[ 
https://issues.apache.org/jira/browse/MESOS-3188?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14652212#comment-14652212
 ] 

Benjamin Mahler commented on MESOS-3188:
----------------------------------------

Sounds good!

> Add tuple-awareness to Future callbacks.
> ----------------------------------------
>
>                 Key: MESOS-3188
>                 URL: https://issues.apache.org/jira/browse/MESOS-3188
>             Project: Mesos
>          Issue Type: Improvement
>          Components: libprocess
>            Reporter: Benjamin Mahler
>
> Future is currently single-valued and so the only way to create a multi-value 
> Future is to use a custom struct or a tuple. Since Future is not currently 
> "tuple-aware", continuations have to take a tuple as well, and manually 
> unpack:
> {code}
>   {
>     // Subprocess example.
>     await(s.get().status(),
>           io::read(s.get().out().get()),
>           io::read(s.get().err().get()))
>       .then(defer(self(), &Self::continue, lambda::_1));
>   }
>   void continue(std::tuple<
>       Future<Option<int>>,
>       Future<string>,
>       Future<string>> results)
>   {
>     Future<Option<int>> status = std::get<0>(results);
>     Future<string> output = std::get<1>(results);
>     Future<string> error = std::get<2>(results);
>   }
> {code}
> Since multi-value Future (i.e. Future<T1, T2, ...>) seems to be a bad design 
> choice, being tuple aware can improve the code cleanliness by unpacking the 
> tuple automatically for the continuation:
> {code}
>   {
>     // Subprocess example.
>     await(s.get().status(),
>           io::read(s.get().out().get()),
>           io::read(s.get().err().get()))
>       .then(defer(self(), &Self::continue, lambda::_1));
>   }
>   void continue(
>       Future<Option<int>> status,
>       Future<string> output,
>       Future<string> error)
>   {
>     ...
>   }
> {code}



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to