Welcome to Software Development on Codidact!
Will you help us build our independent community of developers helping developers? We're small and trying to grow. We welcome questions about all aspects of software development, from design to code to QA and more. Got questions? Got answers? Got code you'd like someone to review? Please join us.
When should I use wait() instead of get() when using C++ threads
I'm trying to understand the purpose of the wait()
function in Class future<> for C++. At the moment I don't see any purpose for calling wait()
instead of get()
. What I have tried in code:
int main()
{
std::future<int>result1(std::async(std::launch::deferred,func1));
// start synchronously.
//result1.get();
result1.wait(); // gives the same output as calling get()
//print character L
for (int i = 0; i < 20; ++i)
{
std::cout.put('L').flush();
}
}
here is what's getting called:
int doSomething(char c)
{
// random-number generator (use c as seed to get different sequences)
std::default_random_engine dre(c);
std::uniform_int_distribution<int> id(10,1000);
//loop to print character after a random period of time
for (int i=0; i<10; ++i)
{
std::cout.put(c).flush();
}
return c;
}
int func1 ()
{
return doSomething('.');
}
But the output I get is the same as I would have had if I called get()
instead. That is, func1 gets called prints some characters to the screen and then character is L
is printed 20 times.
So, I'm trying to figure out in which situations wait()
would be better suited than get()
.
In the book C++ standard library by Nicolai M. Josuttis, it says that wait() function call allows a background operation to finish without processing its outcome. So, when would we need to do that ?
1 answer
tl;dr If you don't have a reason otherwise, you should use get
. More precisely, if you will be done with the future, you should use get
. You should use wait
only if there is some separation between when you want to ensure the visible side-effects of the future have completed and when you will be done with the future.
The purpose of wait
is to cause the thread executing it to block until all visible side-effects of the corresponding future are completed.
The purpose of get
is to get the result (if any) or throw any exceptions for a failed future (which requires waiting first). After get
completes, the future is no longer valid, i.e. get
can be called at most once. This also allows the future to discard some internal state.
You should use wait
if you want to ensure any potential (visible) side-effects of a future are completed before some point in time, but you intend to consume the result later (including throwing any waiting exceptions).
You should use get
when you want to consume the result of the future including throwing any exceptions.
Generally, you should use get
even for future<void>
as opposed to wait
. wait
is to ensure that visible side-effects happen before some other code. In most cases, you might as well get the result at this point, but, particularly for utility code, you may not want to actually consume the results. For example, I may make a utility function to wait on a vector of futures, but I don't actually want to consume the results or cause the exceptions to be thrown. The type of the utility may be:
template <typename T>
vector< future< T > > &wait_all(vector< future< T > > &futures);
Importantly, wait
can be called as often as one wants as long as the future is valid. This means the above code does not have to worry that some of the futures have already had wait
called.
Focusing on future<void>
, the differences between get
and wait
(assuming a valid future) is that 1) get
will potentially throw exceptions, and 2) the future is no longer valid after get
is called (allowing resources to be released).
0 comment threads