespace std;
 using namespace simdjson;
 using error_code=simdjson::error_code;
@@ -382,22 +385,84 @@ bool examplecrt_realloc() {
   TEST_SUCCEED();
 }
 
+#if SIMDJSON_CPLUSPLUS17
+bool big_int_array() {
+  TEST_START();
+  ondemand::parser parser;
+  padded_string docdata = R"([-9223372036854775809, 18446744073709551617, 99999999999999999999999 ])"_padded;
+  std::string expected[] = {"-9223372036854775809", "18446744073709551617", "99999999999999999999999 "};
+  ondemand::document doc = parser.iterate(docdata);
+  ondemand::array arr = doc.get_array();
+  size_t i = 0;
+  for(ondemand::value val : arr) {
+    if(i > 3) {
+      std::cerr << "unexpected number of elements" << std::endl;
+      return false;
+    }
+    if(val.get_number_type() != ondemand::number_type::big_integer) {
+      std::cerr << "unexpected number type" << std::endl;
+      std::cout << val.get_number_type() << std::endl;
+      std::cout << val.raw_json_token() << std::endl;
+      return false;
+    }
+    std::string_view token = val.raw_json_token();
+    std::string_view expected_token = expected[i];
+    if(token != expected_token) {
+      std::cerr << "unexpected token: " << token << " expected: " << expected_token << std::endl;
+      return false;
+    }
+    i++;
+  }
+  if(i != 3) {
+    std::cerr << "unexpected number of elements" << std::endl;
+    return false;
+  }
+  TEST_SUCCEED();
+}
+#endif
+
+bool big_int_array_as_double() {
+  TEST_START();
+  ondemand::parser parser;
+  padded_string docdata = R"([-9223372036854775809, 18446744073709551617, 99999999999999999999999 ])"_padded;
+  double dexpected[] = {-9223372036854775808.0, 18446744073709551616.0, 1e23};
+  ondemand::document doc = parser.iterate(docdata);
+  ondemand::array arr = doc.get_array();
+  size_t i = 0;
+  for(ondemand::value val : arr) {
+    if(i > 3) {
+      std::cerr << "unexpected number of elements" << std::endl;
+      return false;
+    }
+    if((val.get_number_type() != ondemand::number_type::big_integer) || (dexpected[i] != val.get_double())) {
+      return false;
+    }
+    i++;
+  }
+  if(i != 3) {
+    std::cerr << "unexpected number of elements" << std::endl;
+    return false;
+  }
+  TEST_SUCCEED();
+}
+
 bool number_tests() {
   TEST_START();
   ondemand::parser parser;
-  padded_string docdata = R"([1.0, 3, 1, 3.1415,-13231232,9999999999999999999])"_padded;
+  padded_string docdata = R"([1.0, 3, 1, 3.1415,-13231232,9999999999999999999,12345678901234567890123])"_padded;
   ondemand::document doc = parser.iterate(docdata);
   ondemand::array arr = doc.get_array();
   for(ondemand::value val : arr) {
     std::cout << val << " ";
     std::cout << "negative: " << val.is_negative() << " ";
     std::cout << "is_integer: " << val.is_integer() << " ";
-    ondemand::number num = val.get_number();
-    ondemand::number_type t = num.get_number_type();
     // direct computation without materializing the number:
     ondemand::number_type dt = val.get_number_type();
-    if(t != dt) { throw std::runtime_error("bug"); }
-    switch(t) {
+    ondemand::number num;
+    auto res = val.get_number().get(num);
+    ondemand::number_type t = num.get_number_type();
+    if(t != dt && res != BIGINT_ERROR) { throw std::runtime_error("bug"); }
+    switch(dt) {
       case ondemand::number_type::signed_integer:
         std::cout  << "integer: " << int64_t(num) << " ";
         std::cout  << "integer: " << num.get_int64() << std::endl;
@@ -410,6 +475,9 @@ bool number_tests() {
         std::cout  << "float: " << double(num) << " ";
         std::cout << "float: " << num.get_double() << std::endl;
         break;
+      case ondemand::number_type::big_integer:
+        std::cout  << "big-integer: " << val.raw_json_token() << std::endl;
+        break;
     }
   }
   TEST_SUCCEED();
@@ -1778,6 +1846,10 @@ bool value_raw_json_object() {
 bool run() {
   return true
 #if SIMDJSON_EXCEPTIONS
+#if SIMDJSON_CPLUSPLUS17
+    && big_int_array()
+#endif // SIMDJSON_CPLUSPLUS17
+    && big_int_array_as_double()
     && key_raw_json_token()
     && to_optional()
     && value_raw_json_array() && value_raw_json_object()
@@ -1785,7 +1857,7 @@ bool run() {
     && at_end()
     && example1956() && example1958()
     && allow_comma_separated_example()
-//    && basics_1() // Fails because twitter.json isn't in current directory. Compile test only.
+//    && basics_1() // Fails because twitter.json is not in current directory. Compile test only.
     &&  basics_treewalk()
     &&  basics_treewalk_breakline()
     && json_value_with_array_count()

tools/release.py

@@ -16,7 +16,7 @@ def colored(r, g, b, text):
     return "\033[38;2;{};{};{}m{} \033[38;2;255;255;255m".format(r, g, b, text)
 
 def extractnumbers(s):
-    return tuple(map(int,re.findall("(\d+)\.(\d+)\.(\d+)",str(s))[0]))
+    return tuple(map(int,re.findall(r"(\d+)\.(\d+)\.(\d+)",str(s))[0]))
 
 def toversionstring(major, minor, rev):
     return str(major)+"."+str(minor)+"."+str(rev)
@@ -123,7 +123,7 @@ newversionstring = str(newversion[0]) + "." + str(newversion[1]) + "." + str(new
 cmakefile = maindir + os.sep + "CMakeLists.txt"
 
 sonumber = None
-pattern = re.compile("set\(SIMDJSON_LIB_SOVERSION \"(\d+)\" CACHE STRING \"simdjson library soversion\"\)")
+pattern = re.compile(r"set\(SIMDJSON_LIB_SOVERSION \"(\d+)\" CACHE STRING \"simdjson library soversion\"\)")
 with open (cmakefile, 'rt') as myfile:
     for line in myfile:
         m = pattern.search(line)
@@ -138,9 +138,9 @@ if(atleastminor):
     sonumber += 1
 
 for line in fileinput.input(cmakefile, inplace=1, backup='.bak'):
-    line = re.sub('    VERSION \d+\.\d+\.\d+','    VERSION '+newmajorversionstring+'.'+mewminorversionstring+'.'+newrevversionstring, line.rstrip())
-    line = re.sub('SIMDJSON_LIB_VERSION "\d+','SIMDJSON_LIB_VERSION "'+str(sonumber), line)
-    line = re.sub('set\(SIMDJSON_LIB_SOVERSION \"\d+\"','set(SIMDJSON_LIB_SOVERSION \"'+str(sonumber)+'\"', line)
+    line = re.sub(r'    VERSION \d+\.\d+\.\d+','    VERSION '+newmajorversionstring+'.'+mewminorversionstring+'.'+newrevversionstring, line.rstrip())
+    line = re.sub(r'SIMDJSON_LIB_VERSION "\d+','SIMDJSON_LIB_VERSION "'+str(sonumber), line)
+    line = re.sub(r'set\(SIMDJSON_LIB_SOVERSION \"\d+\"','set(SIMDJSON_LIB_SOVERSION \"'+str(sonumber)+'\"', line)
     print(line)
 
 print("modified "+cmakefile+", a backup was made")
@@ -148,7 +148,7 @@ print("modified "+cmakefile+", a backup was made")
 
 doxyfile = maindir + os.sep + "Doxyfile"
 for line in fileinput.input(doxyfile, inplace=1, backup='.bak'):
-    line = re.sub('PROJECT_NUMBER         = "\d+\.\d+\.\d+','PROJECT_NUMBER         = "'+newversionstring, line.rstrip())
+    line = re.sub(r'PROJECT_NUMBER         = "\d+\.\d+\.\d+','PROJECT_NUMBER         = "'+newversionstring, line.rstrip())
     print(line)
 print("modified "+doxyfile+", a backup was made")
 
@@ -168,7 +168,7 @@ if(cp.returncode != 0):
     print("Failed to run doxygen")
 
 
-pattern = re.compile("https://simdjson.org/api/(\d+\.\d+\.\d+)/index.html")
+pattern = re.compile(r"https://simdjson.org/api/(\d+\.\d+\.\d+)/index.html")
 readmefile = maindir + os.sep + "README.md"
 readmedata = open(readmefile).read()
 m = pattern.search(readmedata)