From: Thorsten Schuett (schuett_at_[hidden])
Date: 2005-08-17 07:13:34
On Wednesday 17 August 2005 13:39, Martin Wille wrote:
> Thorsten Schuett wrote:
> > future<int> f3 = f1 || f2;
>
> We'll probably need more than operator ||.
>
> E.g. consider f1 and f2 implement operations that consume time and may
> fail, e.g.:
>
>
> future<whatever> x = query_google_and_check_links_returned ||
> dig_in_local_incomplete_archive;
>
>
> The local archive is faster than searching something over the internet,
> but the search may fail. Having an or-operation that waits for the first
> function to return a *success* would be useful.
Interesting, I never thought of a "failing" future. We could end up with
something like the following:
simple_future<optional<int> > f1;
simple_future<optional<double> > f2;
future<optional<variant<int, double> > > f3 = f1 || f2;
where an empty optional represents failure.
> In another reply, thread-cancellation was mentioned. We don't have
> thread-cancellation currently, but it would be nice if future<> would
> not only accept functors but also allow for an optional interface to be
> passed for signaling an abort request. A user would then be able to
> implement a mechanism that works form her/him. Maybe, an abort policy
> would be the right thing here.
This depends on the cancelation possibilities in boost.thread.
Thorsten