migrate to c++20

.github/workflows/ubuntu_install.yml

@@ -26,8 +26,7 @@ jobs:
     strategy:
       matrix:
         include:
-          - {shared: ON}
-          - {shared: OFF}
+          shared: [ON, OFF]
     steps:
       - uses: actions/checkout@692973e3d937129bcbf40652eb9f2f61becf3332 # v4.1.7
       - name: Setup Ninja

CMakeLists.txt

@@ -5,7 +5,6 @@ project(ada
   LANGUAGES C CXX
   VERSION 2.9.2
 )
-set(CMAKE_CXX_STANDARD 17)
 set(ADA_LIB_VERSION "2.9.2" CACHE STRING "ada library version")
 set(ADA_LIB_SOVERSION "2" CACHE STRING "ada library soversion")
 
@@ -36,20 +35,20 @@ if(ADA_TESTING OR ADA_BENCHMARKS OR ADA_TOOLS)
     CPMAddPackage(
       NAME GTest
       GITHUB_REPOSITORY google/googletest
-      VERSION 1.14.0
+      VERSION 1.15.2
       OPTIONS  "BUILD_GMOCK OFF" "INSTALL_GTEST OFF"
     )
   endif()
   # We use simdjson in both the benchmarks and tests
   if(Git_FOUND AND (ADA_TESTING OR ADA_BENCHMARKS))
-    CPMAddPackage("gh:simdjson/simdjson@3.9.1")
+    CPMAddPackage("gh:simdjson/simdjson@3.10.1")
   endif()
   # We use Google Benchmark, but it does not build under several 32-bit systems.
   if(Git_FOUND AND ADA_BENCHMARKS AND (CMAKE_SIZEOF_VOID_P EQUAL 8))
     CPMAddPackage(
       NAME benchmark
       GITHUB_REPOSITORY google/benchmark
-      GIT_TAG f91b6b4
+      VERSION 1.9.0
       OPTIONS "BENCHMARK_ENABLE_TESTING OFF"
               "BENCHMARK_ENABLE_INSTALL OFF"
               "BENCHMARK_ENABLE_WERROR OFF"

README.md

@@ -28,7 +28,7 @@ standard libraries, follow the RFC 3986. The following table illustrates possibl
 ### Requirements
 
 The project is otherwise self-contained and it has no dependency.
-A recent C++ compiler supporting C++17. We test GCC 9 or better, LLVM 10 or better and Microsoft Visual Studio 2022.
+A recent C++ compiler supporting C++20. We test GCC 12 or better, LLVM 12 or better and Microsoft Visual Studio 2022.
 
 ## Ada is fast.
 
@@ -86,7 +86,7 @@ Linux or macOS users might follow the following instructions if they have a rece
    ```
 2. Compile
    ```
-   c++ -std=c++17 -o demo demo.cpp
+   c++ -std=c++20 -o demo demo.cpp
    ```
 3. `./demo`
 
@@ -279,7 +279,7 @@ E.g., if you grab our single-header C++ files (`ada.cpp` and `ada.h`), as well a
 you can often compile a C program (`demo.c`) as follows under Linux/macOS systems:
 
 ```
-c++ -c ada.cpp -std=c++17
+c++ -c ada.cpp -std=c++20
 cc -c demo.c
 c++ demo.o ada.o -o cdemo
 ./cdemo

cmake/ada-flags.cmake

@@ -50,7 +50,7 @@ endif()
 set(CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/tools/cmake")
 set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
-set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD 20)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 

fuzz/build.sh

@@ -6,39 +6,39 @@ mkdir build
 AMALGAMATE_OUTPUT_PATH=./build/singleheader python3 singleheader/amalgamate.py
 
 $CXX $CFLAGS $CXXFLAGS \
-     -std=c++17 \
+     -std=c++20 \
      -I build/singleheader \
      -c fuzz/parse.cc -o parse.o
 
 $CXX $CFLAGS $CXXFLAGS $LIB_FUZZING_ENGINE parse.o \
      -o $OUT/parse
 
 $CXX $CFLAGS $CXXFLAGS \
-     -std=c++17 \
+     -std=c++20 \
      -I build/singleheader \
      -c fuzz/can_parse.cc -o can_parse.o
 
 $CXX $CFLAGS $CXXFLAGS $LIB_FUZZING_ENGINE can_parse.o \
      -o $OUT/can_parse
 
 $CXX $CFLAGS $CXXFLAGS \
-     -std=c++17 \
+     -std=c++20 \
      -I build/singleheader \
      -c fuzz/idna.cc -o idna.o
 
 $CXX $CFLAGS $CXXFLAGS $LIB_FUZZING_ENGINE idna.o \
      -o $OUT/idna
 
 $CXX $CFLAGS $CXXFLAGS \
-     -std=c++17 \
+     -std=c++20 \
      -I build/singleheader \
      -c fuzz/url_search_params.cc -o url_search_params.o
 
 $CXX $CFLAGS $CXXFLAGS $LIB_FUZZING_ENGINE url_search_params.o \
      -o $OUT/url_search_params
 
 $CXX $CFLAGS $CXXFLAGS \
-     -std=c++17 \
+     -std=c++20 \
      -I build/singleheader \
      -c build/singleheader/ada.cpp -o ada.o
 

fuzz/parse.cc

@@ -219,7 +219,7 @@ extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
   }
 
   // Only used for avoiding dead-code elimination
-  printf("length of url is %d\n", length);
+  printf("length of url is %zu\n", length);
 
   return 0;
 }  // extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {

include/ada/character_sets-inl.h

@@ -512,7 +512,7 @@ constexpr uint8_t WWW_FORM_URLENCODED_PERCENT_ENCODE[32] = {
     // F8     F9     FA     FB     FC     FD     FE     FF
     0x01 | 0x02 | 0x04 | 0x08 | 0x10 | 0x20 | 0x40 | 0x80};
 
-ada_really_inline bool bit_at(const uint8_t a[], const uint8_t i) {
+ada_really_inline constexpr bool bit_at(const uint8_t a[], const uint8_t i) {
   return !!(a[i >> 3] & (1 << (i & 7)));
 }
 

include/ada/character_sets.h

@@ -18,7 +18,7 @@
  * @brief Includes the definitions for unicode character sets.
  */
 namespace ada::character_sets {
-ada_really_inline bool bit_at(const uint8_t a[], uint8_t i);
+ada_really_inline constexpr bool bit_at(const uint8_t a[], uint8_t i);
 }  // namespace ada::character_sets
 
 #endif  // ADA_CHARACTER_SETS_H

include/ada/checkers-inl.h

@@ -5,32 +5,28 @@
 #ifndef ADA_CHECKERS_INL_H
 #define ADA_CHECKERS_INL_H
 
-#include "ada/common_defs.h"
-
-#include <algorithm>
+#include <bit>
 #include <string_view>
-#include <cstring>
 
 namespace ada::checkers {
 
-inline bool has_hex_prefix_unsafe(std::string_view input) {
+constexpr bool has_hex_prefix_unsafe(std::string_view input) {
   // This is actually efficient code, see has_hex_prefix for the assembly.
-  uint32_t value_one = 1;
-  bool is_little_endian = (reinterpret_cast<char*>(&value_one)[0] == 1);
-  uint16_t word0x{};
-  std::memcpy(&word0x, "0x", 2);  // we would use bit_cast in C++20 and the
-                                  // function could be constexpr.
-  uint16_t two_first_bytes{};
-  std::memcpy(&two_first_bytes, input.data(), 2);
-  if (is_little_endian) {
+  constexpr bool is_little_endian = std::endian::native == std::endian::little;
+  constexpr auto word0x =
+      std::bit_cast<uint16_t>(static_cast<uint16_t>(0x7830));
+  uint16_t two_first_bytes =
+      static_cast<uint16_t>(input[0]) |
+      static_cast<uint16_t>((static_cast<uint16_t>(input[1]) << 8));
+  if constexpr (is_little_endian) {
     two_first_bytes |= 0x2000;
   } else {
     two_first_bytes |= 0x020;
   }
   return two_first_bytes == word0x;
 }
 
-inline bool has_hex_prefix(std::string_view input) {
+constexpr bool has_hex_prefix(std::string_view input) {
   return input.size() >= 2 && has_hex_prefix_unsafe(input);
 }
 
@@ -42,26 +38,18 @@ constexpr bool is_alpha(char x) noexcept {
   return (to_lower(x) >= 'a') && (to_lower(x) <= 'z');
 }
 
-inline constexpr bool is_windows_drive_letter(std::string_view input) noexcept {
+constexpr bool is_windows_drive_letter(std::string_view input) noexcept {
   return input.size() >= 2 &&
          (is_alpha(input[0]) && ((input[1] == ':') || (input[1] == '|'))) &&
          ((input.size() == 2) || (input[2] == '/' || input[2] == '\\' ||
                                   input[2] == '?' || input[2] == '#'));
 }
 
-inline constexpr bool is_normalized_windows_drive_letter(
+constexpr bool is_normalized_windows_drive_letter(
     std::string_view input) noexcept {
   return input.size() >= 2 && (is_alpha(input[0]) && (input[1] == ':'));
 }
 
-ada_really_inline bool begins_with(std::string_view view,
-                                   std::string_view prefix) {
-  // in C++20, you have view.begins_with(prefix)
-  // std::equal is constexpr in C++20
-  return view.size() >= prefix.size() &&
-         std::equal(prefix.begin(), prefix.end(), view.begin());
-}
-
 }  // namespace ada::checkers
 
 #endif  // ADA_CHECKERS_INL_H

include/ada/checkers.h

@@ -43,12 +43,12 @@ constexpr bool is_alpha(char x) noexcept;
  *
  * @see has_hex_prefix
  */
-inline bool has_hex_prefix_unsafe(std::string_view input);
+constexpr bool has_hex_prefix_unsafe(std::string_view input);
 /**
  * @private
  * Check whether a string starts with 0x or 0X.
  */
-inline bool has_hex_prefix(std::string_view input);
+constexpr bool has_hex_prefix(std::string_view input);
 
 /**
  * @private
@@ -79,13 +79,6 @@ inline constexpr bool is_windows_drive_letter(std::string_view input) noexcept;
 inline constexpr bool is_normalized_windows_drive_letter(
     std::string_view input) noexcept;
 
-/**
- * @private
- * @warning Will be removed when Ada requires C++20.
- */
-ada_really_inline bool begins_with(std::string_view view,
-                                   std::string_view prefix);
-
 /**
  * @private
  * Returns true if an input is an ipv4 address. It is assumed that the string

include/ada/helpers.h

@@ -88,8 +88,8 @@ ada_really_inline void remove_ascii_tab_or_newline(std::string& input) noexcept;
  * Return the substring from input going from index pos to the end.
  * This function cannot throw.
  */
-ada_really_inline std::string_view substring(std::string_view input,
-                                             size_t pos) noexcept;
+ada_really_inline constexpr std::string_view substring(std::string_view input,
+                                                       size_t pos) noexcept;
 
 /**
  * @private

include/ada/url-inl.h

@@ -81,7 +81,7 @@ inline std::ostream &operator<<(std::ostream &out, const ada::url &u) {
     out.host_start = out.protocol_end;
     out.host_end = out.host_start;
 
-    if (!has_opaque_path && checkers::begins_with(path, "//")) {
+    if (!has_opaque_path && path.starts_with("//")) {
       // If url's host is null, url does not have an opaque path, url's path's
       // size is greater than 1, and url's path[0] is the empty string, then
       // append U+002F (/) followed by U+002E (.) to output.
@@ -196,7 +196,7 @@ inline void url::copy_scheme(const ada::url &u) {
     if (port.has_value()) {
       output += ":" + get_port();
     }
-  } else if (!has_opaque_path && checkers::begins_with(path, "//")) {
+  } else if (!has_opaque_path && path.starts_with("//")) {
     // If url's host is null, url does not have an opaque path, url's path's
     // size is greater than 1, and url's path[0] is the empty string, then
     // append U+002F (/) followed by U+002E (.) to output.

include/ada/url_aggregator-inl.h

@@ -27,7 +27,7 @@ inline void url_aggregator::update_base_authority(
       base.protocol_end, base.host_start - base.protocol_end);
   ada_log("url_aggregator::update_base_authority ", input);
 
-  bool input_starts_with_dash = checkers::begins_with(input, "//");
+  bool input_starts_with_dash = input.starts_with("//");
   uint32_t diff = components.host_start - components.protocol_end;
 
   buffer.erase(components.protocol_end,
@@ -267,7 +267,7 @@ inline void url_aggregator::update_base_pathname(const std::string_view input) {
   ADA_ASSERT_TRUE(!helpers::overlaps(input, buffer));
   ADA_ASSERT_TRUE(validate());
 
-  const bool begins_with_dashdash = checkers::begins_with(input, "//");
+  const bool begins_with_dashdash = input.starts_with("//");
   if (!begins_with_dashdash && has_dash_dot()) {
     ada_log("url_aggregator::update_base_pathname has /.: \n", to_diagram());
     // We must delete the ./

singleheader/README.md

@@ -11,13 +11,13 @@ This will create two new files (ada.h and ada.cpp).
 You can then compile the demo file as follows:
 
 ```
-c++ -std=c++17 -c demo.cpp
+c++ -std=c++20 -c demo.cpp
 ```
 
 It will produce a binary file (e.g., demo.o) which contains ada.cpp.
 
 ```
-c++ -std=c++17 -o demo demo.cpp
+c++ -std=c++20 -o demo demo.cpp
 ./demo
 ```
 
@@ -29,7 +29,7 @@ You may build and link using CMake (--target demo), because CMake can configure
 You may also build a C executable.
 
 ```
-c++ -c ada.cpp -std=c++17
+c++ -c ada.cpp -std=c++20
 cc -c demo.c
 c++ demo.o ada.o -o cdemo
 ./cdemo

src/checkers.cpp

@@ -43,7 +43,7 @@ ada_really_inline ada_constexpr bool is_ipv4(std::string_view view) noexcept {
     return false;
   }
   // It must start with 0x.
-  if (!std::equal(view.begin(), view.begin() + 2, "0x")) {
+  if (!view.starts_with("0x")) {
     return false;
   }
   // We must allow "0x".
@@ -59,7 +59,7 @@ ada_really_inline ada_constexpr bool is_ipv4(std::string_view view) noexcept {
 // for use with path_signature, we include all characters that need percent
 // encoding.
 static constexpr std::array<uint8_t, 256> path_signature_table =
-    []() constexpr {
+    []() consteval {
       std::array<uint8_t, 256> result{};
       for (size_t i = 0; i < 256; i++) {
         if (i <= 0x20 || i == 0x22 || i == 0x23 || i == 0x3c || i == 0x3e ||