Implement a spawn function in a new process module (#2005)

Signed-off-by: Juan Cruz Viotti <[email protected]>

Author: jviotti

.github/workflows/website-build.yml

@@ -17,6 +17,7 @@ jobs:
           cmake -S . -B ./build
           -DCMAKE_BUILD_TYPE:STRING=Release
           -DSOURCEMETA_CORE_LANG_IO:BOOL=OFF
+          -DSOURCEMETA_CORE_LANG_PROCESS:BOOL=OFF
           -DSOURCEMETA_CORE_TIME:BOOL=OFF
           -DSOURCEMETA_CORE_UUID:BOOL=OFF
           -DSOURCEMETA_CORE_REGEX:BOOL=OFF

.github/workflows/website-deploy.yml

@@ -27,6 +27,7 @@ jobs:
           cmake -S . -B ./build
           -DCMAKE_BUILD_TYPE:STRING=Release
           -DSOURCEMETA_CORE_LANG_IO:BOOL=OFF
+          -DSOURCEMETA_CORE_LANG_PROCESS:BOOL=OFF
           -DSOURCEMETA_CORE_TIME:BOOL=OFF
           -DSOURCEMETA_CORE_UUID:BOOL=OFF
           -DSOURCEMETA_CORE_REGEX:BOOL=OFF

CMakeLists.txt

@@ -4,6 +4,7 @@ list(APPEND CMAKE_MODULE_PATH "${PROJECT_SOURCE_DIR}/cmake")
 
 # Options
 option(SOURCEMETA_CORE_LANG_IO "Build the Sourcemeta Core language I/O library" ON)
+option(SOURCEMETA_CORE_LANG_PROCESS "Build the Sourcemeta Core language Process library" ON)
 option(SOURCEMETA_CORE_LANG_PARALLEL "Build the Sourcemeta Core language parallel library" ON)
 option(SOURCEMETA_CORE_TIME "Build the Sourcemeta Core time library" ON)
 option(SOURCEMETA_CORE_UUID "Build the Sourcemeta Core UUID library" ON)
@@ -58,6 +59,10 @@ if(SOURCEMETA_CORE_LANG_IO)
   add_subdirectory(src/lang/io)
 endif()
 
+if(SOURCEMETA_CORE_LANG_PROCESS)
+  add_subdirectory(src/lang/process)
+endif()
+
 if(SOURCEMETA_CORE_LANG_PARALLEL)
   find_package(Threads REQUIRED)
   add_subdirectory(src/lang/parallel)
@@ -160,6 +165,10 @@ if(SOURCEMETA_CORE_TESTS)
     add_subdirectory(test/io)
   endif()
 
+  if(SOURCEMETA_CORE_LANG_PROCESS)
+    add_subdirectory(test/process)
+  endif()
+
   if(SOURCEMETA_CORE_LANG_PARALLEL)
     add_subdirectory(test/parallel)
   endif()

config.cmake.in

@@ -5,6 +5,7 @@ list(APPEND SOURCEMETA_CORE_COMPONENTS ${Core_FIND_COMPONENTS})
 list(APPEND SOURCEMETA_CORE_COMPONENTS ${core_FIND_COMPONENTS})
 if(NOT SOURCEMETA_CORE_COMPONENTS)
   list(APPEND SOURCEMETA_CORE_COMPONENTS io)
+  list(APPEND SOURCEMETA_CORE_COMPONENTS process)
   list(APPEND SOURCEMETA_CORE_COMPONENTS parallel)
   list(APPEND SOURCEMETA_CORE_COMPONENTS time)
   list(APPEND SOURCEMETA_CORE_COMPONENTS uuid)
@@ -28,6 +29,8 @@ include(CMakeFindDependencyMacro)
 foreach(component ${SOURCEMETA_CORE_COMPONENTS})
   if(component STREQUAL "io")
     include("${CMAKE_CURRENT_LIST_DIR}/sourcemeta_core_io.cmake")
+  elseif(component STREQUAL "process")
+    include("${CMAKE_CURRENT_LIST_DIR}/sourcemeta_core_process.cmake")
   elseif(component STREQUAL "parallel")
     find_dependency(Threads)
     include("${CMAKE_CURRENT_LIST_DIR}/sourcemeta_core_parallel.cmake")

src/lang/process/CMakeLists.txt

@@ -0,0 +1,7 @@
+sourcemeta_library(NAMESPACE sourcemeta PROJECT core NAME process
+  PRIVATE_HEADERS error.h
+  SOURCES spawn.cc)
+
+if(SOURCEMETA_CORE_INSTALL)
+  sourcemeta_library_install(NAMESPACE sourcemeta PROJECT core NAME process)
+endif()

src/lang/process/include/sourcemeta/core/process.h

@@ -0,0 +1,72 @@
+#ifndef SOURCEMETA_CORE_PROCESS_H_
+#define SOURCEMETA_CORE_PROCESS_H_
+
+#ifndef SOURCEMETA_CORE_PROCESS_EXPORT
+#include <sourcemeta/core/process_export.h>
+#endif
+
+// NOLINTBEGIN(misc-include-cleaner)
+#include <sourcemeta/core/process_error.h>
+// NOLINTEND(misc-include-cleaner)
+
+#include <filesystem>       // std::filesystem
+#include <initializer_list> // std::initializer_list
+#include <span>             // std::span
+#include <string_view>      // std::string_view
+
+/// @defgroup process Process
+/// @brief Process related utilities
+///
+/// This functionality is included as follows:
+///
+/// ```cpp
+/// #include <sourcemeta/core/process.h>
+/// ```
+
+namespace sourcemeta::core {
+
+/// @ingroup process
+///
+/// Spawn a program piping its output to the current stdio configuration.
+/// The directory parameter specifies the working directory for the spawned
+/// process. It must be an absolute path to an existing directory.
+///
+/// ```cpp
+/// #include <sourcemeta/core/process.h>
+/// #include <cassert>
+///
+/// const auto exit_code{sourcemeta::core::spawn("echo", {"foo"})};
+/// assert(exit_code == 0);
+/// ```
+SOURCEMETA_CORE_PROCESS_EXPORT
+auto spawn(const std::string &program,
+           std::initializer_list<std::string_view> arguments,
+           const std::filesystem::path &directory =
+               std::filesystem::current_path()) -> int;
+
+/// @ingroup process
+///
+/// Spawn a program piping its output to the current stdio configuration.
+/// This overload accepts a span for dynamic argument lists.
+/// The directory parameter specifies the working directory for the spawned
+/// process. It must be an absolute path to an existing directory.
+///
+/// ```cpp
+/// #include <sourcemeta/core/process.h>
+/// #include <vector>
+/// #include <string_view>
+/// #include <cassert>
+///
+/// std::vector<std::string_view> arguments{"foo", "bar"};
+/// const auto exit_code{sourcemeta::core::spawn("echo", arguments)};
+/// assert(exit_code == 0);
+/// ```
+SOURCEMETA_CORE_PROCESS_EXPORT
+auto spawn(
+    const std::string &program, std::span<const std::string_view> arguments,
+    const std::filesystem::path &directory = std::filesystem::current_path())
+    -> int;
+
+} // namespace sourcemeta::core
+
+#endif

src/lang/process/include/sourcemeta/core/process_error.h

@@ -0,0 +1,83 @@
+#ifndef SOURCEMETA_CORE_PROCESS_ERROR_H_
+#define SOURCEMETA_CORE_PROCESS_ERROR_H_
+
+#ifndef SOURCEMETA_CORE_PROCESS_EXPORT
+#include <sourcemeta/core/process_export.h>
+#endif
+
+#include <exception>        // std::exception
+#include <initializer_list> // std::initializer_list
+#include <span>             // std::span
+#include <string>           // std::string
+#include <string_view>      // std::string_view
+#include <utility>          // std::move
+#include <vector>           // std::vector
+
+namespace sourcemeta::core {
+
+// Exporting symbols that depends on the standard C++ library is considered
+// safe.
+// https://learn.microsoft.com/en-us/cpp/error-messages/compiler-warnings/compiler-warning-level-2-c4275?view=msvc-170&redirectedfrom=MSDN
+#if defined(_MSC_VER)
+#pragma warning(disable : 4251 4275)
+#endif
+
+/// @ingroup process
+/// An executable program could not be found
+class SOURCEMETA_CORE_PROCESS_EXPORT ProcessProgramNotNotFoundError
+    : public std::exception {
+public:
+  ProcessProgramNotNotFoundError(std::string program)
+      : program_{std::move(program)} {}
+
+  [[nodiscard]] auto what() const noexcept -> const char * override {
+    return "Could not locate the requested program";
+  }
+
+  [[nodiscard]] auto program() const noexcept -> std::string_view {
+    return this->program_;
+  }
+
+private:
+  std::string program_;
+};
+
+/// @ingroup process
+/// A spawned process terminated abnormally
+class SOURCEMETA_CORE_PROCESS_EXPORT ProcessSpawnError : public std::exception {
+public:
+  ProcessSpawnError(std::string program,
+                    std::initializer_list<std::string_view> arguments)
+      : program_{std::move(program)},
+        arguments_{arguments.begin(), arguments.end()} {}
+
+  ProcessSpawnError(std::string program,
+                    std::span<const std::string_view> arguments)
+      : program_{std::move(program)},
+        arguments_{arguments.begin(), arguments.end()} {}
+
+  [[nodiscard]] auto what() const noexcept -> const char * override {
+    return "Process terminated abnormally";
+  }
+
+  [[nodiscard]] auto program() const noexcept -> std::string_view {
+    return this->program_;
+  }
+
+  [[nodiscard]] auto arguments() const noexcept
+      -> const std::vector<std::string> & {
+    return this->arguments_;
+  }
+
+private:
+  std::string program_;
+  std::vector<std::string> arguments_;
+};
+
+#if defined(_MSC_VER)
+#pragma warning(default : 4251 4275)
+#endif
+
+} // namespace sourcemeta::core
+
+#endif

src/lang/process/spawn.cc

@@ -0,0 +1,152 @@
+#include <sourcemeta/core/process.h>
+
+#include <cassert>          // assert
+#include <filesystem>       // std::filesystem
+#include <initializer_list> // std::initializer_list
+#include <span>             // std::span
+#include <vector>           // std::vector
+
+#if defined(_WIN32) && !defined(__MSYS__) && !defined(__CYGWIN__) &&           \
+    !defined(__MINGW32__) && !defined(__MINGW64__)
+#define WIN32_LEAN_AND_MEAN
+#include <sstream>   // std::ostringstream
+#include <windows.h> // CreateProcess, PROCESS_INFORMATION, STARTUPINFO, WaitForSingleObject, GetExitCodeProcess
+#else
+#include <spawn.h> // posix_spawnp, posix_spawnattr_t, posix_spawnattr_init, posix_spawnattr_destroy, posix_spawn_file_actions_t, posix_spawn_file_actions_init, posix_spawn_file_actions_destroy, pid_t
+#include <sys/wait.h> // waitpid, WIFEXITED, WEXITSTATUS
+
+#if defined(__MSYS__) || defined(__CYGWIN__) || defined(__MINGW32__) ||        \
+    defined(__MINGW64__)
+#include <unistd.h> // chdir
+#endif
+
+extern char **environ;
+#endif
+
+namespace sourcemeta::core {
+
+auto spawn(const std::string &program,
+           std::span<const std::string_view> arguments,
+           const std::filesystem::path &directory) -> int {
+  assert(directory.is_absolute());
+  assert(std::filesystem::exists(directory));
+  assert(std::filesystem::is_directory(directory));
+
+#if defined(_WIN32) && !defined(__MSYS__) && !defined(__CYGWIN__) &&           \
+    !defined(__MINGW32__) && !defined(__MINGW64__)
+  std::ostringstream command_line;
+  command_line << program;
+
+  for (const auto &argument : arguments) {
+    command_line << " ";
+    // Quote arguments that contain spaces
+    const std::string arg_str{argument};
+    if (arg_str.find(' ') != std::string::npos) {
+      command_line << "\"" << arg_str << "\"";
+    } else {
+      command_line << arg_str;
+    }
+  }
+
+  std::string cmd_line_str = command_line.str();
+  std::vector<char> cmd_line(cmd_line_str.begin(), cmd_line_str.end());
+  cmd_line.push_back('\0');
+
+  STARTUPINFOA startup_info{};
+  startup_info.cb = sizeof(startup_info);
+  PROCESS_INFORMATION process_info{};
+  const std::string working_dir = directory.string();
+  const BOOL success =
+      CreateProcessA(nullptr,             // lpApplicationName
+                     cmd_line.data(),     // lpCommandLine (modifiable)
+                     nullptr,             // lpProcessAttributes
+                     nullptr,             // lpThreadAttributes
+                     TRUE,                // bInheritHandles
+                     0,                   // dwCreationFlags
+                     nullptr,             // lpEnvironment
+                     working_dir.c_str(), // lpCurrentDirectory
+                     &startup_info,       // lpStartupInfo
+                     &process_info        // lpProcessInformation
+      );
+
+  if (!success) {
+    throw ProcessProgramNotNotFoundError{program};
+  }
+
+  WaitForSingleObject(process_info.hProcess, INFINITE);
+
+  DWORD exit_code;
+  if (!GetExitCodeProcess(process_info.hProcess, &exit_code)) {
+    CloseHandle(process_info.hProcess);
+    CloseHandle(process_info.hThread);
+    throw ProcessSpawnError{program, arguments};
+  }
+
+  CloseHandle(process_info.hProcess);
+  CloseHandle(process_info.hThread);
+
+  return static_cast<int>(exit_code);
+#else
+  std::vector<const char *> argv;
+  argv.reserve(arguments.size() + 2);
+  argv.push_back(program.c_str());
+
+  for (const auto &argument : arguments) {
+    argv.push_back(argument.data());
+  }
+
+  argv.push_back(nullptr);
+
+  posix_spawnattr_t attributes;
+  posix_spawnattr_init(&attributes);
+
+  posix_spawn_file_actions_t file_actions;
+  posix_spawn_file_actions_init(&file_actions);
+
+#if defined(__MSYS__) || defined(__CYGWIN__) || defined(__MINGW32__) ||        \
+    defined(__MINGW64__)
+  const std::filesystem::path original_directory{
+      std::filesystem::current_path()};
+  std::filesystem::current_path(directory);
+#else
+  posix_spawn_file_actions_addchdir_np(&file_actions, directory.c_str());
+#endif
+
+  pid_t process_id;
+  const int spawn_result{
+      posix_spawnp(&process_id, program.c_str(), &file_actions, &attributes,
+                   const_cast<char *const *>(argv.data()), environ)};
+
+  posix_spawn_file_actions_destroy(&file_actions);
+  posix_spawnattr_destroy(&attributes);
+
+#if defined(__MSYS__) || defined(__CYGWIN__) || defined(__MINGW32__) ||        \
+    defined(__MINGW64__)
+  std::filesystem::current_path(original_directory);
+#endif
+
+  if (spawn_result != 0) {
+    throw ProcessProgramNotNotFoundError{program};
+  }
+
+  int status;
+  waitpid(process_id, &status, 0);
+
+  if (WIFEXITED(status)) {
+    return WEXITSTATUS(status);
+  }
+
+  throw ProcessSpawnError{program, arguments};
+#endif
+}
+
+auto spawn(const std::string &program,
+           std::initializer_list<std::string_view> arguments,
+           const std::filesystem::path &directory) -> int {
+  return spawn(
+      program,
+      std::span<const std::string_view>{arguments.begin(), arguments.size()},
+      directory);
+}
+
+} // namespace sourcemeta::core

test/parallel/parallel_for_each_test.cc

@@ -1,6 +1,6 @@
 #include <gtest/gtest.h>
 
-#include "sourcemeta/core/parallel_for_each.h"
+#include <sourcemeta/core/parallel.h>
 
 #include <algorithm>
 #include <atomic>