30 Thread support library [thread]

[futures] describes components that a C++ program can use to retrieve in one thread the result (value or exception) from a function that has run in the same thread or another thread. [ Note: These components are not restricted to multi-threaded programs but can be useful in single-threaded programs as well.  — end note ]

Header <future> synopsis

namespace std {
  enum class future_errc {
    broken_promise = implementation-defined,
    future_already_retrieved = implementation-defined,
    promise_already_satisfied = implementation-defined,
    no_state = implementation-defined
  };

  enum class launch : unspecified {
    async = unspecified,
    deferred = unspecified,
    implementation-defined
  };

  enum class future_status {
    ready,
    timeout,
    deferred
  };

  template <> struct is_error_code_enum<future_errc> : public true_type { };
  error_code make_error_code(future_errc e) noexcept;
  error_condition make_error_condition(future_errc e) noexcept;

  const error_category& future_category() noexcept;

  class future_error;

  template <class R> class promise;
  template <class R> class promise<R&>;
  template <> class promise<void>;

  template <class R>
    void swap(promise<R>& x, promise<R>& y) noexcept;

  template <class R, class Alloc>
    struct uses_allocator<promise<R>, Alloc>;

  template <class R> class future;
  template <class R> class future<R&>;
  template <> class future<void>;

  template <class R> class shared_future;
  template <class R> class shared_future<R&>;
  template <> class shared_future<void>;

  template <class> class packaged_task;   // undefined
  template <class R, class... ArgTypes>
    class packaged_task<R(ArgTypes...)>;

  template <class R, class... ArgTypes>
    void swap(packaged_task<R(ArgTypes...)>&, packaged_task<R(ArgTypes...)>&) noexcept;

  template <class R, class Alloc>
    struct uses_allocator<packaged_task<R>, Alloc>;

  template <class F, class... Args>
    future<result_of_t<decay_t<F>(decay_t<Args>...)>>
    async(F&& f, Args&&... args);
  template <class F, class... Args>
    future<result_of_t<decay_t<F>(decay_t<Args>...)>>
    async(launch policy, F&& f, Args&&... args);
}

The enum type launch is a bitmask type ([bitmask.types]) with launch::async and launch::deferred denoting individual bits. [ Note: Implementations can provide bitmasks to specify restrictions on task interaction by functions launched by async() applicable to a corresponding subset of available launch policies. Implementations can extend the behavior of the first overload of async() by adding their extensions to the launch policy under the “as if” rule.  — end note ]

The enum values of future_errc are distinct and not zero.

Many of the classes introduced in this sub-clause use some state to communicate results. This shared state consists of some state information and some (possibly not yet evaluated) result, which can be a (possibly void) value or an exception. [ Note: Futures, promises, and tasks defined in this clause reference such shared state.  — end note ]

[ Note: The result can be any kind of object including a function to compute that result, as used by async when policy is launch::deferred.  — end note ]

An asynchronous return object is an object that reads results from a shared state. A waiting function of an asynchronous return object is one that potentially blocks to wait for the shared state to be made ready. If a waiting function can return before the state is made ready because of a timeout ([thread.req.lockable]), then it is a timed waiting function, otherwise it is a non-timed waiting function.

An asynchronous provider is an object that provides a result to a shared state. The result of a shared state is set by respective functions on the asynchronous provider. [ Note: Such as promises or tasks.  — end note ] The means of setting the result of a shared state is specified in the description of those classes and functions that create such a state object.

When an asynchronous return object or an asynchronous provider is said to release its shared state, it means:

• (5.1)

  if the return object or provider holds the last reference to its shared state, the shared state is destroyed; and

• (5.2)

  the return object or provider gives up its reference to its shared state; and

• (5.3)

  these actions will not block for the shared state to become ready, except that it may block if all of the following are true: the shared state was created by a call to std::async, the shared state is not yet ready, and this was the last reference to the shared state.

When an asynchronous provider is said to make its shared state ready, it means:

• (6.1)

  first, the provider marks its shared state as ready; and

• (6.2)

  second, the provider unblocks any execution agents waiting for its shared state to become ready.

When an asynchronous provider is said to abandon its shared state, it means:

• (7.1)

  first, if that state is not ready, the provider

  • (7.1.1)

    stores an exception object of type future_error with an error condition of broken_promise within its shared state; and then

  • (7.1.2)

    makes its shared state ready;

• (7.2)

  second, the provider releases its shared state.

A shared state is ready only if it holds a value or an exception ready for retrieval. Waiting for a shared state to become ready may invoke code to compute the result on the waiting thread if so specified in the description of the class or function that creates the state object.

Calls to functions that successfully set the stored result of a shared state synchronize with ([intro.multithread]) calls to functions successfully detecting the ready state resulting from that setting. The storage of the result (whether normal or exceptional) into the shared state synchronizes with ([intro.multithread]) the successful return from a call to a waiting function on the shared state.

Some functions (e.g., promise::set_value_at_thread_exit) delay making the shared state ready until the calling thread exits. The destruction of each of that thread's objects with thread storage duration ([basic.stc.thread]) is sequenced before making that shared state ready.

Access to the result of the same shared state may conflict ([intro.multithread]). [ Note: this explicitly specifies that the result of the shared state is visible in the objects that reference this state in the sense of data race avoidance ([res.on.data.races]). For example, concurrent accesses through references returned by shared_future::get() ([futures.shared_future]) must either use read-only operations or provide additional synchronization.  — end note ]

The implementation shall provide the template promise and two specializations, promise<R&> and promise<void>. These differ only in the argument type of the member function set_value, as set out in its description, below.

The set_value, set_exception, set_value_at_thread_exit, and set_exception_at_thread_exit member functions behave as though they acquire a single mutex associated with the promise object while updating the promise object.

template <class R, class Alloc> struct uses_allocator<promise<R>, Alloc> : true_type { };

The class template future defines a type for asynchronous return objects which do not share their shared state with other asynchronous return objects. A default-constructed future object has no shared state. A future object with shared state can be created by functions on asynchronous providers ([futures.state]) or by the move constructor and shares its shared state with the original asynchronous provider. The result (value or exception) of a future object can be set by calling a respective function on an object that shares the same shared state.

[ Note: Member functions of future do not synchronize with themselves or with member functions of shared_future.  — end note ]

The effect of calling any member function other than the destructor, the move-assignment operator, or valid on a future object for which valid() == false is undefined. [ Note: Implementations are encouraged to detect this case and throw an object of type future_error with an error condition of future_errc::no_state.  — end note ]

namespace std {
  template <class R>
  class future {
  public:
    future() noexcept;
    future(future &&) noexcept;
    future(const future& rhs) = delete;
    ~future();
    future& operator=(const future& rhs) = delete;
    future& operator=(future&&) noexcept;
    shared_future<R> share();

    // retrieving the value
    see below get();

    // functions to check state
    bool valid() const noexcept;

    void wait() const;
    template <class Rep, class Period>
      future_status wait_for(const chrono::duration<Rep, Period>& rel_time) const;
    template <class Clock, class Duration>
      future_status wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
  };
}

The implementation shall provide the template future and two specializations, future<R&> and future<void>. These differ only in the return type and return value of the member function get, as set out in its description, below.

future() noexcept;

The class template shared_future defines a type for asynchronous return objects which may share their shared state with other asynchronous return objects. A default-constructed shared_future object has no shared state. A shared_future object with shared state can be created by conversion from a future object and shares its shared state with the original asynchronous provider ([futures.state]) of the shared state. The result (value or exception) of a shared_future object can be set by calling a respective function on an object that shares the same shared state.

[ Note: Member functions of shared_future do not synchronize with themselves, but they synchronize with the shared shared state.  — end note ]

The effect of calling any member function other than the destructor, the move-assignment operator, or valid() on a shared_future object for which valid() == false is undefined. [ Note: Implementations are encouraged to detect this case and throw an object of type future_error with an error condition of future_errc::no_state.  — end note ]

namespace std {
  template <class R>
  class shared_future {
  public:
    shared_future() noexcept;
    shared_future(const shared_future& rhs);
    shared_future(future<R>&&) noexcept;
    shared_future(shared_future&& rhs) noexcept;
    ~shared_future();
    shared_future& operator=(const shared_future& rhs);
    shared_future& operator=(shared_future&& rhs) noexcept;

    // retrieving the value
    see below get() const;

    // functions to check state
    bool valid() const noexcept;

    void wait() const;
    template <class Rep, class Period>
      future_status wait_for(const chrono::duration<Rep, Period>& rel_time) const;
    template <class Clock, class Duration>
      future_status wait_until(const chrono::time_point<Clock, Duration>& abs_time) const;
  };
}

The implementation shall provide the template shared_future and two specializations, shared_future<R&> and shared_future<void>. These differ only in the return type and return value of the member function get, as set out in its description, below.

shared_future() noexcept;

The function template async provides a mechanism to launch a function potentially in a new thread and provides the result of the function in a future object with which it shares a shared state.

template <class F, class... Args> future<result_of_t<decay_t<F>(decay_t<Args>...)>> async(F&& f, Args&&... args); template <class F, class... Args> future<result_of_t<decay_t<F>(decay_t<Args>...)>> async(launch policy, F&& f, Args&&... args);

[ Example:

int work1(int value);
int work2(int value);
int work(int value) {
  auto handle = std::async([=]{ return work2(value); });
  int tmp = work1(value);
  return tmp + handle.get();    // #1
}

[ Note: Line #1 might not result in concurrency because the async call uses the default policy, which may use launch::deferred, in which case the lambda might not be invoked until the get() call; in that case, work1 and work2 are called on the same thread and there is no concurrency.  — end note ]  — end example ]

The class template packaged_task defines a type for wrapping a function or callable object so that the return value of the function or callable object is stored in a future when it is invoked.

When the packaged_task object is invoked, its stored task is invoked and the result (whether normal or exceptional) stored in the shared state. Any futures that share the shared state will then be able to access the stored result.

namespace std {
  template<class> class packaged_task; // undefined

  template<class R, class... ArgTypes>
  class packaged_task<R(ArgTypes...)> {
  public:
    // construction and destruction
    packaged_task() noexcept;
    template <class F>
      explicit packaged_task(F&& f);
    template <class F, class Allocator>
      explicit packaged_task(allocator_arg_t, const Allocator& a, F&& f);
    ~packaged_task();

    // no copy
    packaged_task(const packaged_task&) = delete;
    packaged_task& operator=(const packaged_task&) = delete;

    // move support
    packaged_task(packaged_task&& rhs) noexcept;
    packaged_task& operator=(packaged_task&& rhs) noexcept;
    void swap(packaged_task& other) noexcept;

    bool valid() const noexcept;

    // result retrieval
    future<R> get_future();        

    // execution
    void operator()(ArgTypes... );
    void make_ready_at_thread_exit(ArgTypes...);

    void reset();
  };
  template <class R, class... ArgTypes>
    void swap(packaged_task<R(ArgTypes...)>& x, packaged_task<R(ArgTypes...)>& y) noexcept;
  template <class R, class Alloc>
    struct uses_allocator<packaged_task<R>, Alloc>;
}