c++-gtk-utils
Public Member Functions | Static Public Member Functions | Public Attributes | List of all members
Cgu::Thread::Future< Val > Class Template Reference

A class representing a pthread thread which will provide a value. More...

#include <c++-gtk-utils/future.h>

Inheritance diagram for Cgu::Thread::Future< Val >:
Cgu::IntrusiveLockCounter

Public Member Functions

 Future (const Future &)=delete
 
Futureoperator= (const Future &)=delete
 
bool run ()
 
Val get ()
 
Val move_get ()
 
bool cancel () noexcept
 
Cgu::Callback::SafeFunctor when (const Cgu::Callback::CallbackArg< const Val & > *cb, gint priority=G_PRIORITY_DEFAULT, GMainContext *context=0)
 
template<class When , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<When>::type, const Cgu::Callback::CallbackArg<const Val&>*>::value>::type>
Cgu::Callback::SafeFunctor when (When &&w, gint priority=G_PRIORITY_DEFAULT, GMainContext *context=0)
 
Cgu::Callback::SafeFunctor when (const Cgu::Callback::CallbackArg< const Val & > *cb, Cgu::Releaser &r, gint priority=G_PRIORITY_DEFAULT, GMainContext *context=0)
 
template<class When , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<When>::type, const Cgu::Callback::CallbackArg<const Val&>*>::value>::type>
Cgu::Callback::SafeFunctor when (When &&w, Cgu::Releaser &r, gint priority=G_PRIORITY_DEFAULT, GMainContext *context=0)
 
void fail (const Cgu::Callback::Callback *cb, GMainContext *context=0)
 
template<class Fail , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<Fail>::type, const Cgu::Callback::Callback*>::value>::type>
void fail (Fail &&f, GMainContext *context=0)
 
void fail (const Cgu::Callback::Callback *cb, Cgu::Releaser &r, GMainContext *context=0)
 
template<class Fail , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<Fail>::type, const Cgu::Callback::Callback*>::value>::type>
void fail (Fail &&f, Cgu::Releaser &r, GMainContext *context=0)
 
bool is_done () const noexcept
 
bool is_emitter_done () const noexcept
 
bool is_error () const noexcept
 
bool is_emitter_error () const noexcept
 
- Public Member Functions inherited from Cgu::IntrusiveLockCounter
 IntrusiveLockCounter (const IntrusiveLockCounter &)=delete
 
IntrusiveLockCounteroperator= (const IntrusiveLockCounter &)=delete
 
void ref () noexcept
 
void unref ()
 
 IntrusiveLockCounter ()
 
virtual ~IntrusiveLockCounter ()
 

Static Public Member Functions

template<class Ret , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (T &t, Ret(T::*func)())
 
template<class Ret , class Param1 , class Arg1 , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (T &t, Ret(T::*func)(Param1), Arg1 &&arg1)
 
template<class Ret , class Param1 , class Param2 , class Arg1 , class Arg2 , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (T &t, Ret(T::*func)(Param1, Param2), Arg1 &&arg1, Arg2 &&arg2)
 
template<class Ret , class Param1 , class Param2 , class Param3 , class Arg1 , class Arg2 , class Arg3 , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (T &t, Ret(T::*func)(Param1, Param2, Param3), Arg1 &&arg1, Arg2 &&arg2, Arg3 &&arg3)
 
template<class Ret , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (const T &t, Ret(T::*func)() const)
 
template<class Ret , class Param1 , class Arg1 , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (const T &t, Ret(T::*func)(Param1) const, Arg1 &&arg1)
 
template<class Ret , class Param1 , class Param2 , class Arg1 , class Arg2 , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (const T &t, Ret(T::*func)(Param1, Param2) const, Arg1 &&arg1, Arg2 &&arg2)
 
template<class Ret , class Param1 , class Param2 , class Param3 , class Arg1 , class Arg2 , class Arg3 , class T >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (const T &t, Ret(T::*func)(Param1, Param2, Param3) const, Arg1 &&arg1, Arg2 &&arg2, Arg3 &&arg3)
 
template<class Ret >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (Ret(*func)())
 
template<class Ret , class Param1 , class Arg1 >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (Ret(*func)(Param1), Arg1 &&arg1)
 
template<class Ret , class Param1 , class Param2 , class Arg1 , class Arg2 >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (Ret(*func)(Param1, Param2), Arg1 &&arg1, Arg2 &&arg2)
 
template<class Ret , class Param1 , class Param2 , class Param3 , class Arg1 , class Arg2 , class Arg3 >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (Ret(*func)(Param1, Param2, Param3), Arg1 &&arg1, Arg2 &&arg2, Arg3 &&arg3)
 
template<class Ret , class Param1 , class Param2 , class Param3 , class Param4 , class Arg1 , class Arg2 , class Arg3 , class Arg4 >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (Ret(*func)(Param1, Param2, Param3, Param4), Arg1 &&arg1, Arg2 &&arg2, Arg3 &&arg3, Arg4 &&arg4)
 
template<class Func >
static Cgu::IntrusivePtr< Cgu::Thread::Future< Val > > make (Func &&func)
 

Public Attributes

SafeEmitter done_emitter
 

Detailed Description

template<class Val>
class Cgu::Thread::Future< Val >

A class representing a pthread thread which will provide a value.

See also
Cgu::Thread::Thread Cgu::Thread::JoinableHandle Cgu::AsyncResult Cgu::Thread::make_future() Cgu::Thread::TaskManager

The Thread::Future class will launch a worker thread, run the function it represents in that thread until it returns, and store the return value so that it can be waited on and/or extracted by another thread. A new Thread::Future object representing the function to be called is normally created by calling Cgu::Thread::make_future() with a callable object, such as a lambda expression or the return value of std::bind. The worker thread is then started by calling run(), and the value extracted or waited for by calling get(). The run() method can only be called once, but any number of threads can wait for and/or extract the return value by calling the get() method. The class also provides a move_get() method, and a SafeEmitter done_emitter public object which emits when the worker thread has finished, and an associated when() function.

The template parameter type of Thread::Future is the type of the return value of the function or callable object called by the Thread::Future object. The return value can be any type, including any arbitrarily large tuple or other struct or standard C++ container.

A Thread::Future object cannot represent a function with a void return type - a compilation error will result if that is attempted. If no return value is wanted, then the Thread::Thread class can be used directly. (However, if in a particular usage this class is thought to be more convenient, the function to be represented by it can be wrapped by another function which provides a dummy return value, such as a dummy int. One possible case for this is where more than one thread wants to wait for the worker thread to terminate, as pthread_join() and so Thread::Thread::join() only give defined behaviour when called by one thread.)

A future object can also be constructed with Thread::make_future() and Thread::Future::make() functions which take a function pointer (or an object reference and member function pointer) with bound arguments but these are deprecated in the 2.2 series of the library as they offer little advantage over using std::bind. (Although deprecated, there is no plan to remove these functions as they are there and they work - the deprecation is in effect guidance.) These deprecated functions can take up to three bound arguments in the case of a non-static member function, and four bound arguments in the case of any other function. In the case of a non-static member function, the referenced object whose member function is to be called must remain in existence until the worker thread has completed. The target function passed by pointer (or member function pointer) can take a reference to const argument, as a copy of the object to be passed to the argument is taken to avoid dangling references, but it cannot take a reference to non-const argument.

It is to be noted that the target function or callable object to be represented by a Thread::Future object must not allow any exception other than Thread::Exit, an exception deriving from std::exception or a cancellation pseudo-exception to escape from it when it is executed. This includes ensuring that, for any function's bound argument which is of class type and not taken by reference, the argument's copy constructor does not throw anything other than these, and that the move assignment operator (or if none, copy assignment operator) of the return value (if of class type) of the target function or callable object does not throw anything other than these. (If the target function or callable object, or the copy constructor of a bound value argument or the move or copy assignment operator of the return value, throws Thread::Exit or an exception deriving from std::exception, the exception is safely consumed and the Thread::Future object's error flag is set. However, if the move assignment operator or copy assignment operator, as the case may be, of the return value throws, it should leave the movee/assignee in a state in which it can safely be destroyed and in which, if that movee/assignee is further copied or moved from, the copy or move either throws an exception or produces an object which can also be destroyed – but these are minimum requirements for any reasonable assignment operator, and met by any assignment operator offering the basic exception guarantee.)

The Thread::Future object will store the return value of the target function or callable object, so that it is available to the get() and move_get() methods and any 'when' callback, and therefore either move it, or if it has no move assignment operator, copy it once.

For safety reasons, the get() method returns by value and so will cause the return value to be copied once more, so for return values comprising complex class objects which are to be extracted using the get() method, it is often better if the function represented by the Thread::Future object allocates the return value on free store and returns it by pointer, by Cgu::SharedLockPtr, or by a std::shared_ptr implementation which has a thread-safe reference count. Alternatively, from version 2.0.11 a move_get() method is provided which will make a move operation instead of a copy if the return type implements a move constructor, but see the documentation on move_get() for the caveats with respect to its use: in particular, if move_get() is to be called by a thread, then get() may not normally be called by another thread, nor should the when() method be called.

It should be noted that where the when() method is used, the return value is passed to the 'when' callback by reference to const and so without the copying carried out by the get() method: therefore, if the return value has a move assignment operator and the when() method is to be employed, and the 'when' callback only needs to call const methods of the return value, it may be more efficient not to allocate the return value on free store.

This is a usage example:

std::vector<long> get_primes(int n); // calculates the first n primes
// get the first 1,000 primes
using namespace Cgu;
auto future = Thread::make_future([] () {return get_primes(1000);});
future->run();
... [ do something else ] ...
std::vector<long> result(future->move_get());
std::for_each(result.begin(), result.end(), [](long l) {std::cout << l << std::endl;});

The Cgu::Thread::Future::when() functions

The return value of the thread function represented by Cgu::Thread::Future can be obtained asynchronously using Cgu::Thread::Future::when() to execute a function in a glib main loop when the thread function completes. The above example could be reimplemented as:

std::vector<long> get_primes(int n); // calculates the first n primes
using namespace Cgu;
auto future = Thread::make_future([] () {return get_primes(1000);});
future->when([](const std::vector<long>& vec) {
for (const auto& elt: vec) {std::cout << elt << std::endl;}
});
future->run();

The Cgu::Thread::Future::fail() functions

The Thread::Future::when() functions have an associated optional Thread::Future::fail() function which causes a 'fail' callback to execute in a glib main loop in the event of certain exceptions arising in executing the thread function or a thread being cancelled (the documentation on Thread::Future::fail() gives further details). The 'fail' callback must be fully bound. Whilst a worker thread can pass error status to the 'fail' callback via shared data bound to both the thread function and the 'fail' callback (held by, say, a SharedLockPtr object), or a global error stack, 'fail' callbacks are generally best reserved either for use with entirely unexpected exceptions, where the most reasonable course is to perform some orderly logging and shutdown, or to report thread cancellation. For handlable exceptions, in an asynchronous environment the best course is often to catch them and deal with them in the thread function itself and return a value of the return type for the 'when' callback indicating no result.

Constructor & Destructor Documentation

◆ Future()

template<class Val >
Cgu::Thread::Future< Val >::Future ( const Future< Val > &  )
delete

This class cannot be copied (except by smart pointer). The copy constructor is deleted.

Member Function Documentation

◆ cancel()

template<class Val >
bool Cgu::Thread::Future< Val >::cancel ( )
noexcept

Cancels the worker thread in which the function or callable object represented by this object runs, if it has not yet finished. If this method is called and the worker thread is still running and is cancelled in response to a call to this method, then the error flag will be set so that a method calling get() or move_get() can examine whether the result is valid. If run() has not yet been called or the worker thread has already finished executing the function or callable object represented by this object then this function does nothing and returns false. This method is thread safe and may be called by any thread. It will not throw.

Returns
true if run() has previously been called and the worker thread has not yet finished executing the function or callable object represented by this object, otherwise false (in which case this method does nothing).
Note
1. Use this method with care. When cancelling a thread not all thread implementations will unwind the stack, and so run the destructors of local objects. This is discussed further in the documentation on Cgu::Thread::Thread::cancel().
2. This method might return true because the worker thread has not yet finished, but the error flag might still not be set. This is because the worker thread may not meet a cancellation point before it ends naturally. It is the error flag which indicates definitively whether the worker thread terminated prematurely in response to a call to this method.

◆ fail() [1/4]

template<class Val >
void Cgu::Thread::Future< Val >::fail ( const Cgu::Callback::Callback cb,
Cgu::Releaser r,
GMainContext *  context = 0 
)

This is a version of the fail() utility for use in conjunction with the when() methods, which takes a Releaser object for automatic disconnection of the callback functor passed as an argument to this method if the object having the callback function as a member is destroyed. For this to be race free, the lifetime of that object must be controlled by the thread in whose main loop the 'fail' callback will execute.

This method enables a callback to be executed in a glib main loop if memory is exhausted and std::bad_alloc was thrown by the thread wrapper of Cgu::Thread::Future after calling run() or by done_emitter when emitting, or if the thread function represented by this Cgu::Thread::Future object threw Cgu::Thread::Exit, exited with an uncaught exception deriving from std::exception or was cancelled (or that function took an argument of class type by value whose copy constructor threw such an exception or had a return value of class type whose move assignment operator, or if none copy assignment operator, threw such an exception), or any callback connected to done_emitter exited with an uncaught exception. It therefore enables errors to be detected and acted on without having a thread wait on the get() method in order to test is_error() or is_emitter_error().

This method can be called before or after the run() method has been called, and whether or not the thread function represented by this Cgu::Thread::Future object has completed.

The documentation for the version of this method which does not take a Releaser object gives further details of how this method is used.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
cbThe 'fail' callback (the callback to be executed if the thread function represented by this Cgu::Thread::Future object or a done_emitter emission has failed to complete). Ownership is taken of this object, and it will be deleted when it has been finished with.
rA Releaser object for automatic disconnection of the 'fail' callback before it executes in a main loop (mainly relevant if the callback represents a non-static member function of an object which may be destroyed before the callback executes).
contextThe glib main context of the thread in whose main loop the 'fail' callback is to be executed (the default of NULL will cause the functor to be executed in the main program loop).
Exceptions
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case. If it does so, the 'fail' callback will be disposed of.
Cgu::Thread::MutexErrorThis method will throw Cgu:Thread::MutexError if initialisation of the mutex in a SafeEmitterArg object constructed by Cgu::start_timeout() fails. If it does so, the 'fail' callback will be disposed of. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Note
By virtue of the Releaser object, it is in theory possible (if memory is exhausted and the system throws in that case) that an internal SafeEmitterArg object will throw std::bad_alloc when emitting/executing the 'fail' callback in the glib main loop, with the result that the relevant callback will not execute (instead the exception will be consumed and a g_critical() warning will be issued). This is rarely of any relevance because glib will abort the program if it is itself unable to obtain memory from the operating system. However, where it is relevant, design the program so that it is not necessary to provide a releaser object.

Since 2.0.2

◆ fail() [2/4]

template<class Val >
void Cgu::Thread::Future< Val >::fail ( const Cgu::Callback::Callback cb,
GMainContext *  context = 0 
)

A utility intended to be used where relevant in conjunction with the when() methods. It enables a callback to be executed in a glib main loop (referred to below as the 'fail' callback) if memory is exhausted and std::bad_alloc was thrown by the thread wrapper of Cgu::Thread::Future after calling run() or by done_emitter when emitting, or if the thread function represented by this Cgu::Thread::Future object threw Cgu::Thread::Exit, exited with an uncaught exception deriving from std::exception or was cancelled (or that function took an argument of class type by value whose copy constructor threw such an exception or had a return value of class type whose move assignment operator, or if none copy assignment operator, threw such an exception), or any callback connected to done_emitter exited with an uncaught exception. It therefore enables errors to be detected and acted on without having a thread wait on the get() method in order to test is_error() or is_emitter_error().

It is implemented by attaching a timeout to the main loop which polls at 100 millisecond intervals and tests is_done()/is_error() and is_emitter_done()/is_emitter_error(). The timeout is automatically removed by the implementation once it has been detected that an error has occurred and the 'fail' callback is executed, or if the thread function represented by this Cgu::Future object and all done_emitter emissions (including execution of any 'when' callback) have completed successfully.

This method can be called before or after the run() method has been called, and whether or not the thread function represented by this Cgu::Thread::Future object has completed.

Once this method has been called, this Cgu::Thread::Future object will always stay in existence until the timeout has been automatically removed by the implementation. Accordingly it is safe to use this method even if the intrusive pointer object returned by the make() methods will go out of scope before the 'fail' callback has executed: the callback will execute correctly irrespective of that.

This method does not have a priority argument: as a polling timeout is created, a particular priority will normally have no significance (in fact, the 'fail' callback will execute in the main loop with a priority of G_PRIORITY_DEFAULT). If in a special case a different polling interval than 100 milliseconds or a different priority is required, users can attach their own polling timeouts to a main loop and carry out the tests by hand.

Four other points should be noted. First, if as well as the when() method being called some other callback has been connected to done_emitter, and that other callback throws, the 'fail' callback will execute. Therefore, if the particular program design requires that the 'fail' callback should only execute if the 'when' callback is not executed (and the 'when' callback only execute if the 'fail' callback does not execute), no other callbacks which throw should be connected to done_emitter.

Secondly, as mentioned in the documentation on the when() method, if the 'when' callback exits with an uncaught exception upon being executed by the main loop or it represents a function which takes an argument by value whose copy constructor throws, the 'fail' callback will not execute (the exception will have been consumed internally in order to protect the main loop and a g_critical message issued).

Thirdly, avoid if possible having a 'fail' callback which might throw, or representing a function which takes an argument by value whose copy constructor might throw: such an exception would be consumed internally in order to protect the main loop and a g_critical message issued, but no other error indication apart from the g_critical message will be provided.

Fourthly, unlike the 'when' callback, a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() methods may safely be bound to the 'fail' callback, which would enable the 'fail' callback to determine whether it is is_error() or is_emitter_error() which returns false.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
cbThe 'fail' callback (the callback to be executed if the thread function represented by this Cgu::Thread::Future object or a done_emitter emission has failed to complete). Ownership is taken of this object, and it will be deleted when it has been finished with.
contextThe glib main context of the thread in whose main loop the 'fail' callback is to be executed (the default of NULL will cause the functor to be executed in the main program loop).
Exceptions
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case. If it does so, the 'fail' callback will be disposed of.

Since 2.0.2

◆ fail() [3/4]

template<class Val >
template<class Fail , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<Fail>::type, const Cgu::Callback::Callback*>::value>::type>
void Cgu::Thread::Future< Val >::fail ( Fail &&  f,
Cgu::Releaser r,
GMainContext *  context = 0 
)
inline

This is a version of the fail() utility for use in conjunction with the when() methods, which takes a Releaser object for automatic disconnection of the callable object passed as an argument to this method if the 'fail' callback represents or calls into an object which is destroyed. For this to be race free, the lifetime of the object called into must be controlled by the thread in whose main loop the 'fail' callback will execute.

This method enables a callback to be executed in a glib main loop if memory is exhausted and std::bad_alloc was thrown by the thread wrapper of Cgu::Thread::Future after calling run() or by done_emitter when emitting, or if the thread function represented by this Cgu::Thread::Future object threw Cgu::Thread::Exit, exited with an uncaught exception deriving from std::exception or was cancelled (or that function took an argument of class type by value whose copy constructor threw such an exception or had a return value of class type whose move assignment operator, or if none copy assignment operator, threw such an exception), or any callback connected to done_emitter exited with an uncaught exception. It therefore enables errors to be detected and acted on without having a thread wait on the get() method in order to test is_error() or is_emitter_error().

This method can be called before or after the run() method has been called, and whether or not the thread function represented by this Cgu::Thread::Future object has completed.

The documentation for the version of this method which does not take a Releaser object gives further details of how this method is used.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
fA callable object (such as formed by a lambda expression or the result of std::bind) representing the 'fail' callback (the callback to be executed if the thread function represented by this Cgu::Thread::Future object or a done_emitter emission has failed to complete). The callable object should be fully bound (it should take no arguments when called).
rA Releaser object for automatic disconnection of the 'fail' callback before it executes in a main loop (mainly relevant if the callback represents or calls into a non-static member function of an object which may be destroyed before the callback executes).
contextThe glib main context of the thread in whose main loop the 'fail' callback is to be executed (the default of NULL will cause the functor to be executed in the main program loop).
Exceptions
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case.
Cgu::Thread::MutexErrorThis method will throw Cgu:Thread::MutexError if initialisation of the mutex in a SafeEmitterArg object constructed by Cgu::start_timeout() fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Note
1. This method will also throw if the copy or move constructor of the 'fail' callable object throws.
2. By virtue of the Releaser object, it is in theory possible (if memory is exhausted and the system throws in that case) that an internal SafeEmitterArg object will throw std::bad_alloc when emitting/executing the 'fail' callback in the glib main loop, with the result that the relevant callback will not execute (instead the exception will be consumed and a g_critical() warning will be issued). This is rarely of any relevance because glib will abort the program if it is itself unable to obtain memory from the operating system. However, where it is relevant, design the program so that it is not necessary to provide a releaser object.

Since 2.1.0

◆ fail() [4/4]

template<class Val >
template<class Fail , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<Fail>::type, const Cgu::Callback::Callback*>::value>::type>
void Cgu::Thread::Future< Val >::fail ( Fail &&  f,
GMainContext *  context = 0 
)
inline

A utility intended to be used where relevant in conjunction with the when() methods. It enables a callable object to be executed in a glib main loop (referred to below as the 'fail' callback) if memory is exhausted and std::bad_alloc was thrown by the thread wrapper of Cgu::Thread::Future after calling run() or by done_emitter when emitting, or if the thread function represented by this Cgu::Thread::Future object threw Cgu::Thread::Exit, exited with an uncaught exception deriving from std::exception or was cancelled (or that function took an argument of class type by value whose copy constructor threw such an exception or had a return value of class type whose move assignment operator, or if none copy assignment operator, threw such an exception), or any callback connected to done_emitter exited with an uncaught exception. It therefore enables errors to be detected and acted on without having a thread wait on the get() method in order to test is_error() or is_emitter_error().

It is implemented by attaching a timeout to the main loop which polls at 100 millisecond intervals and tests is_done()/is_error() and is_emitter_done()/is_emitter_error(). The timeout is automatically removed by the implementation once it has been detected that an error has occurred and the 'fail' callback is executed, or if the thread function represented by this Cgu::Future object and all done_emitter emissions (including execution of any 'when' callback) have completed successfully.

This method can be called before or after the run() method has been called, and whether or not the thread function represented by this Cgu::Thread::Future object has completed.

Once this method has been called, this Cgu::Thread::Future object will always stay in existence until the timeout has been automatically removed by the implementation. Accordingly it is safe to use this method even if the intrusive pointer object returned by the make() methods will go out of scope before the 'fail' callback has executed: the callback will execute correctly irrespective of that.

This method does not have a priority argument: as a polling timeout is created, a particular priority will normally have no significance (in fact, the 'fail' callback will execute in the main loop with a priority of G_PRIORITY_DEFAULT). If in a special case a different polling interval than 100 milliseconds or a different priority is required, users can attach their own polling timeouts to a main loop and carry out the tests by hand.

Four other points should be noted. First, if as well as the when() method being called some other callback has been connected to done_emitter, and that other callback throws, the 'fail' callback will execute. Therefore, if the particular program design requires that the 'fail' callback should only execute if the 'when' callback is not executed (and the 'when' callback only execute if the 'fail' callback does not execute), no other callbacks which throw should be connected to done_emitter.

Secondly, as mentioned in the documentation on the when() method, if the 'when' callback exits with an uncaught exception upon being executed by the main loop or it represents a function which takes an argument by value whose copy constructor throws, the 'fail' callback will not execute (the exception will have been consumed internally in order to protect the main loop and a g_critical message issued).

Thirdly, avoid if possible having a 'fail' callback which might throw: such an exception would be consumed internally in order to protect the main loop and a g_critical message issued, but no other error indication apart from the g_critical message will be provided.

Fourthly, unlike the 'when' callback, a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() methods may safely be bound to the 'fail' callback, which would enable the 'fail' callback to determine whether it is is_error() or is_emitter_error() which returns false.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
fA callable object (such as formed by a lambda expression or the result of std::bind) representing the 'fail' callback (the callback to be executed if the thread function represented by this Cgu::Thread::Future object or a done_emitter emission has failed to complete). The callable object should be fully bound (it should take no arguments when called).
contextThe glib main context of the thread in whose main loop the 'fail' callback is to be executed (the default of NULL will cause the functor to be executed in the main program loop).
Exceptions
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case.
Note
This method will also throw if the copy or move constructor of the 'fail' callable object throws.

Since 2.1.0

◆ get()

template<class Val >
Val Cgu::Thread::Future< Val >::get ( )

Gets the stored value obtained from the function or callable object which is represented by this object. If the worker thread launched by the call to run() has not completed, then this method will block until it has completed. If run() has not been called, then run() will be called (and this method will block until the launched worker thread completes). If the function or callable object which is represented by this Cgu::Thread::Future object throws Cgu::Thread::Exit or an uncaught exception derived from std::exception, or if any of those exceptions are thrown either by the copy constructor of an argument taken by value by that function or object, or by the move assignment operator (or if none, copy assignment operator) of the return value of that function or object, then the exception will have been consumed by this Cgu::Thread::Future object and the error flag will have been set. The error flag will also have been set if the worker thread is cancelled or the thread wrapper in this Cgu::Thread::Future object threw std::bad_alloc. On the error flag being set, this method will unblock and return a default constructed object of the return type. This method is thread safe and may be called by any thread (and by more than one thread). It is a cancellation point if it blocks, and from version 2.0.11 is cancellation safe if the stack unwinds on cancellation. It is also strongly exception safe: no data will be lost if extracting the value fails.

Returns
The value obtained from the function which is represented by this object, or a default constructed object of the return type if the error flag has been set.
Exceptions
Cgu::Thread::FutureThreadErrorThis method might throw Cgu::Thread::FutureThreadError if run() has not previously been called and the thread did not start properly when this function called run().
std::bad_allocThis method might throw std::bad_alloc if run() has not previously been called, memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Note
1. This method might also throw if the copy constructor of the returned value type throws.
2. This method calls Thread::Cond::wait(). Between versions 2.2.3 and 2.2.13 inclusive, Thread::Cond::wait() was marked 'noexcept'. This was a mistake because it prevented a thread being cancelled while in a wait, including in this method (the cancellation pseudo-exception conflicted with the noexcept specifier). This was fixed in version 2.2.14.
3. Question: Couldn't this method return the stored value by lvalue reference to const? Answer: It could. However, because of return value optimization, which will be implemented by any compiler capable of compiling this library, no advantage would be gained by doing so when initializing a local variable with the return value of this method (the copy constructor will only be called once whether returning by value or const reference). The advantage of returning by value is that the call to the copy constructor is forced to be within this Thread::Future object's mutex, so different threads' calls to the copy constructor are serialized, and also with blocked cancellation, so this method is cancellation safe. All calls to this method by different threads are therefore isolated and we do not have to worry about the thread safety of direct access to the stored value via its const methods outside the mutex (which would not be thread safe if the stored value has data members declared mutable) nor about the cancellation safety of the copy constructor. Of course, for objects which do not have mutable data, a hit arises by returning by value in cases where it is not intended to initialize a local variable at all nor to cancel a thread: where, say, only const methods are to be called on the return value (which could be done directly if this method returned by const reference). However, in many use cases this will be mitigated by the move_get() method.

◆ is_done()

template<class Val >
bool Cgu::Thread::Future< Val >::is_done ( ) const
noexcept
Returns
true if the function or callable object represented by this Cgu::Thread::Future object has finished, either by returning normally, by cancellation or by virtue of having thrown Cgu::Thread::Exit or some exception derived from std::exception. Once this method returns true, then it is guaranteed that the get() or move_get() method will not block (except as incidental to any contention between threads calling get()). Once this method has returned true or get() or move_get() has unblocked, then the result of is_error() is definitive. This method is thread safe and may be called by any thread. It will not throw.
Note
This method will return true even though any callbacks connected to done_emitter are still executing or waiting to execute. From version 2.0.2 the is_emitter_done() method will indicate when done_emitter callbacks (if any) have also completed.

◆ is_emitter_done()

template<class Val >
bool Cgu::Thread::Future< Val >::is_emitter_done ( ) const
noexcept
Returns
true if both the function or callable object represented by this Cgu::Thread::Future object has finished and any callbacks connected to done_emitter have completed. Once this method returns true, then the result of is_emitter_error() is definitive. This method is thread safe and may be called by any thread. It will not throw.
Note
This method will return true automatically if is_error() and is_done() return true, because if the function or callable object represented by this Cgu::Thread::Future object was cancelled or exited with an uncaught exception, done_emitter is never emitted. In addition, if this method returns true, then is_done() must also return true.

Since 2.0.2

◆ is_emitter_error()

template<class Val >
bool Cgu::Thread::Future< Val >::is_emitter_error ( ) const
noexcept
Returns
true if an uncaught exception arose in emitting done_emitter when executing callbacks connected to it. Otherwise this method returns false. The result of this method is definitive once is_emitter_done() returns true. This method is thread safe and may be called by any thread. It will not throw.
Note
This method will return false automatically if is_error() returns true, because if the function or callable object represented by this Cgu::Thread::Future object was cancelled or exited with an uncaught exception, done_emitter is never emitted. It follows that if this method returns true, is_error() must return false.

◆ is_error()

template<class Val >
bool Cgu::Thread::Future< Val >::is_error ( ) const
noexcept
Returns
true if (a) a Cgu::Thread::Exit exception has been thrown by the function or callable object represented by this Cgu::Thread::Future object (which will have been consumed by this Cgu::Thread::Future object), (b) an exception derived from std::exception has been thrown on invoking that function or object which was not caught by it (which will have been consumed by this Cgu::Thread::Future object), (c) any of those exceptions have been thrown either by the copy constructor of an argument taken by value by that function or object, or by the move assignment operator (or if none, copy assignment operator) of the return value of that function or object (which will have been consumed by this Cgu::Thread::Future object), (d) the worker thread in which that function or callable object executes was cancelled in mid-course with a call to cancel() or (e) the thread wrapper implementing the worker thread in this Cgu::Thread::Future object threw and then consumed std::bad_alloc (this is different from the run() method throwing std::bad_alloc). In these cases the value obtained by get() or move_get() will not be valid (it will be a default constructed object of the return type of the function represented by this Cgu::Thread::Future object). Otherwise this method returns false. The result of this method is definitive once get() or move_get() has unblocked or is_done() returns true. This method is thread safe and may be called by any thread. It will not throw.

◆ make() [1/14]

template<class Val >
template<class Ret , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( const T &  t,
Ret(T::*)() const  func 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the default constructor of the return value type throws.

◆ make() [2/14]

template<class Val >
template<class Ret , class Param1 , class Arg1 , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( const T &  t,
Ret(T::*)(Param1) const  func,
Arg1 &&  arg1 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of the bound argument throws, or the default constructor of the return value type throws.

◆ make() [3/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Arg1 , class Arg2 , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( const T &  t,
Ret(T::*)(Param1, Param2) const  func,
Arg1 &&  arg1,
Arg2 &&  arg2 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ make() [4/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Param3 , class Arg1 , class Arg2 , class Arg3 , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( const T &  t,
Ret(T::*)(Param1, Param2, Param3) const  func,
Arg1 &&  arg1,
Arg2 &&  arg2,
Arg3 &&  arg3 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ make() [5/14]

template<class Val >
template<class Func >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( Func &&  func)
static

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. It will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Parameters
funcA callable object, such as formed by a lambda expression or the result of std::bind. It must be fully bound (that is, its must take no arguments when called). It should return the Val type.
Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
1. This method will also throw if the copy or move constructor of the callable object passed as an argument throws, or the default constructor of the return value type throws.
2. If the callable object passed as an argument has both const and non-const operator()() methods, the non-const version will be called even if the callable object passed is a const object.

Since 2.0.14

◆ make() [6/14]

template<class Val >
template<class Ret >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( Ret(*)()  func)
static

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the default constructor of the return value type throws.

◆ make() [7/14]

template<class Val >
template<class Ret , class Param1 , class Arg1 >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( Ret(*)(Param1)  func,
Arg1 &&  arg1 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of the bound argument throws, or the default constructor of the return value type throws.

◆ make() [8/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Arg1 , class Arg2 >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( Ret(*)(Param1, Param2)  func,
Arg1 &&  arg1,
Arg2 &&  arg2 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ make() [9/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Param3 , class Arg1 , class Arg2 , class Arg3 >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( Ret(*)(Param1, Param2, Param3)  func,
Arg1 &&  arg1,
Arg2 &&  arg2,
Arg3 &&  arg3 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ make() [10/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Param3 , class Param4 , class Arg1 , class Arg2 , class Arg3 , class Arg4 >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( Ret(*)(Param1, Param2, Param3, Param4)  func,
Arg1 &&  arg1,
Arg2 &&  arg2,
Arg3 &&  arg3,
Arg4 &&  arg4 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ make() [11/14]

template<class Val >
template<class Ret , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( T &  t,
Ret(T::*)()  func 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the default constructor of the return value type throws.

◆ make() [12/14]

template<class Val >
template<class Ret , class Param1 , class Arg1 , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( T &  t,
Ret(T::*)(Param1)  func,
Arg1 &&  arg1 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of the bound argument throws, or the default constructor of the return value type throws.

◆ make() [13/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Arg1 , class Arg2 , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( T &  t,
Ret(T::*)(Param1, Param2)  func,
Arg1 &&  arg1,
Arg2 &&  arg2 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ make() [14/14]

template<class Val >
template<class Ret , class Param1 , class Param2 , class Param3 , class Arg1 , class Arg2 , class Arg3 , class T >
static Cgu::IntrusivePtr<Cgu::Thread::Future<Val> > Cgu::Thread::Future< Val >::make ( T &  t,
Ret(T::*)(Param1, Param2, Param3)  func,
Arg1 &&  arg1,
Arg2 &&  arg2,
Arg3 &&  arg3 
)
static
Deprecated:

DEPRECATED. Use the version of Future::make() which takes a callable object.

Constructs a new Cgu::Thread::Future object (returned by Cgu::IntrusivePtr<Cgu::Thread::Future<Val>>). The type parameter Val represents the return value of the function to be represented by the new object. From version 2.0.4, it will usually be more convenient to call the Cgu::Thread::make_future() function, which is a convenience wrapper for this static method.

Exceptions
std::bad_allocIt might throw std::bad_alloc if memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Cgu::Thread::MutexErrorIt might throw Cgu::Thread::MutexError if initialisation of the contained mutex fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Cgu::Thread::CondErrorIt might throw Cgu::Thread::CondError if initialisation of the contained condition variable fails. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new condition variables.)
Note
This method will also throw if the copy or move constructor of a bound argument throws, or the default constructor of the return value type throws.

◆ move_get()

template<class Val >
Val Cgu::Thread::Future< Val >::move_get ( )

Gets the stored value obtained from the function or callable object which is represented by this object by a move operation, if the return type implements a move constructor (otherwise this method does the same as the get() method). It is provided as an option for cases where a move is required for efficiency reasons, but although it may be called by any thread, a move from this Thread::Future object may normally only be made once (except where the return type has been designed to be moved more than once for the limited purpose of inspecting a flag indicating whether its value is valid or not). If this method is to be called then no calls to get() by another thread should normally be made and no calls to when() should be made. If the worker thread launched by the call to run() has not completed, then this method will block until it has completed. If run() has not been called, then run() will be called (and this method will block until the launched worker thread completes). If the function or callable object which is represented by this Cgu::Thread::Future object throws Cgu::Thread::Exit or an uncaught exception derived from std::exception, or if any of those exceptions are thrown either by the copy constructor of an argument taken by value by that function or object, or by the move assignment operator (or if none, copy assignment operator) of the return value of that function or object, then the exception will have been consumed by this Cgu::Thread::Future object and the error flag will have been set. The error flag will also have been set if the worker thread is cancelled or the thread wrapper in this Cgu::Thread::Future object threw std::bad_alloc. On the error flag being set, this method will unblock and return a default constructed object of the return type. This method is a cancellation point if it blocks, and is cancellation safe if the stack unwinds on cancellation. This method is only exception safe if the return type's move constructor is exception safe.

Returns
The value obtained from the function which is represented by this object, or a default constructed object of the return type if the error flag has been set.
Exceptions
Cgu::Thread::FutureThreadErrorThis method might throw Cgu::Thread::FutureThreadError if run() has not previously been called and the thread did not start properly when this function called run().
std::bad_allocThis method might throw std::bad_alloc if run() has not previously been called, memory is exhausted and the system throws in that case. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Note
1. This method might also throw if the copy or move constructor of the returned value type throws.
2. This method calls Thread::Cond::wait(). Between versions 2.2.3 and 2.2.13 inclusive, Thread::Cond::wait() was marked 'noexcept'. This was a mistake because it prevented a thread being cancelled while in a wait, including in this method (the cancellation pseudo-exception conflicted with the noexcept specifier). This was fixed in version 2.2.14.
3. Question: Couldn't this method return the stored value by rvalue reference? Answer: It could. However, because of return value optimization, which will be implemented by any compiler capable of compiling this library, no advantage would be gained by doing so when initializing a local variable with the return value of this method (the move constructor will only be called once, and no call will be made to the copy constructor, whether returning by value or rvalue reference). The advantage of returning by value is that the call to the move constructor is forced to be within this Thread::Future object's mutex, so different threads' calls to the move constructor are serialized, and also with blocked cancellation, so this method is cancellation safe. All calls to this method by different threads are therefore isolated and we do not have to worry about the thread safety of the mutating first call to this method, nor about direct access to the stored value via a rvalue reference outside the mutex nor the cancellation safety of the move constructor.

Since 2.0.11

◆ operator=()

template<class Val >
Future& Cgu::Thread::Future< Val >::operator= ( const Future< Val > &  )
delete

This class cannot be copied (except by smart pointer). The assignment operator is deleted.

◆ run()

template<class Val >
bool Cgu::Thread::Future< Val >::run ( )

Runs the function or callable object represented by this Cgu::Thread::Future object in a new worker thread. That function will only be run once. If this is the first time this method has been called, it will start the worker thread and return true, and if it has previously been called, this method will do nothing and return false. This method will not wait for the worker thread to complete before returning. This method is thread safe and may be called by a different thread from the one which called make().

Returns
true if this is the first time this method has been called, or false if this method has previously been called.
Exceptions
Cgu::Thread::FutureThreadErrorThis method might throw Cgu::Thread::FutureThreadError if it has not previously been called and the thread did not start properly. If it does throw, this Cgu::Thread::Future object is defunct and further attempts to call this method will return immediately with a false value. (It is often not worth checking for this exception, as it means either memory is exhausted, the pthread thread limit has been reached or pthread has run out of other resources to start new threads.)
std::bad_allocThis method might throw std::bad_alloc if it has not previously been called, and memory is exhausted and the system throws in that case. If it does throw, this Cgu::Thread::Future object is defunct and further attempts to call this method will return immediately with a false value. (This exception will not be thrown if the library has been installed using the --with-glib-memory-slices-no-compat configuration option: instead glib will terminate the program if it is unable to obtain memory from the operating system.)
Note
1. Any Cgu::Thread::Exit exception, or any uncaught exception derived from std::exception, which is thrown from the worker thread will be caught and consumed and the error flag will be set. The worker thread will safely terminate and unwind the stack in so doing.
2. As this wrapper class can provide error reporting in a way that Cgu::Thread::Thread of itself cannot, it would be desirable to consume any other uncaught exceptions. However, this cannot be done: annoyingly, NPTL's forced stack unwinding does not allow this if thread cancellation is to be made available. If an uncaught exception propagates out of a thread when the thread exits, the C++11/14 standard will cause std::terminate() to be called and so terminate the entire program. Accordingly, a user must make sure that no exceptions, other than Cgu::Thread::Exit or those derived from std::exception or any cancellation pseudo-exception, can propagate from the function which this Cgu::Thread::Future object represents, nor from the copy constructor of any argument type that that function takes by value nor from the move assignment operator (or if none, copy assignment operator) of the return value of that function.
3. If the worker thread is cancelled by a call to cancel() while in the middle of executing the function which this Cgu::Thread::Future object represents, the error flag will be set.

◆ when() [1/4]

template<class Val >
Cgu::Callback::SafeFunctor Cgu::Thread::Future< Val >::when ( const Cgu::Callback::CallbackArg< const Val & > *  cb,
Cgu::Releaser r,
gint  priority = G_PRIORITY_DEFAULT,
GMainContext *  context = 0 
)

This is a version of the utility enabling the value returned by the thread function represented by this Cgu::Thread::Future object to be dealt with asynchronously, which takes a Releaser object for automatic disconnection of the callback passed as an argument to this method (referred to below as the 'when' callback), if the object having the 'when' callback function as a member is destroyed. For this to be race free, the lifetime of that object must be controlled by the thread in whose main loop the 'when' callback will execute.

If the 'when' callback has not been released, this method causes it to be executed by a thread's main loop if and when the thread function represented by this Cgu::Thread::Future object finishes correctly - the 'when' callback is passed that thread function's return value when it is invoked. This method is thread safe, and may be called by any thread.

This functionality is implemented by connecting an internal dispatching callback to the done_emitter object.

The 'when' callback should take a single unbound argument comprising a const reference to the return type of the thread function represented by this Cgu::Thread::Future object. (So, in the case of a Future<int> object, the callback function should take a const int& argument as the unbound argument.) The 'when' callback can have any number of bound arguments, except that a bound argument may not include a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() methods (that would result in this Cgu::Thread::Future object owning, via done_emitter, a reference to itself and so become incapable of being freed). The 'when' callback may, however, take a pointer to this Cgu::Thread::Future object, as obtained by the Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future object is guaranteed to remain in existence until the callback has completed executing.

This method cannot be called after the thread function represented by this Cgu::Thread::Future object has completed (either successfully or unsuccessfully) so that is_done() would return true, and if this is attempted a Cgu::Thread::FutureWhenError exception will be thrown. Therefore, generally this method should be called before the run() method has been called.

The documentation for the version of this method which does not take a Releaser object gives further details of how this method is used.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
cbThe 'when' callback (the callback to be executed when the function represented by this Cgu::Thread::Future object has successfully completed). Ownership is taken of this object, and it will be deleted when it has been finished with.
rA Releaser object for automatic disconnection of the 'when' callback before it executes in a main loop (mainly relevant if the callback represents a non-static member function of an object which may be destroyed before the callback executes).
priorityThe priority to be given to the 'when' callback in the main loop after the thread function represented by this Cgu::Thread::Future object has successfully completed. In ascending order of priorities, priorities are G_PRIORITY_LOW, G_PRIORITY_DEFAULT_IDLE, G_PRIORITY_HIGH_IDLE, G_PRIORITY_DEFAULT and G_PRIORITY_HIGH. The default is G_PRIORITY_DEFAULT. This determines the order in which the callback will appear in the event list in the main loop, not the priority which the OS will adopt.
contextThe glib main context of the thread in whose main loop the 'when' callback is to be executed (the default of NULL will cause the callback to be executed in the main program loop).
Returns
The internal dispatching callback created by this method and connected to done_emitter. It is made available as a return value so that if wanted it can be disconnected programmatically from done_emitter, or block()/unblock() can be called on it (but if that is to be done, it must be done before the thread function represented by this Cgu::Thread::Future object has completed in order for it to be effective).
Exceptions
Cgu::Thread::FutureWhenErrorThis method will throw Cgu::Thread::FutureWhenError if it is called after the thread function represented by this Cgu::Thread::Future object has completed. If it does so, the 'when' callback will be disposed of.
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case. If it does so, the 'when' callback will be disposed of.
Cgu::Thread::MutexErrorThis method will throw Cgu:Thread::MutexError if initialisation of the mutex in a SafeEmitterArg object constructed by this method fails. If it does so, the 'when' callback will be disposed of. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Note
1. The return value of the function represented by this Cgu::Thread::Future object is stored and passed as an argument to the 'when' callback by const reference. If other threads might concurrently call this object's get() method, which copies the stored value, the stored type's copy constructor must be thread safe with respect to the stored type's const methods. This would be relevant if the stored type has data members declared mutable which would be copied by its copy constructor.
2. By virtue of the Releaser object, it is in theory possible (if memory is exhausted and the system throws in that case) that an internal SafeEmitterArg object will throw std::bad_alloc when emitting/executing the 'when' callback in the glib main loop, with the result that the relevant callback will not execute (instead the exception will be consumed and a g_critical() warning will be issued). This is rarely of any relevance because glib will abort the program if it is itself unable to obtain memory from the operating system. However, where it is relevant, design the program so that it is not necessary to provide a releaser object.

Since 2.0.2

◆ when() [2/4]

template<class Val >
Cgu::Callback::SafeFunctor Cgu::Thread::Future< Val >::when ( const Cgu::Callback::CallbackArg< const Val & > *  cb,
gint  priority = G_PRIORITY_DEFAULT,
GMainContext *  context = 0 
)

A utility enabling the value returned by the thread function represented by this Cgu::Thread::Future object to be dealt with asynchronously rather than by (or in addition to) a call to the get() method. It causes the callback passed as an argument to this method (referred to below as the 'when' callback) to be executed by a thread's main loop if and when the thread function represented by this Cgu::Thread::Future object finishes correctly - the 'when' callback is passed that thread function's return value when it is invoked. This method is thread safe, and may be called by any thread.

This functionality is implemented by connecting an internal dispatching callback to the done_emitter object.

The 'when' callback should take a single unbound argument comprising a const reference to the return type of the thread function represented by this Cgu::Thread::Future object. (So, in the case of a Future<int> object, the callback function should take a const int& argument as the unbound argument.) The 'when' callback can have any number of bound arguments, except that a bound argument may not include a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() methods (that would result in this Cgu::Thread::Future object owning, via done_emitter, a reference to itself and so become incapable of being freed). The 'when' callback may, however, take a pointer to this Cgu::Thread::Future object, as obtained by the Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future object is guaranteed to remain in existence until the callback has completed executing.

This method cannot be called after the thread function represented by this Cgu::Thread::Future object has completed (either successfully or unsuccessfully) so that is_done() would return true, and if this is attempted a Cgu::Thread::FutureWhenError exception will be thrown. Therefore, generally this method should be called before the run() method has been called.

Once the run() method has been called, this Cgu::Thread::Future object will always stay in existence until the thread function represented by it has completed (whether correctly, by cancellation or by a thrown exception), and any 'when' callback (and any other callbacks connected to the done_emitter object) and any 'fail' callback have completed. Accordingly it is safe to use this method even if the intrusive pointer object returned by the make() methods will go out of scope before the 'when' callback has executed: the callback will execute correctly irrespective of that.

Summary: use of this method is safe and has been implemented in a way which does not give rise to timing issues.

If memory is exhausted and std::bad_alloc is thrown by the thread wrapper of Cgu::Thread::Future after run() is called or by done_emitter when emitting, or if the thread function represented by this Cgu::Thread::Future object throws Cgu::Thread::Exit, is cancelled, exits with an uncaught exception deriving from std::exception, takes an argument by value whose copy constructor throws such an exception or has a return value whose move assignment operator (or if none, copy assignment operator) throws such an exception, or if the 'when' callback represents a function taking a non-reference argument whose copy constructor throws an exception, or if any other callback has been connected to done_emitter before this method is called which exits with an uncaught exception, then the 'when' callback will not execute (instead the exception concerned will be consumed and an error indicated). With many systems, swap memory combined with memory over-commit makes it pointless to check for std::bad_alloc (and even more so in programs using glib, as glib aborts a program where it cannot obtain memory from the operating system). So subject to that, if the user program is designed so that the thread function represented by this Cgu::Thread::Future object does not exit with uncaught exceptions, does not take an argument by value which throws, does not have a return value whose move assignment operator (or if none, copy assignment operator) throws, does not throw Cgu::Thread::Exit and is not cancelled, and so that the 'when' callback does not exit with an uncaught exception (and a function represented by that callback either takes no arguments of class type by value or the copy constructors of any of its value arguments do not throw), and if this method is called before any other callbacks are connected to done_emitter, the possibility of failure can be disregarded.

In cases where that is not true and detecting whether a failure has occurred is required, a fail() method is provided. It should be noted that a callback handed to the fail() method will not execute in a case of error if the error comprises the 'when' callback exiting with an uncaught exception when it is executed by the main loop, or the copy constructor of any value argument of a function represented by the 'when' callback throwing (such exceptions would be consumed internally in order to protect the main loop and a g_critical message issued). If the 'when' callback might exit with an uncaught exception when executing or have the copy constructor of a value argument throw, and doing something other than consuming the exception and issuing a g_critical message is required, then a different approach is to start a new thread to wait on the get() method which can act on the result of is_error() directly.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
cbThe 'when' callback (the callback to be executed when the function represented by this Cgu::Thread::Future object has successfully completed). Ownership is taken of this object, and it will be deleted when it has been finished with.
priorityThe priority to be given to the 'when' callback in the main loop after the thread function represented by this Cgu::Thread::Future object has successfully completed. In ascending order of priorities, priorities are G_PRIORITY_LOW, G_PRIORITY_DEFAULT_IDLE, G_PRIORITY_HIGH_IDLE, G_PRIORITY_DEFAULT and G_PRIORITY_HIGH. The default is G_PRIORITY_DEFAULT. This determines the order in which the callback will appear in the event list in the main loop, not the priority which the OS will adopt.
contextThe glib main context of the thread in whose main loop the 'when' callback is to be executed (the default of NULL will cause the callback to be executed in the main program loop).
Returns
The internal dispatching callback created by this method and connected to done_emitter. It is made available as a return value so that if wanted it can be disconnected programmatically from done_emitter, or block()/unblock() can be called on it (but if that is to be done, it must be done before the thread function represented by this Cgu::Thread::Future object has completed in order for it to be effective).
Exceptions
Cgu::Thread::FutureWhenErrorThis method will throw Cgu::Thread::FutureWhenError if it is called after the thread function represented by this Cgu::Thread::Future object has completed. If it does so, the 'when' callback will be disposed of.
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case. If it does so, the 'when' callback will be disposed of.
Note
The return value of the function represented by this Cgu::Thread::Future object is stored and passed as an argument to the 'when' callback by const reference. If other threads might concurrently call this object's get() method, which copies the stored value, the stored type's copy constructor must be thread safe with respect to the stored type's const methods. This would be relevant if the stored type has data members declared mutable which would be copied by its copy constructor.

Since 2.0.2

◆ when() [3/4]

template<class Val >
template<class When , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<When>::type, const Cgu::Callback::CallbackArg<const Val&>*>::value>::type>
Cgu::Callback::SafeFunctor Cgu::Thread::Future< Val >::when ( When &&  w,
Cgu::Releaser r,
gint  priority = G_PRIORITY_DEFAULT,
GMainContext *  context = 0 
)
inline

This is a version of the utility enabling the value returned by the thread function represented by this Cgu::Thread::Future object to be dealt with asynchronously, which takes a Releaser object for automatic disconnection of the callable object passed as an argument to this method (referred to below as the 'when' callback), if the 'when' callback represents or calls into an object which is destroyed. For this to be race free, the lifetime of the object called into must be controlled by the thread in whose main loop the 'when' callback will execute.

If the 'when' callback has not been released, this method causes it to be executed by a thread's main loop if and when the thread function represented by this Cgu::Thread::Future object finishes correctly - the 'when' callback is passed that thread function's return value when it is invoked. This method is thread safe, and may be called by any thread.

This functionality is implemented by connecting an internal dispatching callback to the done_emitter object.

The 'when' callback should take a single unbound argument comprising a const reference to the return type of the thread function represented by this Cgu::Thread::Future object. (So, in the case of a Future<int> object, the callback function should take a const int& argument as the unbound argument.) The 'when' callback can have any number of bound arguments, except that a bound argument may not include a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() methods (that would result in this Cgu::Thread::Future object owning, via done_emitter, a reference to itself and so become incapable of being freed). The 'when' callback may, however, take a pointer to this Cgu::Thread::Future object, as obtained by the Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future object is guaranteed to remain in existence until the callback has completed executing.

This method cannot be called after the thread function represented by this Cgu::Thread::Future object has completed (either successfully or unsuccessfully) so that is_done() would return true, and if this is attempted a Cgu::Thread::FutureWhenError exception will be thrown. Therefore, generally this method should be called before the run() method has been called.

The documentation for the version of this method which does not take a Releaser object gives further details of how this method is used.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
wA callable object (such as formed by a lambda expression or the result of std::bind) representing the 'when' callback (the callback to be executed when the function represented by this Cgu::Thread::Future object has successfully completed). It should take a single unbound argument, namely a reference to const to the return type of the thread function represented by this Cgu::Thread::Future object.
rA Releaser object for automatic disconnection of the 'when' callback before it executes in a main loop (mainly relevant if the callback represents or calls into a non-static member function of an object which may be destroyed before the callback executes).
priorityThe priority to be given to the 'when' callback in the main loop after the thread function represented by this Cgu::Thread::Future object has successfully completed. In ascending order of priorities, priorities are G_PRIORITY_LOW, G_PRIORITY_DEFAULT_IDLE, G_PRIORITY_HIGH_IDLE, G_PRIORITY_DEFAULT and G_PRIORITY_HIGH. The default is G_PRIORITY_DEFAULT. This determines the order in which the callback will appear in the event list in the main loop, not the priority which the OS will adopt.
contextThe glib main context of the thread in whose main loop the 'when' callback is to be executed (the default of NULL will cause the callback to be executed in the main program loop).
Returns
The internal dispatching callback created by this method and connected to done_emitter. It is made available as a return value so that if wanted it can be disconnected programmatically from done_emitter, or block()/unblock() can be called on it (but if that is to be done, it must be done before the thread function represented by this Cgu::Thread::Future object has completed in order for it to be effective).
Exceptions
Cgu::Thread::FutureWhenErrorThis method will throw Cgu::Thread::FutureWhenError if it is called after the thread function represented by this Cgu::Thread::Future object has completed.
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case.
Cgu::Thread::MutexErrorThis method will throw Cgu:Thread::MutexError if initialisation of the mutex in a SafeEmitterArg object constructed by this method fails. If it does so, the 'when' callback will be disposed of. (It is often not worth checking for this, as it means either memory is exhausted or pthread has run out of other resources to create new mutexes.)
Note
1. This method will also throw if the copy or move constructor of the 'when' callable object throws.
2. The return value of the function represented by this Cgu::Thread::Future object is stored and passed as an argument to the 'when' callback by const reference. If other threads might concurrently call this object's get() method, which copies the stored value, the stored type's copy constructor must be thread safe with respect to the stored type's const methods. This would be relevant if the stored type has data members declared mutable which would be copied by its copy constructor.
3. By virtue of the Releaser object, it is in theory possible (if memory is exhausted and the system throws in that case) that an internal SafeEmitterArg object will throw std::bad_alloc when emitting/executing the 'when' callback in the glib main loop, with the result that the relevant callback will not execute (instead the exception will be consumed and a g_critical() warning will be issued). This is rarely of any relevance because glib will abort the program if it is itself unable to obtain memory from the operating system. However, where it is relevant, design the program so that it is not necessary to provide a releaser object.

Since 2.1.0

◆ when() [4/4]

template<class Val >
template<class When , class = typename std::enable_if<!std::is_convertible<typename std::remove_reference<When>::type, const Cgu::Callback::CallbackArg<const Val&>*>::value>::type>
Cgu::Callback::SafeFunctor Cgu::Thread::Future< Val >::when ( When &&  w,
gint  priority = G_PRIORITY_DEFAULT,
GMainContext *  context = 0 
)
inline

A utility enabling the value returned by the thread function represented by this Cgu::Thread::Future object to be dealt with asynchronously rather than by (or in addition to) a call to the get() method. It causes the callable object passed as an argument to this method (referred to below as the 'when' callback) to be executed by a thread's main loop if and when the thread function represented by this Cgu::Thread::Future object finishes correctly - the 'when' callback is passed that thread function's return value when it is invoked. This method is thread safe, and may be called by any thread.

This functionality is implemented by connecting an internal dispatching callback to the done_emitter object.

The 'when' callback should take a single unbound argument comprising a const reference to the return type of the thread function represented by this Cgu::Thread::Future object. (So, in the case of a Future<int> object, the callback function should take a const int& argument as the unbound argument.) The 'when' callback can have any number of bound arguments, except that a bound argument may not include a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() methods (that would result in this Cgu::Thread::Future object owning, via done_emitter, a reference to itself and so become incapable of being freed). The 'when' callback may, however, take a pointer to this Cgu::Thread::Future object, as obtained by the Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future object is guaranteed to remain in existence until the callback has completed executing.

This method cannot be called after the thread function represented by this Cgu::Thread::Future object has completed (either successfully or unsuccessfully) so that is_done() would return true, and if this is attempted a Cgu::Thread::FutureWhenError exception will be thrown. Therefore, generally this method should be called before the run() method has been called.

Once the run() method has been called, this Cgu::Thread::Future object will always stay in existence until the thread function represented by it has completed (whether correctly, by cancellation or by a thrown exception), and any 'when' callback (and any other callbacks connected to the done_emitter object) and any 'fail' callback have completed. Accordingly it is safe to use this method even if the intrusive pointer object returned by the make() methods will go out of scope before the 'when' callback has executed: the callback will execute correctly irrespective of that.

Summary: use of this method is safe and has been implemented in a way which does not give rise to timing issues.

If memory is exhausted and std::bad_alloc is thrown by the thread wrapper of Cgu::Thread::Future after run() is called or by done_emitter when emitting, or if the thread function represented by this Cgu::Thread::Future object throws Cgu::Thread::Exit, is cancelled, exits with an uncaught exception deriving from std::exception, takes an argument by value whose copy constructor throws such an exception or has a return value whose move assignment operator (or if none, copy assignment operator) throws such an exception, or if any other callback has been connected to done_emitter before this method is called which exits with an uncaught exception, then the 'when' callback will not execute (instead the exception concerned will be consumed and an error indicated). With many systems, swap memory combined with memory over-commit makes it pointless to check for std::bad_alloc (and even more so in programs using glib, as glib aborts a program where it cannot obtain memory from the operating system). So subject to that, if the user program is designed so that the thread function represented by this Cgu::Thread::Future object does not exit with uncaught exceptions, does not take an argument by value which throws, does not have a return value whose move assignment operator (or if none, copy assignment operator) throws, does not throw Cgu::Thread::Exit and is not cancelled, and so that the 'when' callback does not exit with an uncaught exception, and if this method is called before any other callbacks are connected to done_emitter, the possibility of failure can be disregarded.

In cases where that is not true and detecting whether a failure has occurred is required, a fail() method is provided. It should be noted that a callback handed to the fail() method will not execute in a case of error if the error comprises the 'when' callback exiting with an uncaught exception when it is executed by the main loop (such exceptions would be consumed internally in order to protect the main loop and a g_critical message issued). If the 'when' callback might exit with an uncaught exception when executing, and doing something other than consuming the exception and issuing a g_critical message is required, then a different approach is to start a new thread to wait on the get() method which can act on the result of is_error() directly.

If glib < 2.32 is used, the glib main loop must have been made thread-safe by a call to g_thread_init() before this function is called. glib >= 2.32 does not require g_thread_init() to be called in order to be thread safe.

Parameters
wA callable object (such as formed by a lambda expression or the result of std::bind) representing the 'when' callback (the callback to be executed when the function represented by this Cgu::Thread::Future object has successfully completed). It should take a single unbound argument, namely a reference to const to the return type of the thread function represented by this Cgu::Thread::Future object.
priorityThe priority to be given to the 'when' callback in the main loop after the thread function represented by this Cgu::Thread::Future object has successfully completed. In ascending order of priorities, priorities are G_PRIORITY_LOW, G_PRIORITY_DEFAULT_IDLE, G_PRIORITY_HIGH_IDLE, G_PRIORITY_DEFAULT and G_PRIORITY_HIGH. The default is G_PRIORITY_DEFAULT. This determines the order in which the callback will appear in the event list in the main loop, not the priority which the OS will adopt.
contextThe glib main context of the thread in whose main loop the 'when' callback is to be executed (the default of NULL will cause the callback to be executed in the main program loop).
Returns
The internal dispatching callback created by this method and connected to done_emitter. It is made available as a return value so that if wanted it can be disconnected programmatically from done_emitter, or block()/unblock() can be called on it (but if that is to be done, it must be done before the thread function represented by this Cgu::Thread::Future object has completed in order for it to be effective).
Exceptions
Cgu::Thread::FutureWhenErrorThis method will throw Cgu::Thread::FutureWhenError if it is called after the thread function represented by this Cgu::Thread::Future object has completed.
std::bad_allocThis method might throw std::bad_alloc if memory is exhausted and the system throws in that case.
Note
1. This method will also throw if the copy or move constructor of the 'when' callable object throws.
2. The return value of the function represented by this Cgu::Thread::Future object is stored and passed as an argument to the 'when' callback by const reference. If other threads might concurrently call this object's get() method, which copies the stored value, the stored type's copy constructor must be thread safe with respect to the stored type's const methods. This would be relevant if the stored type has data members declared mutable which would be copied by its copy constructor.

Since 2.1.0

Member Data Documentation

◆ done_emitter

template<class Val >
SafeEmitter Cgu::Thread::Future< Val >::done_emitter

A Cgu::SafeEmitter object which is emitted when the function or callable object represented by this Cgu::Thread::Future object finishes correctly (that is, it is not cancelled and does not throw any uncaught exceptions). By itself this emission does not do too much as it is emitted (and connected callbacks execute in) the same worker thread immediately after the Future function has completed. However, any thread can connect a callback object to this Cgu::SafeEmitter object and a connected callback can, say, cause another callback to be executed in a thread's main loop using Cgu::Callback::post(), and from version 2.0.2 when() methods are provided which will do this for users automatically. Once the run() method has been called, this Cgu::Thread::Future object (and so done_emitter) will always stay in existence until the function or callable object represented by it has completed (whether correctly, by cancellation or by a thrown exception) and any callbacks connected to the done_emitter object have completed, irrespective of whether the intrusive pointer returned by the make() or make_future() functions has gone out of scope.

Note
1. Cancellation is blocked while the Cgu::SafeEmitter object emits and any connected callback executes.
2. A connected callback can however terminate the worker thread by throwing Cgu::Thread::Exit (in which case no subsequent callbacks to be executed on that emission will execute either: the worker thread will safely terminate and unwind the stack in so doing). In that event, the emitter_error flag will be set.
3. All other uncaught exceptions which might be thrown by the Cgu::SafeEmitter object emitting, or by a connected callback function executing, are consumed to retain the integrity of the Thread::Future object. In the event of such an exception being thrown, the emitter_error flag will be set. In summary, the emitter_error flag will be set if (a) a connected callback function throws Cgu::Thread::Exit, (b) some other uncaught exception escapes from a connected callback function or (c) Cgu::SafeEmitter::emit() throws std::bad_alloc or the copy constructor of a bound argument which is not a reference argument has thrown. If the user knows that the callback function does not throw Cgu::Thread::Exit and does not allow any other exception to escape, then the cause must be a std::bad_alloc memory exception in Cgu::SafeEmitter::emit() or the copy constructor of a non-reference bound argument throwing.
4. An emission is thread safe if the connected callback functions are thread safe.
5. This Cgu::Thread::Future object's mutex is released while the Cgu::SafeEmitter object emits. This means that any connected callbacks can safely call, say, the Future object's get() or is_error() methods. However, a connected callback should not hold a bound argument comprising a copy of this Cgu::Thread::Future object held by intrusive pointer as returned by the make() or make_future() methods (that would result in this Cgu::Thread::Future object owning, via done_emitter, a reference to itself and so become incapable of being freed). The callback may, however, take a pointer to this Cgu::Thread::Future object as a bound argument, as obtained by the Cgu::IntrusivePtr::get() method, because this Cgu::Thread::Future object is guaranteed to remain in existence until all callbacks connected to done_emitter have completed executing.

The documentation for this class was generated from the following file:
Cgu
Definition: application.h:44
Cgu::Thread::make_future
Cgu::IntrusivePtr< Cgu::Thread::Future< Ret > > make_future(Obj &obj, Ret(Obj::*func)(Params...), Args &&... args)
Definition: future.h:2209