version bump

* [utf8-validator] eliminate unnecessary comparison from must_be_2_3_continuation

* Fix comment in

.github/workflows/alpine.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/debian.yaml

@@ -16,7 +16,7 @@ jobs:
       image: debian:testing
 
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - name: Install dependencies
       run: |

.github/workflows/documentation.yml

@@ -19,7 +19,7 @@ jobs:
       url: ${{ steps.deployment.outputs.page_url }}
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - name: Install Doxygen
         run: sudo apt-get install doxygen graphviz -y
       - run: mkdir docs

.github/workflows/fix-trailing-whitespace.yml

@@ -6,7 +6,7 @@ jobs:
   whitespace:
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - name: Remove whitespace and check the diff
         run: |
           set -eu

.github/workflows/fuzzers.yml

@@ -38,7 +38,7 @@ jobs:
         chmod +x llvm.sh
         sudo ./llvm.sh $CLANGVERSION
 
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
 
     - uses: actions/cache@v3
       with:

.github/workflows/macos.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: macos-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/msys2-clang.yml

@@ -24,7 +24,7 @@ jobs:
       CMAKE_GENERATOR: Ninja
 
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/msys2.yml

@@ -26,7 +26,7 @@ jobs:
       CMAKE_GENERATOR: Ninja
 
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/power-fuzz.yml

@@ -15,7 +15,7 @@ jobs:
     runs-on: ubuntu-20.04
     name: Build on ubuntu-20.04 ppc64le
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: uraimo/run-on-arch-action@v2.0.5
         name: Run commands
         id: runcmd

.github/workflows/ppc64.yml

@@ -12,7 +12,7 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: uraimo/run-on-arch-action@v2
         name: Test
         id: runcmd

.github/workflows/riscv64.yml

@@ -12,7 +12,7 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: uraimo/run-on-arch-action@v2
         name: Test
         id: runcmd

.github/workflows/s390x.yml

@@ -12,7 +12,7 @@ jobs:
   build:
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: uraimo/run-on-arch-action@v2
         name: Test
         id: runcmd

.github/workflows/ubuntu20-checkperf.yml

@@ -15,7 +15,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu20-gcc8.yml

@@ -12,7 +12,7 @@ jobs:
       CXX: g++-8
       CC: gcc-8
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu20-noexcept.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu20-nothread.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu20-sani.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache
@@ -27,7 +27,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu20.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-20.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu22-clang13.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu22-clang14.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu22-cxx20.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu22-gcc12.yml

@@ -9,13 +9,11 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache
           key: ${{ hashFiles('dependencies/CMakeLists.txt') }}
-      - name: Install gcc12
-        run: sudo apt-get install -y g++-12
       - name: Use cmake
         run: |
           mkdir build &&

.github/workflows/ubuntu22-gcc13.yml

@@ -0,0 +1,23 @@
+name: Ubuntu 22.04 CI (GCC 13)
+
+on: [push, pull_request]
+
+jobs:
+  ubuntu-build:
+    if: >-
+      ! contains(toJSON(github.event.commits.*.message), '[skip ci]') &&
+      ! contains(toJSON(github.event.commits.*.message), '[skip github]')
+    runs-on: ubuntu-22.04
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/cache@v3
+        with:
+          path: dependencies/.cache
+          key: ${{ hashFiles('dependencies/CMakeLists.txt') }}
+      - name: Use cmake
+        run: |
+          mkdir build &&
+          cd build &&
+          CXX=g++-13 cmake -DSIMDJSON_DEVELOPER_MODE=ON ..  &&
+          cmake --build .   &&
+          ctest --output-on-failure -LE explicitonly -j

.github/workflows/ubuntu22-glibcxxassertions.yml

@@ -8,7 +8,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu22-threadsani.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/ubuntu22.yml

@@ -9,7 +9,7 @@ jobs:
       ! contains(toJSON(github.event.commits.*.message), '[skip github]')
     runs-on: ubuntu-22.04
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - uses: actions/cache@v3
         with:
           path: dependencies/.cache

.github/workflows/vs17-arm-ci.yml

@@ -14,7 +14,7 @@ jobs:
            - {arch: ARM64}
      steps:
        - name: checkout
-         uses: actions/checkout@v3
+         uses: actions/checkout@v4
        - name: Use cmake
          run: |
            cmake -A ${{ matrix.arch }} -DCMAKE_CROSSCOMPILING=1 -DSIMDJSON_DEVELOPER_MODE=ON -D SIMDJSON_GOOGLE_BENCHMARKS=OFF -DSIMDJSON_EXCEPTIONS=OFF -B build  &&

.github/workflows/vs17-ci-cxx20.yml

@@ -19,7 +19,7 @@ jobs:
           - {gen: Visual Studio 17 2022, arch: x64, shared: OFF}
     steps:
     - name: checkout
-      uses: actions/checkout@v3
+      uses: actions/checkout@v4
     - name: Configure
       run: |
         cmake -DSIMDJSON_CXX_STANDARD=20 -G "${{matrix.gen}}" -A ${{matrix.arch}} -DSIMDJSON_DEVELOPER_MODE=ON -DSIMDJSON_COMPETITION=OFF -DBUILD_SHARED_LIBS=${{matrix.shared}} -B build

.github/workflows/vs17-ci.yml

@@ -19,7 +19,7 @@ jobs:
           - {gen: Visual Studio 17 2022, arch: x64, shared: OFF}
     steps:
     - name: checkout
-      uses: actions/checkout@v3
+      uses: actions/checkout@v4
     - name: Configure
       run: |
         cmake -G "${{matrix.gen}}" -A ${{matrix.arch}} -DSIMDJSON_DEVELOPER_MODE=ON -DSIMDJSON_COMPETITION=OFF -DBUILD_SHARED_LIBS=${{matrix.shared}} -B build

.github/workflows/vs17-clang-ci.yml

@@ -16,7 +16,7 @@ jobs:
           - {gen: Visual Studio 17 2022, arch: x64}
     steps:
     - name: checkout
-      uses: actions/checkout@v3
+      uses: actions/checkout@v4
     - name: Configure
       run: |
         cmake -G "${{matrix.gen}}" -A ${{matrix.arch}}  -T ClangCL -DSIMDJSON_DEVELOPER_MODE=ON -DSIMDJSON_COMPETITION=OFF -B build

.github/workflows/vs17-noexcept-ci.yml

@@ -7,7 +7,7 @@ jobs:
     name: windows-vs17
     runs-on: windows-latest
     steps:
-    - uses: actions/checkout@v3
+    - uses: actions/checkout@v4
     - uses: actions/cache@v3
       with:
         path: dependencies/.cache

CMakeLists.txt

@@ -3,7 +3,7 @@ cmake_minimum_required(VERSION 3.14)
 project(
     simdjson
     # The version number is modified by tools/release.py
-    VERSION 3.6.2
+    VERSION 3.6.4
     DESCRIPTION "Parsing gigabytes of JSON per second"
     HOMEPAGE_URL "https://simdjson.org/"
     LANGUAGES CXX C

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         = "3.6.2"
+PROJECT_NUMBER         = "3.6.4"
 
 # 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

README.md

@@ -29,6 +29,7 @@ Table of Contents
 * [Real-world usage](#real-world-usage)
 * [Quick Start](#quick-start)
 * [Documentation](#documentation)
+* [Godbolt](#godbolt)
 * [Performance results](#performance-results)
 * [Bindings and Ports of simdjson](#bindings-and-ports-of-simdjson)
 * [About simdjson](#about-simdjson)
@@ -68,9 +69,9 @@ The simdjson library is easily consumable with a single .h and .cpp file.
 
 0. Prerequisites: `g++` (version 7 or better) or `clang++` (version 6 or better), and a 64-bit
    system with a command-line shell (e.g., Linux, macOS, freeBSD). We also support programming
-   environments like Visual Studio and Xcode, but different steps are needed.
+   environments like Visual Studio and Xcode, but different steps are needed. Users of clang++ may need to specify the C++ version (e.g., `c++ -std=c++17`) since clang++ tends to default on C++98.
 1. Pull [simdjson.h](singleheader/simdjson.h) and [simdjson.cpp](singleheader/simdjson.cpp) into a
-   directory, along with the sample file [twitter.json](jsonexamples/twitter.json).
+   directory, along with the sample file [twitter.json](jsonexamples/twitter.json). You can download them with the `wget` utility:
 
    ```
    wget https://raw.githubusercontent.com/simdjson/simdjson/master/singleheader/simdjson.h https://raw.githubusercontent.com/simdjson/simdjson/master/singleheader/simdjson.cpp https://raw.githubusercontent.com/simdjson/simdjson/master/jsonexamples/twitter.json
@@ -95,6 +96,7 @@ int main(void) {
    100 results.
   ```
 
+
 Documentation
 -------------
 
@@ -106,11 +108,12 @@ Usage documentation is available:
   how you can work with it.
 * [API](https://simdjson.github.io/simdjson/) contains the automatically generated API documentation.
 
+Godbolt
+-------------
 
 Some users may want to browse code along with the compiled assembly. You want to check out the following lists of examples:
-
-* [simdjson examples with errors handled through exceptions](https://godbolt.org/z/98Kx9Kqjn)
-* [simdjson examples with errors without exceptions](https://godbolt.org/z/PKG7GdbPo)
+* [simdjson examples with errors handled through exceptions](https://godbolt.org/z/7G5qE4sr9)
+* [simdjson examples with errors without exceptions](https://godbolt.org/z/e9dWb9E4v)
 
 Performance results
 -------------------
@@ -174,7 +177,11 @@ The simdjson library takes advantage of modern microarchitectures, parallelizing
 instructions, reducing branch misprediction, and reducing data dependency to take advantage of each
 CPU's multiple execution cores.
 
-Some people [enjoy reading our paper](https://arxiv.org/abs/1902.08318): A description of the design
+Our default front-end is called On Demand, and we wrote a paper about it:
+
+- John Keiser, Daniel Lemire, [On-Demand JSON: A Better Way to Parse Documents?](http://arxiv.org/abs/2312.17149), Software: Practice and Experience (to appear)
+
+Some people [enjoy reading the first (2019) simdjson paper](https://arxiv.org/abs/1902.08318): A description of the design
 and implementation of simdjson is in our research article:
 - Geoff Langdale, Daniel Lemire, [Parsing Gigabytes of JSON per Second](https://arxiv.org/abs/1902.08318), VLDB Journal 28 (6), 2019.
 

benchmark/CMakeLists.txt

@@ -26,6 +26,9 @@ if (TARGET benchmark::benchmark)
     if(TARGET nlohmann_json)
       target_link_libraries(bench_ondemand PRIVATE nlohmann_json)
     endif()
+    if(TARGET boostjson)
+      target_link_libraries(bench_ondemand PRIVATE boostjson)
+    endif()
   endif()
 endif()
 

benchmark/bench_ondemand.cpp

@@ -21,6 +21,10 @@ SIMDJSON_PUSH_DISABLE_ALL_WARNINGS
 #include <nlohmann/json.hpp>
 #endif
 
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+#include <boost/json.hpp>
+#endif
+
 // This has to be last, for reasons I don't yet understand
 #include <benchmark/benchmark.h>
 
@@ -33,6 +37,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #include "json2msgpack/sajson.h"
 #endif // SIMDJSON_COMPETITION_ONDEMAND_SAJSON
 #include "json2msgpack/nlohmann_json.h"
+#include "json2msgpack/boostjson.h"
 
 #include "partial_tweets/simdjson_ondemand.h"
 #include "partial_tweets/simdjson_dom.h"
@@ -48,6 +53,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #if SIMDJSON_COMPETITION_SAX
 #include "partial_tweets/nlohmann_json_sax.h"
 #endif // SIMDJSON_COMPETITION_SAX
+#include "partial_tweets/boostjson.h"
 
 
 #include "distinct_user_id/simdjson_ondemand.h"
@@ -66,6 +72,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #if SIMDJSON_COMPETITION_SAX
 #include "distinct_user_id/nlohmann_json_sax.h"
 #endif // SIMDJSON_COMPETITION_SAX
+#include "distinct_user_id/boostjson.h"
 
 #include "find_tweet/simdjson_ondemand.h"
 #include "find_tweet/simdjson_dom.h"
@@ -81,6 +88,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #if SIMDJSON_COMPETITION_SAX
 #include "find_tweet/nlohmann_json_sax.h"
 #endif // SIMDJSON_COMPETITION_SAX
+#include "find_tweet/boostjson.h"
 
 #include "top_tweet/simdjson_ondemand.h"
 #include "top_tweet/simdjson_dom.h"
@@ -96,6 +104,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #if SIMDJSON_COMPETITION_SAX
 #include "top_tweet/nlohmann_json_sax.h"
 #endif // SIMDJSON_COMPETITION_SAX
+#include "top_tweet/boostjson.h"
 
 
 #include "kostya/simdjson_ondemand.h"
@@ -112,6 +121,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #if SIMDJSON_COMPETITION_SAX
 #include "kostya/nlohmann_json_sax.h"
 #endif // SIMDJSON_COMPETITION_SAX
+#include "kostya/boostjson.h"
 
 #include "large_random/simdjson_ondemand.h"
 #if SIMDJSON_COMPETITION_ONDEMAND_UNORDERED
@@ -130,6 +140,7 @@ SIMDJSON_POP_DISABLE_WARNINGS
 #if SIMDJSON_COMPETITION_SAX
 #include "large_random/nlohmann_json_sax.h"
 #endif // SIMDJSON_COMPETITION_SAX
+#include "large_random/boostjson.h"
 
 #include "amazon_cellphones/simdjson_dom.h"
 #include "amazon_cellphones/simdjson_ondemand.h"

benchmark/distinct_user_id/boostjson.h

@@ -0,0 +1,29 @@
+#pragma once
+
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "distinct_user_id.h"
+
+namespace distinct_user_id {
+
+struct boostjson {
+  bool run(simdjson::padded_string &json, std::vector<uint64_t> &result) {
+
+    auto root = boost::json::parse(json);
+    for (const auto &tweet : root.at("statuses").as_array()) {
+      result.push_back(tweet.at("user").at("id").to_number<uint64_t>());
+
+      if (tweet.as_object().if_contains("retweeted_status")) {
+        result.push_back(tweet.at("retweeted_status").at("user").at("id").to_number<uint64_t>());
+      }
+    }
+
+    return true;
+  }
+};
+
+BENCHMARK_TEMPLATE(distinct_user_id, boostjson)->UseManualTime();
+
+} // namespace distinct_user_id
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

benchmark/find_tweet/boostjson.h

@@ -0,0 +1,30 @@
+#pragma once
+
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "find_tweet.h"
+
+namespace find_tweet {
+
+struct boostjson {
+  using StringType=std::string;
+
+  bool run(simdjson::padded_string &json, uint64_t find_id, std::string &result) {
+
+    auto root = boost::json::parse(json);
+    for (const auto &tweet : root.at("statuses").as_array()) {
+      if (tweet.at("id") == find_id) {
+        result = tweet.at("text").as_string();
+        return true;
+      }
+    }
+
+    return false;
+  }
+};
+
+BENCHMARK_TEMPLATE(find_tweet, boostjson)->UseManualTime();
+
+} // namespace find_tweet
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

benchmark/json2msgpack/boostjson.h

@@ -0,0 +1,104 @@
+#pragma once
+
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "json2msgpack.h"
+
+namespace json2msgpack {
+
+struct boostjson2msgpack {
+  inline std::string_view to_msgpack(const boost::json::value &root, uint8_t *buf) {
+    buff = buf;
+    recursive_processor(root);
+    return std::string_view(reinterpret_cast<char *>(buf), size_t(buff - buf));
+  }
+
+private:
+  uint8_t *buff{};
+
+  inline void write_double(const double d) noexcept {
+    *buff++ = 0xcb;
+    ::memcpy(buff, &d, sizeof(d));
+    buff += sizeof(d);
+  }
+
+  inline void write_byte(const uint8_t b) noexcept {
+    *buff = b;
+    buff++;
+  }
+
+  inline void write_uint32(const uint32_t w) noexcept {
+    ::memcpy(buff, &w, sizeof(w));
+    buff += sizeof(w);
+  }
+
+  inline void write_string(const std::string & str) {
+    write_byte(0xdb);
+    write_uint32(uint32_t(str.size()));
+    ::memcpy(buff, str.data(), str.size());
+    buff += str.size();
+  }
+
+  inline void recursive_processor(const boost::json::value &element) {
+    switch(element.kind()) {
+    case boost::json::kind::array: {
+      write_byte(0xdd);
+      const auto &array = element.as_array();
+      write_uint32(static_cast<uint32_t>(array.size()));
+      for (const auto &child : array) {
+        recursive_processor(child);
+      }
+    } break;
+
+    case boost::json::kind::object: {
+      write_byte(0xdf);
+      const auto &object = element.as_object();
+      write_uint32(static_cast<uint32_t>(object.size()));
+      for (const auto &child : object) {
+        write_string(child.key_c_str());
+        recursive_processor(child.value());
+      }
+    } break;
+
+    case boost::json::kind::int64:
+    case boost::json::kind::uint64:
+    case boost::json::kind::double_:
+      write_double(element.to_number<double>());
+      break;
+
+    case boost::json::kind::string:
+      write_string(element.as_string().c_str());
+      break;
+
+    case boost::json::kind::bool_:
+      write_byte(0xc2 + element.as_bool());
+      break;
+
+    case boost::json::kind::null:
+      write_byte(0xc0);
+      break;
+
+    default:
+      printf("unexpected\n");
+      break;
+    }
+  }
+};
+
+struct boostjson {
+  using StringType=std::string;
+
+  boostjson2msgpack parser{};
+
+  bool run(simdjson::padded_string &json, char *buffer, std::string_view &result) {
+    auto root = boost::json::parse(json);
+    result = parser.to_msgpack(root, reinterpret_cast<uint8_t *>(buffer));
+    return true;
+  }
+};
+
+BENCHMARK_TEMPLATE(json2msgpack, boostjson)->UseManualTime();
+
+} // namespace json2msgpack
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

benchmark/kostya/boostjson.h

@@ -0,0 +1,29 @@
+#pragma once
+
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "kostya.h"
+
+namespace kostya {
+
+struct boostjson {
+  static constexpr diff_flags DiffFlags = diff_flags::IMPRECISE_FLOATS;
+
+  bool run(simdjson::padded_string &json, std::vector<point> &result) {
+    auto root = boost::json::parse(json);
+    for (const auto &point : root.at("coordinates").as_array()) {
+      result.emplace_back(json_benchmark::point{
+        point.at("x").to_number<double>(),
+        point.at("y").to_number<double>(),
+        point.at("z").to_number<double>()
+      });
+    }
+    return true;
+  }
+};
+
+BENCHMARK_TEMPLATE(kostya, boostjson)->UseManualTime();
+
+} // namespace kostya
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

benchmark/large_random/boostjson.h

@@ -0,0 +1,29 @@
+#pragma once
+
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "large_random.h"
+
+namespace large_random {
+
+struct boostjson {
+  static constexpr diff_flags DiffFlags = diff_flags::IMPRECISE_FLOATS;
+
+  bool run(simdjson::padded_string &json, std::vector<point> &result) {
+    auto root = boost::json::parse(json);
+    for (const auto &point : root.as_array()) {
+      result.emplace_back(json_benchmark::point{
+        point.at("x").to_number<double>(),
+        point.at("y").to_number<double>(),
+        point.at("z").to_number<double>()
+      });
+    }
+    return true;
+  }
+};
+
+BENCHMARK_TEMPLATE(large_random, boostjson)->UseManualTime();
+
+} // namespace large_random
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

benchmark/partial_tweets/boostjson.h

@@ -0,0 +1,43 @@
+#pragma once
+
+#ifdef SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "partial_tweets.h"
+
+namespace partial_tweets {
+
+struct boostjson {
+  using StringType=std::string;
+
+  bool run(simdjson::padded_string &json, std::vector<tweet<StringType>> &result) {
+
+    auto root = boost::json::parse(json);
+    for (const auto &tweet : root.at("statuses").as_array()) {
+      const auto &user = tweet.at("user");
+
+      auto in_reply_to_status_id = tweet.as_object().if_contains("in_reply_to_status_id")
+        ? tweet.at("in_reply_to_status_id") : boost::json::value();
+
+      result.emplace_back(partial_tweets::tweet<StringType>{
+        tweet.at("created_at").as_string().c_str(),
+        tweet.at("id").to_number<uint64_t>(),
+        tweet.at("text").as_string().c_str(),
+        in_reply_to_status_id.is_null() ? 0 : in_reply_to_status_id.to_number<uint64_t>(),
+        {
+          user.at("id").to_number<uint64_t>(),
+          user.at("screen_name").as_string().c_str()
+        },
+        tweet.at("retweet_count").to_number<uint64_t>(),
+        tweet.at("favorite_count").to_number<uint64_t>()
+      });
+    }
+
+    return true;
+  }
+};
+
+BENCHMARK_TEMPLATE(partial_tweets, boostjson)->UseManualTime();
+
+} // namespace partial_tweets
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

benchmark/top_tweet/boostjson.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#if SIMDJSON_COMPETITION_BOOSTJSON
+
+#include "top_tweet.h"
+
+namespace top_tweet {
+
+using namespace simdjson;
+
+struct boostjson {
+  using StringType=std::string;
+
+  bool run(simdjson::padded_string &json, int64_t max_retweet_count, top_tweet_result<StringType> &result) {
+    result.retweet_count = -1;
+    boost::json::value top_tweet{};
+
+    auto root = boost::json::parse(json);
+    for (const auto &tweet : root.at("statuses").as_array()) {
+      int64_t retweet_count = tweet.at("retweet_count").as_int64();
+      if (retweet_count <= max_retweet_count && retweet_count >= result.retweet_count) {
+        result.retweet_count = retweet_count;
+        top_tweet = tweet;
+      }
+    }
+
+    result.text = top_tweet.at("text").as_string();
+    result.screen_name = top_tweet.at("user").at("screen_name").as_string();
+    return result.retweet_count != -1;
+  }
+};
+
+BENCHMARK_TEMPLATE(top_tweet, boostjson)->UseManualTime();
+
+} // namespace top_tweet
+
+#endif // SIMDJSON_COMPETITION_BOOSTJSON

doc/basics.md

@@ -36,6 +36,7 @@ An overview of what you need to know to use simdjson, with examples.
   - [Backwards Compatibility](#backwards-compatibility)
   - [Examples](#examples)
   - [Performance Tips](#performance-tips)
+  - [Further Reading](#further-reading)
 
 
 Requirements
@@ -1656,8 +1657,8 @@ JSON string to a user-provided buffer:
 General Direct Access to the Raw JSON String
 --------------------------------
 If your value is a string, the `raw_json_string` you with `get_raw_json_string()` gives you direct access to the unprocessed
-string. The simdjson library allows you to have access to the raw underlying JSON
-more generally.
+string. But the simdjson library allows you to have access to the raw underlying JSON
+more generally, not just for strings.
 
 The simdjson library makes explicit assumptions about types. For examples, numbers
 must be integers (up to 64-bit integers) or binary64 floating-point numbers. Some users
@@ -2127,3 +2128,9 @@ Performance Tips
 	std::string_view rating = data["rating"];
   ```
 - To better understand the operation of your On Demand parser, and whether it is performing as well as you think it should be, there is a  logger feature built in to simdjson! To use it, define the pre-processor directive `SIMDJSON_VERBOSE_LOGGING` prior to including the `simdjson.h` header, which enables logging in simdjson. Run your code. It may generate a lot of logging output; adding printouts from your application that show each section may be helpful. The log's output will show step-by-step information on state, buffer pointer position, depth, and key retrieval status. Importantly, unless `SIMDJSON_VERBOSE_LOGGING` is defined, logging is entirely disabled and thus carries no overhead.
+
+Further reading
+--------
+
+
+- John Keiser, Daniel Lemire, [On-Demand JSON: A Better Way to Parse Documents?](http://arxiv.org/abs/2312.17149), Software: Practice and Experience (to appear)

include/simdjson/dom/element.h

@@ -214,7 +214,7 @@ public:
     static_assert(!sizeof(T), "The get method with given type is not implemented by the simdjson library. "
       "The supported types are Boolean (bool), numbers (double, uint64_t, int64_t), "
       "strings (std::string_view, const char *), arrays (dom::array) and objects (dom::object). "
-      "We recommand you use get_double(), get_bool(), get_uint64(), get_int64(), "
+      "We recommend you use get_double(), get_bool(), get_uint64(), get_int64(), "
       "get_object(), get_array() or get_string() instead of the get template.");
   }
 

include/simdjson/dom/parser-inl.h

@@ -92,10 +92,14 @@ inline simdjson_result<size_t> parser::read_file(const std::string &path) noexce
 }
 
 inline simdjson_result<element> parser::load(const std::string &path) & noexcept {
+  return load_into_document(doc, path);
+}
+
+inline simdjson_result<element> parser::load_into_document(document& provided_doc, const std::string &path) & noexcept {
   size_t len;
   auto _error = read_file(path).get(len);
   if (_error) { return _error; }
-  return parse(loaded_bytes.get(), len, false);
+  return parse_into_document(provided_doc, loaded_bytes.get(), len, false);
 }
 
 inline simdjson_result<document_stream> parser::load_many(const std::string &path, size_t batch_size) noexcept {

include/simdjson/dom/parser.h

@@ -95,6 +95,53 @@ public:
    */
   inline simdjson_result<element> load(const std::string &path) & noexcept;
   inline simdjson_result<element> load(const std::string &path) &&  = delete ;
+
+  /**
+   * Load a JSON document from a file into a provide document instance and return a temporary reference to it.
+   * It is similar to the function `load` except that instead of parsing into the internal
+   * `document` instance associated with the parser, it allows the user to provide a document
+   * instance.
+   *
+   *   dom::parser parser;
+   *   dom::document doc;
+   *   element doc_root = parser.load_into_document(doc, "jsonexamples/twitter.json");
+   *
+   * The function is eager: the file's content is loaded in memory inside the parser instance
+   * and immediately parsed. The file can be deleted after the `parser.load_into_document` call.
+   *
+   * ### IMPORTANT: Document Lifetime
+   *
+   * After the call to load_into_document, the parser is no longer needed.
+   *
+   * The JSON document lives in the document instance: you must keep the document
+   * instance alive while you navigate through it (i.e., used the returned value from
+   * load_into_document). You are encourage to reuse the document instance
+   * many times with new data to avoid reallocations:
+   *
+   *   dom::document doc;
+   *   element doc_root1 = parser.load_into_document(doc, "jsonexamples/twitter.json");
+   *   //... doc_root1 is a pointer inside doc
+   *   element doc_root2 = parser.load_into_document(doc, "jsonexamples/twitter.json");
+   *   //... doc_root2 is a pointer inside doc
+   *   // at this point doc_root1 is no longer safe
+   *
+   * Moving the document instance is safe, but it invalidates the element instances. After
+   * moving a document, you can recover safe access to the document root with its `root()` method.
+   *
+   * @param doc The document instance where the parsed data will be stored (on success).
+   * @param path The path to load.
+   * @return The document, or an error:
+   *         - IO_ERROR if there was an error opening or reading the file.
+   *           Be mindful that on some 32-bit systems,
+   *           the file size might be limited to 2 GB.
+   *         - MEMALLOC if the parser does not have enough capacity and memory allocation fails.
+   *         - CAPACITY if the parser does not have enough capacity and len > max_capacity.
+   *         - other json errors if parsing fails. You should not rely on these errors to always the same for the
+   *           same document: they may vary under runtime dispatch (so they may vary depending on your system and hardware).
+   */
+  inline simdjson_result<element> load_into_document(document& doc, const std::string &path) & noexcept;
+  inline simdjson_result<element> load_into_document(document& doc, const std::string &path) && =delete;
+
   /**
    * Parse a JSON document and return a temporary reference to it.
    *

include/simdjson/generic/numberparsing.h

@@ -446,11 +446,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -474,13 +471,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -1004,7 +1001,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -1203,7 +1200,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;

include/simdjson/generic/ondemand/json_type.h

@@ -92,8 +92,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);

include/simdjson/nonstd/string_view.hpp

@@ -6,7 +6,7 @@
 // Distributed under the Boost Software License, Version 1.0.
 // (See accompanying file LICENSE.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
 
-#pragma once
+// #pragma once // We remove #pragma once here as it generates a warning in some cases. We rely on the include guard.
 
 #ifndef NONSTD_SV_LITE_H_INCLUDED
 #define NONSTD_SV_LITE_H_INCLUDED
@@ -69,6 +69,10 @@
 # define nssv_CONFIG_NO_STREAM_INSERTION  0
 #endif
 
+#ifndef  nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+# define nssv_CONFIG_CONSTEXPR11_STD_SEARCH  1
+#endif
+
 // Control presence of exception handling (try and auto discover):
 
 #ifndef nssv_CONFIG_NO_EXCEPTIONS
@@ -267,7 +271,7 @@ using std::operator<<;
 # define nssv_HAS_CPP0X  0
 #endif
 
-// Unless defined otherwise below, consider VC14 as C++11 for variant-lite:
+// Unless defined otherwise below, consider VC14 as C++11 for string-view-lite:
 
 #if nssv_COMPILER_MSVC_VER >= 1900
 # undef  nssv_CPP11_OR_GREATER
@@ -434,9 +438,9 @@ using std::operator<<;
 # pragma clang diagnostic ignored "-Wreserved-user-defined-literal"
 # pragma clang diagnostic push
 # pragma clang diagnostic ignored "-Wuser-defined-literals"
-#elif defined(__GNUC__)
-# pragma  GCC  diagnostic push
-# pragma  GCC  diagnostic ignored "-Wliteral-suffix"
+#elif nssv_COMPILER_GNUC_VERSION >= 480
+#  pragma  GCC  diagnostic push
+#  pragma  GCC  diagnostic ignored "-Wliteral-suffix"
 #endif // __clang__
 
 #if nssv_COMPILER_MSVC_VERSION >= 140
@@ -451,8 +455,8 @@ using std::operator<<;
 
 #if defined(__clang__)
 # define nssv_RESTORE_WARNINGS()  _Pragma("clang diagnostic pop")
-#elif defined(__GNUC__)
-# define nssv_RESTORE_WARNINGS()  _Pragma("GCC diagnostic pop")
+#elif nssv_COMPILER_GNUC_VERSION >= 480
+#  define nssv_RESTORE_WARNINGS()  _Pragma("GCC diagnostic pop")
 #elif nssv_COMPILER_MSVC_VERSION >= 140
 # define nssv_RESTORE_WARNINGS()  __pragma(warning(pop ))
 #else
@@ -567,12 +571,31 @@ constexpr const CharT* search( basic_string_view<CharT, Traits> haystack, basic_
 
 // non-recursive:
 
+#if nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
 template< class CharT, class Traits = std::char_traits<CharT> >
 constexpr const CharT* search( basic_string_view<CharT, Traits> haystack, basic_string_view<CharT, Traits> needle )
 {
     return std::search( haystack.begin(), haystack.end(), needle.begin(), needle.end() );
 }
 
+#else // nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
+template< class CharT, class Traits = std::char_traits<CharT> >
+nssv_constexpr14 const CharT* search( basic_string_view<CharT, Traits> haystack, basic_string_view<CharT, Traits> needle )
+{
+    while ( needle.size() <= haystack.size() )
+    {
+        if  ( haystack.starts_with(needle) )
+        {
+            return haystack.cbegin();
+        }
+        haystack = basic_string_view<CharT, Traits>{ haystack.begin() + 1, haystack.size() - 1U };
+    }
+    return haystack.cend();
+}
+#endif // nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
 #endif // OPTIMIZE
 #endif // nssv_CPP11_OR_GREATER && ! nssv_CPP17_OR_GREATER
 
@@ -845,7 +868,7 @@ public:
 
     // find(), 4x:
 
-    nssv_constexpr size_type find( basic_string_view v, size_type pos = 0 ) const nssv_noexcept  // (1)
+    nssv_constexpr14 size_type find( basic_string_view v, size_type pos = 0 ) const nssv_noexcept  // (1)
     {
         return assert( v.size() == 0 || v.data() != nssv_nullptr )
             , pos >= size()

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 "3.6.2"
+#define SIMDJSON_VERSION "3.6.4"
 
 namespace simdjson {
 enum {
@@ -19,7 +19,7 @@ enum {
   /**
    * The revision (major.minor.REVISION) of simdjson being used.
    */
-  SIMDJSON_VERSION_REVISION = 2
+  SIMDJSON_VERSION_REVISION = 4
 };
 } // namespace simdjson
 

singleheader/simdjson.cpp

@@ -1,4 +1,4 @@
-/* auto-generated on 2023-12-01 13:55:28 -0500. Do not edit! */
+/* auto-generated on 2024-01-28 12:42:59 -0500. Do not edit! */
 /* including simdjson.cpp:  */
 /* begin file simdjson.cpp */
 #define SIMDJSON_SRC_SIMDJSON_CPP
@@ -525,7 +525,7 @@ SIMDJSON_PUSH_DISABLE_ALL_WARNINGS
 // Distributed under the Boost Software License, Version 1.0.
 // (See accompanying file LICENSE.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
 
-#pragma once
+// #pragma once // We remove #pragma once here as it generates a warning in some cases. We rely on the include guard.
 
 #ifndef NONSTD_SV_LITE_H_INCLUDED
 #define NONSTD_SV_LITE_H_INCLUDED
@@ -588,6 +588,10 @@ SIMDJSON_PUSH_DISABLE_ALL_WARNINGS
 # define nssv_CONFIG_NO_STREAM_INSERTION  0
 #endif
 
+#ifndef  nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+# define nssv_CONFIG_CONSTEXPR11_STD_SEARCH  1
+#endif
+
 // Control presence of exception handling (try and auto discover):
 
 #ifndef nssv_CONFIG_NO_EXCEPTIONS
@@ -786,7 +790,7 @@ using std::operator<<;
 # define nssv_HAS_CPP0X  0
 #endif
 
-// Unless defined otherwise below, consider VC14 as C++11 for variant-lite:
+// Unless defined otherwise below, consider VC14 as C++11 for string-view-lite:
 
 #if nssv_COMPILER_MSVC_VER >= 1900
 # undef  nssv_CPP11_OR_GREATER
@@ -953,9 +957,9 @@ using std::operator<<;
 # pragma clang diagnostic ignored "-Wreserved-user-defined-literal"
 # pragma clang diagnostic push
 # pragma clang diagnostic ignored "-Wuser-defined-literals"
-#elif defined(__GNUC__)
-# pragma  GCC  diagnostic push
-# pragma  GCC  diagnostic ignored "-Wliteral-suffix"
+#elif nssv_COMPILER_GNUC_VERSION >= 480
+#  pragma  GCC  diagnostic push
+#  pragma  GCC  diagnostic ignored "-Wliteral-suffix"
 #endif // __clang__
 
 #if nssv_COMPILER_MSVC_VERSION >= 140
@@ -970,8 +974,8 @@ using std::operator<<;
 
 #if defined(__clang__)
 # define nssv_RESTORE_WARNINGS()  _Pragma("clang diagnostic pop")
-#elif defined(__GNUC__)
-# define nssv_RESTORE_WARNINGS()  _Pragma("GCC diagnostic pop")
+#elif nssv_COMPILER_GNUC_VERSION >= 480
+#  define nssv_RESTORE_WARNINGS()  _Pragma("GCC diagnostic pop")
 #elif nssv_COMPILER_MSVC_VERSION >= 140
 # define nssv_RESTORE_WARNINGS()  __pragma(warning(pop ))
 #else
@@ -1086,12 +1090,31 @@ constexpr const CharT* search( basic_string_view<CharT, Traits> haystack, basic_
 
 // non-recursive:
 
+#if nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
 template< class CharT, class Traits = std::char_traits<CharT> >
 constexpr const CharT* search( basic_string_view<CharT, Traits> haystack, basic_string_view<CharT, Traits> needle )
 {
     return std::search( haystack.begin(), haystack.end(), needle.begin(), needle.end() );
 }
 
+#else // nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
+template< class CharT, class Traits = std::char_traits<CharT> >
+nssv_constexpr14 const CharT* search( basic_string_view<CharT, Traits> haystack, basic_string_view<CharT, Traits> needle )
+{
+    while ( needle.size() <= haystack.size() )
+    {
+        if  ( haystack.starts_with(needle) )
+        {
+            return haystack.cbegin();
+        }
+        haystack = basic_string_view<CharT, Traits>{ haystack.begin() + 1, haystack.size() - 1U };
+    }
+    return haystack.cend();
+}
+#endif // nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
 #endif // OPTIMIZE
 #endif // nssv_CPP11_OR_GREATER && ! nssv_CPP17_OR_GREATER
 
@@ -1364,7 +1387,7 @@ public:
 
     // find(), 4x:
 
-    nssv_constexpr size_type find( basic_string_view v, size_type pos = 0 ) const nssv_noexcept  // (1)
+    nssv_constexpr14 size_type find( basic_string_view v, size_type pos = 0 ) const nssv_noexcept  // (1)
     {
         return assert( v.size() == 0 || v.data() != nssv_nullptr )
             , pos >= size()
@@ -9025,11 +9048,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -9053,13 +9073,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -9583,7 +9603,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -9782,7 +9802,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -10852,7 +10872,7 @@ namespace simdjson {
 namespace arm64 {
 namespace {
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
 simdjson_inline bool is_ascii(const simd8x64<uint8_t>& input);
 
 } // unnamed namespace
@@ -11419,7 +11439,7 @@ using namespace simd;
       const simd8<uint8_t> prev_input, const simd8<uint8_t> sc) {
     simd8<uint8_t> prev2 = input.prev<2>(prev_input);
     simd8<uint8_t> prev3 = input.prev<3>(prev_input);
-    simd8<uint8_t> must23 = simd8<uint8_t>(must_be_2_3_continuation(prev2, prev3));
+    simd8<uint8_t> must23 = must_be_2_3_continuation(prev2, prev3);
     simd8<uint8_t> must23_80 = must23 & uint8_t(0x80);
     return must23_80 ^ sc;
   }
@@ -13614,10 +13634,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
     return is_second_byte ^ is_third_byte ^ is_fourth_byte;
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-    simd8<bool> is_third_byte  = prev2 >= uint8_t(0xe0u);
-    simd8<bool> is_fourth_byte = prev3 >= uint8_t(0xf0u);
-    return is_third_byte ^ is_fourth_byte;
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+    simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+    simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+    return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace
@@ -14823,11 +14843,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -14851,13 +14868,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -15381,7 +15398,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -15580,7 +15597,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -19226,11 +19243,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -19254,13 +19268,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -19784,7 +19798,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -19983,7 +19997,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -20930,7 +20944,7 @@ namespace simdjson {
 namespace haswell {
 namespace {
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
 simdjson_inline bool is_ascii(const simd8x64<uint8_t>& input);
 
 } // unnamed namespace
@@ -21497,7 +21511,7 @@ using namespace simd;
       const simd8<uint8_t> prev_input, const simd8<uint8_t> sc) {
     simd8<uint8_t> prev2 = input.prev<2>(prev_input);
     simd8<uint8_t> prev3 = input.prev<3>(prev_input);
-    simd8<uint8_t> must23 = simd8<uint8_t>(must_be_2_3_continuation(prev2, prev3));
+    simd8<uint8_t> must23 = must_be_2_3_continuation(prev2, prev3);
     simd8<uint8_t> must23_80 = must23 & uint8_t(0x80);
     return must23_80 ^ sc;
   }
@@ -23689,11 +23703,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace
@@ -25390,11 +25403,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -25418,13 +25428,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -25948,7 +25958,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -26147,7 +26157,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -27092,7 +27102,7 @@ namespace simdjson {
 namespace icelake {
 namespace {
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
 simdjson_inline bool is_ascii(const simd8x64<uint8_t>& input);
 
 } // unnamed namespace
@@ -27659,7 +27669,7 @@ using namespace simd;
       const simd8<uint8_t> prev_input, const simd8<uint8_t> sc) {
     simd8<uint8_t> prev2 = input.prev<2>(prev_input);
     simd8<uint8_t> prev3 = input.prev<3>(prev_input);
-    simd8<uint8_t> must23 = simd8<uint8_t>(must_be_2_3_continuation(prev2, prev3));
+    simd8<uint8_t> must23 = must_be_2_3_continuation(prev2, prev3);
     simd8<uint8_t> must23_80 = must23 & uint8_t(0x80);
     return must23_80 ^ sc;
   }
@@ -29850,11 +29860,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace
@@ -31710,11 +31719,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -31738,13 +31744,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -32268,7 +32274,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -32467,7 +32473,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -33525,7 +33531,7 @@ namespace simdjson {
 namespace ppc64 {
 namespace {
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
 simdjson_inline bool is_ascii(const simd8x64<uint8_t>& input);
 
 } // unnamed namespace
@@ -34092,7 +34098,7 @@ using namespace simd;
       const simd8<uint8_t> prev_input, const simd8<uint8_t> sc) {
     simd8<uint8_t> prev2 = input.prev<2>(prev_input);
     simd8<uint8_t> prev3 = input.prev<3>(prev_input);
-    simd8<uint8_t> must23 = simd8<uint8_t>(must_be_2_3_continuation(prev2, prev3));
+    simd8<uint8_t> must23 = must_be_2_3_continuation(prev2, prev3);
     simd8<uint8_t> must23_80 = must23 & uint8_t(0x80);
     return must23_80 ^ sc;
   }
@@ -36257,11 +36263,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace
@@ -38396,11 +38401,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -38424,13 +38426,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -38954,7 +38956,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -39153,7 +39155,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -40532,7 +40534,7 @@ namespace simdjson {
 namespace westmere {
 namespace {
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
 simdjson_inline bool is_ascii(const simd8x64<uint8_t>& input);
 
 } // unnamed namespace
@@ -41099,7 +41101,7 @@ using namespace simd;
       const simd8<uint8_t> prev_input, const simd8<uint8_t> sc) {
     simd8<uint8_t> prev2 = input.prev<2>(prev_input);
     simd8<uint8_t> prev3 = input.prev<3>(prev_input);
-    simd8<uint8_t> must23 = simd8<uint8_t>(must_be_2_3_continuation(prev2, prev3));
+    simd8<uint8_t> must23 = must_be_2_3_continuation(prev2, prev3);
     simd8<uint8_t> must23_80 = must23 & uint8_t(0x80);
     return must23_80 ^ sc;
   }
@@ -43295,11 +43297,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace

singleheader/simdjson.h

@@ -1,4 +1,4 @@
-/* auto-generated on 2023-12-01 13:55:28 -0500. Do not edit! */
+/* auto-generated on 2024-01-28 12:42:59 -0500. Do not edit! */
 /* including simdjson.h:  */
 /* begin file simdjson.h */
 #ifndef SIMDJSON_H
@@ -545,7 +545,7 @@ SIMDJSON_PUSH_DISABLE_ALL_WARNINGS
 // Distributed under the Boost Software License, Version 1.0.
 // (See accompanying file LICENSE.txt or copy at http://www.boost.org/LICENSE_1_0.txt)
 
-#pragma once
+// #pragma once // We remove #pragma once here as it generates a warning in some cases. We rely on the include guard.
 
 #ifndef NONSTD_SV_LITE_H_INCLUDED
 #define NONSTD_SV_LITE_H_INCLUDED
@@ -608,6 +608,10 @@ SIMDJSON_PUSH_DISABLE_ALL_WARNINGS
 # define nssv_CONFIG_NO_STREAM_INSERTION  0
 #endif
 
+#ifndef  nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+# define nssv_CONFIG_CONSTEXPR11_STD_SEARCH  1
+#endif
+
 // Control presence of exception handling (try and auto discover):
 
 #ifndef nssv_CONFIG_NO_EXCEPTIONS
@@ -806,7 +810,7 @@ using std::operator<<;
 # define nssv_HAS_CPP0X  0
 #endif
 
-// Unless defined otherwise below, consider VC14 as C++11 for variant-lite:
+// Unless defined otherwise below, consider VC14 as C++11 for string-view-lite:
 
 #if nssv_COMPILER_MSVC_VER >= 1900
 # undef  nssv_CPP11_OR_GREATER
@@ -973,9 +977,9 @@ using std::operator<<;
 # pragma clang diagnostic ignored "-Wreserved-user-defined-literal"
 # pragma clang diagnostic push
 # pragma clang diagnostic ignored "-Wuser-defined-literals"
-#elif defined(__GNUC__)
-# pragma  GCC  diagnostic push
-# pragma  GCC  diagnostic ignored "-Wliteral-suffix"
+#elif nssv_COMPILER_GNUC_VERSION >= 480
+#  pragma  GCC  diagnostic push
+#  pragma  GCC  diagnostic ignored "-Wliteral-suffix"
 #endif // __clang__
 
 #if nssv_COMPILER_MSVC_VERSION >= 140
@@ -990,8 +994,8 @@ using std::operator<<;
 
 #if defined(__clang__)
 # define nssv_RESTORE_WARNINGS()  _Pragma("clang diagnostic pop")
-#elif defined(__GNUC__)
-# define nssv_RESTORE_WARNINGS()  _Pragma("GCC diagnostic pop")
+#elif nssv_COMPILER_GNUC_VERSION >= 480
+#  define nssv_RESTORE_WARNINGS()  _Pragma("GCC diagnostic pop")
 #elif nssv_COMPILER_MSVC_VERSION >= 140
 # define nssv_RESTORE_WARNINGS()  __pragma(warning(pop ))
 #else
@@ -1106,12 +1110,31 @@ constexpr const CharT* search( basic_string_view<CharT, Traits> haystack, basic_
 
 // non-recursive:
 
+#if nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
 template< class CharT, class Traits = std::char_traits<CharT> >
 constexpr const CharT* search( basic_string_view<CharT, Traits> haystack, basic_string_view<CharT, Traits> needle )
 {
     return std::search( haystack.begin(), haystack.end(), needle.begin(), needle.end() );
 }
 
+#else // nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
+template< class CharT, class Traits = std::char_traits<CharT> >
+nssv_constexpr14 const CharT* search( basic_string_view<CharT, Traits> haystack, basic_string_view<CharT, Traits> needle )
+{
+    while ( needle.size() <= haystack.size() )
+    {
+        if  ( haystack.starts_with(needle) )
+        {
+            return haystack.cbegin();
+        }
+        haystack = basic_string_view<CharT, Traits>{ haystack.begin() + 1, haystack.size() - 1U };
+    }
+    return haystack.cend();
+}
+#endif // nssv_CONFIG_CONSTEXPR11_STD_SEARCH
+
 #endif // OPTIMIZE
 #endif // nssv_CPP11_OR_GREATER && ! nssv_CPP17_OR_GREATER
 
@@ -1384,7 +1407,7 @@ public:
 
     // find(), 4x:
 
-    nssv_constexpr size_type find( basic_string_view v, size_type pos = 0 ) const nssv_noexcept  // (1)
+    nssv_constexpr14 size_type find( basic_string_view v, size_type pos = 0 ) const nssv_noexcept  // (1)
     {
         return assert( v.size() == 0 || v.data() != nssv_nullptr )
             , pos >= size()
@@ -2321,7 +2344,7 @@ namespace std {
 #define SIMDJSON_SIMDJSON_VERSION_H
 
 /** The version of simdjson being used (major.minor.revision) */
-#define SIMDJSON_VERSION "3.6.2"
+#define SIMDJSON_VERSION "3.6.4"
 
 namespace simdjson {
 enum {
@@ -2336,7 +2359,7 @@ enum {
   /**
    * The revision (major.minor.REVISION) of simdjson being used.
    */
-  SIMDJSON_VERSION_REVISION = 2
+  SIMDJSON_VERSION_REVISION = 4
 };
 } // namespace simdjson
 
@@ -4449,6 +4472,53 @@ public:
    */
   inline simdjson_result<element> load(const std::string &path) & noexcept;
   inline simdjson_result<element> load(const std::string &path) &&  = delete ;
+
+  /**
+   * Load a JSON document from a file into a provide document instance and return a temporary reference to it.
+   * It is similar to the function `load` except that instead of parsing into the internal
+   * `document` instance associated with the parser, it allows the user to provide a document
+   * instance.
+   *
+   *   dom::parser parser;
+   *   dom::document doc;
+   *   element doc_root = parser.load_into_document(doc, "jsonexamples/twitter.json");
+   *
+   * The function is eager: the file's content is loaded in memory inside the parser instance
+   * and immediately parsed. The file can be deleted after the `parser.load_into_document` call.
+   *
+   * ### IMPORTANT: Document Lifetime
+   *
+   * After the call to load_into_document, the parser is no longer needed.
+   *
+   * The JSON document lives in the document instance: you must keep the document
+   * instance alive while you navigate through it (i.e., used the returned value from
+   * load_into_document). You are encourage to reuse the document instance
+   * many times with new data to avoid reallocations:
+   *
+   *   dom::document doc;
+   *   element doc_root1 = parser.load_into_document(doc, "jsonexamples/twitter.json");
+   *   //... doc_root1 is a pointer inside doc
+   *   element doc_root2 = parser.load_into_document(doc, "jsonexamples/twitter.json");
+   *   //... doc_root2 is a pointer inside doc
+   *   // at this point doc_root1 is no longer safe
+   *
+   * Moving the document instance is safe, but it invalidates the element instances. After
+   * moving a document, you can recover safe access to the document root with its `root()` method.
+   *
+   * @param doc The document instance where the parsed data will be stored (on success).
+   * @param path The path to load.
+   * @return The document, or an error:
+   *         - IO_ERROR if there was an error opening or reading the file.
+   *           Be mindful that on some 32-bit systems,
+   *           the file size might be limited to 2 GB.
+   *         - MEMALLOC if the parser does not have enough capacity and memory allocation fails.
+   *         - CAPACITY if the parser does not have enough capacity and len > max_capacity.
+   *         - other json errors if parsing fails. You should not rely on these errors to always the same for the
+   *           same document: they may vary under runtime dispatch (so they may vary depending on your system and hardware).
+   */
+  inline simdjson_result<element> load_into_document(document& doc, const std::string &path) & noexcept;
+  inline simdjson_result<element> load_into_document(document& doc, const std::string &path) && =delete;
+
   /**
    * Parse a JSON document and return a temporary reference to it.
    *
@@ -5484,7 +5554,7 @@ public:
     static_assert(!sizeof(T), "The get method with given type is not implemented by the simdjson library. "
       "The supported types are Boolean (bool), numbers (double, uint64_t, int64_t), "
       "strings (std::string_view, const char *), arrays (dom::array) and objects (dom::object). "
-      "We recommand you use get_double(), get_bool(), get_uint64(), get_int64(), "
+      "We recommend you use get_double(), get_bool(), get_uint64(), get_int64(), "
       "get_object(), get_array() or get_string() instead of the get template.");
   }
 
@@ -7556,10 +7626,14 @@ inline simdjson_result<size_t> parser::read_file(const std::string &path) noexce
 }
 
 inline simdjson_result<element> parser::load(const std::string &path) & noexcept {
+  return load_into_document(doc, path);
+}
+
+inline simdjson_result<element> parser::load_into_document(document& provided_doc, const std::string &path) & noexcept {
   size_t len;
   auto _error = read_file(path).get(len);
   if (_error) { return _error; }
-  return parse(loaded_bytes.get(), len, false);
+  return parse_into_document(provided_doc, loaded_bytes.get(), len, false);
 }
 
 inline simdjson_result<document_stream> parser::load_many(const std::string &path, size_t batch_size) noexcept {
@@ -12109,11 +12183,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -12137,13 +12208,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -12667,7 +12738,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -12866,7 +12937,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -14161,11 +14232,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -14189,13 +14257,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -14719,7 +14787,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -14918,7 +14986,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -16705,11 +16773,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -16733,13 +16798,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -17263,7 +17328,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -17462,7 +17527,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -19248,11 +19313,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -19276,13 +19338,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -19806,7 +19868,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -20005,7 +20067,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -21906,11 +21968,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -21934,13 +21993,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -22464,7 +22523,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -22663,7 +22722,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -24887,11 +24946,8 @@ simdjson_inline size_t significant_digits(const uint8_t * start_digits, size_t d
 } // unnamed namespace
 
 /** @private */
-template<typename W>
-error_code slow_float_parsing(simdjson_unused const uint8_t * src, W writer) {
-  double d;
-  if (parse_float_fallback(src, &d)) {
-    writer.append_double(d);
+static error_code slow_float_parsing(simdjson_unused const uint8_t * src, double* answer) {
+  if (parse_float_fallback(src, answer)) {
     return SUCCESS;
   }
   return INVALID_NUMBER(src);
@@ -24915,13 +24971,13 @@ simdjson_inline error_code write_float(const uint8_t *const src, bool negative,
     // 10000000000000000000000000000000000000000000e+308
     // 3.1415926535897932384626433832795028841971693993751
     //
-    // NOTE: This makes a *copy* of the writer and passes it to slow_float_parsing. This happens
-    // because slow_float_parsing is a non-inlined function. If we passed our writer reference to
-    // it, it would force it to be stored in memory, preventing the compiler from picking it apart
-    // and putting into registers. i.e. if we pass it as reference, it gets slow.
-    // This is what forces the skip_double, as well.
-    error_code error = slow_float_parsing(src, writer);
-    writer.skip_double();
+    // NOTE: We do not pass a reference to the to slow_float_parsing. If we passed our writer
+    // reference to it, it would force it to be stored in memory, preventing the compiler from
+    // picking it apart and putting into registers. i.e. if we pass it as reference,
+    // it gets slow.
+    double d;
+    error_code error = slow_float_parsing(src, &d);
+    writer.append_double(d);
     return error;
   }
   // NOTE: it's weird that the simdjson_unlikely() only wraps half the if, but it seems to get slower any other
@@ -25445,7 +25501,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double(const uint8
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -25644,7 +25700,7 @@ simdjson_unused simdjson_inline simdjson_result<double> parse_double_in_string(c
       // Skip leading 0.00000 and see if it still overflows
       const uint8_t *start_digits = src + 2;
       while (*start_digits == '0') { start_digits++; }
-      overflow = start_digits-src > 19;
+      overflow = p-start_digits > 19;
     }
   } else {
     overflow = p-src > 19;
@@ -28613,8 +28669,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);
@@ -38534,8 +38588,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);
@@ -48947,8 +48999,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);
@@ -59359,8 +59409,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);
@@ -69886,8 +69934,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);
@@ -80736,8 +80782,6 @@ protected:
    */
   friend class value_iterator;
   template<typename W>
-  friend error_code numberparsing::slow_float_parsing(simdjson_unused const uint8_t * src, W writer);
-  template<typename W>
   friend error_code numberparsing::write_float(const uint8_t *const src, bool negative, uint64_t i, const uint8_t * start_digits, size_t digit_count, int64_t exponent, W &writer);
   template<typename W>
   friend error_code numberparsing::parse_number(const uint8_t *const src, W &writer);

src/arm64.cpp

@@ -108,10 +108,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
     return is_second_byte ^ is_third_byte ^ is_fourth_byte;
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-    simd8<bool> is_third_byte  = prev2 >= uint8_t(0xe0u);
-    simd8<bool> is_fourth_byte = prev3 >= uint8_t(0xf0u);
-    return is_third_byte ^ is_fourth_byte;
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+    simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+    simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+    return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace

src/generic/dom_parser_implementation.h

@@ -11,7 +11,7 @@ namespace simdjson {
 namespace SIMDJSON_IMPLEMENTATION {
 namespace {
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3);
 simdjson_inline bool is_ascii(const simd8x64<uint8_t>& input);
 
 } // unnamed namespace

src/generic/stage1/utf8_lookup4_algorithm.h

@@ -107,7 +107,7 @@ using namespace simd;
       const simd8<uint8_t> prev_input, const simd8<uint8_t> sc) {
     simd8<uint8_t> prev2 = input.prev<2>(prev_input);
     simd8<uint8_t> prev3 = input.prev<3>(prev_input);
-    simd8<uint8_t> must23 = simd8<uint8_t>(must_be_2_3_continuation(prev2, prev3));
+    simd8<uint8_t> must23 = must_be_2_3_continuation(prev2, prev3);
     simd8<uint8_t> must23_80 = must23 & uint8_t(0x80);
     return must23_80 ^ sc;
   }

src/haswell.cpp

@@ -105,11 +105,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace

src/icelake.cpp

@@ -107,11 +107,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace

src/ppc64.cpp

@@ -78,11 +78,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace

src/westmere.cpp

@@ -109,11 +109,10 @@ simdjson_unused simdjson_inline simd8<bool> must_be_continuation(const simd8<uin
   return simd8<int8_t>(is_second_byte | is_third_byte | is_fourth_byte) > int8_t(0);
 }
 
-simdjson_inline simd8<bool> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
-  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-1); // Only 111_____ will be > 0
-  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-1); // Only 1111____ will be > 0
-  // Caller requires a bool (all 1's). All values resulting from the subtraction will be <= 64, so signed comparison is fine.
-  return simd8<int8_t>(is_third_byte | is_fourth_byte) > int8_t(0);
+simdjson_inline simd8<uint8_t> must_be_2_3_continuation(const simd8<uint8_t> prev2, const simd8<uint8_t> prev3) {
+  simd8<uint8_t> is_third_byte  = prev2.saturating_sub(0xe0u-0x80); // Only 111_____ will be >= 0x80
+  simd8<uint8_t> is_fourth_byte = prev3.saturating_sub(0xf0u-0x80); // Only 1111____ will be >= 0x80
+  return is_third_byte | is_fourth_byte;
 }
 
 } // unnamed namespace

tests/dom/basictests.cpp

@@ -452,14 +452,18 @@ namespace parse_api_tests {
     element doc_root1 = parser.parse_into_document(doc, input);
     if(simdjson::to_string(doc_root1) != "[1,2,3]") { return false; }
     //... doc_root1 is a pointer inside doc
-    element doc_root2 = parser.parse_into_document(doc, input);
+    element doc_root2 = parser.load_into_document(doc, TWITTER_JSON);
     //... doc_root2 is a pointer inside doc
-    if(simdjson::to_string(doc_root2) != "[1,2,3]") { return false; }
+    if(uint64_t(doc_root2["search_metadata"]["count"]) != 100) { return false; }
+    if(uint64_t(doc.root()["search_metadata"]["count"]) != 100) { return false; }
+    element doc_root3 = parser.parse_into_document(doc, input);
+    //... doc_root3 is a pointer inside doc
+    if(simdjson::to_string(doc_root3) != "[1,2,3]") { return false; }
 
     // Here let us take moving the document:
     dom::document docm = std::move(doc);
-    element doc_root3 = docm.root();
-    if(simdjson::to_string(doc_root3) != "[1,2,3]") { return false; }
+    element doc_root4 = docm.root();
+    if(simdjson::to_string(doc_root4) != "[1,2,3]") { return false; }
     return true;
   }
 

tests/ondemand/ondemand_number_tests.cpp

@@ -278,6 +278,19 @@ namespace number_tests {
     TEST_SUCCEED();
   }
 
+  bool issue2099() {
+    TEST_START();
+    ondemand::parser parser;
+    auto json = "1000000000.000000001"_padded;
+    ondemand::document doc;
+    ASSERT_SUCCESS(parser.iterate(json).get(doc));
+    ondemand::number number;
+    ASSERT_SUCCESS(doc.get_number().get(number));
+    ASSERT_EQUAL(number.get_number_type(), ondemand::number_type::floating_point_number);
+    ASSERT_EQUAL(number.get_double(), 1e9);
+    TEST_SUCCEED();
+}
+
   bool issue1878() {
     TEST_START();
     ondemand::parser parser;
@@ -415,8 +428,23 @@ namespace number_tests {
     TEST_SUCCEED();
   }
 
+
+  bool issue2093() {
+    TEST_START();
+    ondemand::parser parser;
+    ondemand::document doc;
+    padded_string docdata = R"(0.95000000000000000000)"_padded;
+    ASSERT_SUCCESS(parser.iterate(docdata).get(doc));
+    double d;
+    ASSERT_SUCCESS(doc.get_double().get(d));
+    ASSERT_EQUAL(d, 0.95);
+    TEST_SUCCEED();
+  }
+
   bool run() {
-    return issue2045() &&
+    return issue2099() &&
+           issue2093() &&
+           issue2045() &&
            issue2017() &&
            issue_1898() &&
            issue1878() &&