Skip to content
Permalink
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
336 lines (237 sloc) 10.4 KB
=begin pod :kind("Type") :subkind("class") :category("domain-specific")
=TITLE class Promise
=SUBTITLE Status/result of an asynchronous computation
my enum PromiseStatus (:Planned(0), :Kept(1), :Broken(2));
class Promise {}
A I<Promise> is used to handle the result of a computation that might not have
finished. It allows the user to execute code once the computation is done
(with the C<then> method), execution after a time delay (with C<in>),
combining promises, and waiting for results.
my $p = Promise.start({ sleep 2; 42});
$p.then({ say .result }); # will print 42 once the block finished
say $p.status; # OUTPUT: «Planned␤»
$p.result; # waits for the computation to finish
say $p.status; # OUTPUT: «Kept␤»
There are two typical scenarios for using promises. The first is to use a
factory method (C<start>, C<in>, C<at>, C<anyof>, C<allof>, C<kept>, C<broken>)
on the type object; those will make sure that the promise is automatically kept
or broken for you, and you can't call C<break> or C<keep> on these promises
yourself.
The second is to create your promises yourself with C<Promise.new>. If you
want to ensure that only your code can keep or break the promise, you can use
the C<vow> method to get a unique handle, and call C<keep> or C<break> on it:
=begin code
sub async-get-with-promise($user-agent, $url) {
my $p = Promise.new;
my $v = $p.vow;
# do an asynchronous call on a fictive user agent,
# and return the promise:
$user-agent.async-get($url,
on-error => -> $error {
$v.break($error);
},
on-success => -> $response {
$v.keep($response);
}
);
return $p;
}
=end code
Further examples can be found in the L<concurrency page|/language/concurrency#Promises>.
=head1 Methods
=head2 method start
method start(Promise:U: &code, :$scheduler = $*SCHEDULER --> Promise:D)
Creates a new Promise that runs the given code object. The promise will be
kept when the code terminates normally, or broken if it throws an exception.
The return value or exception can be inspected with the C<result> method.
The scheduler that handles this promise can be passed as a named argument.
There is also a statement prefix C<start> that provides syntactic sugar for
this method:
# these two are equivalent:
my $p1 = Promise.start({ ;#`( do something here ) });
my $p2 = start { ;#`( do something here ) };
As of the 6.d version of the language, C<start> statement prefix used in
L<sink|/routine/sink> context will automatically attach an exceptions handler. If an
exception occurs in the given code, it will be printed and the program
will then exit, like if it were thrown without any C<start> statement
prefixes involved.
=begin code :solo
use v6.c;
start { die }; sleep ⅓; say "hello"; # OUTPUT: «hello␤»
=end code
=begin code :solo
use v6.d;
start { die }; sleep ⅓; say "hello";
# OUTPUT:
# Unhandled exception in code scheduled on thread 4
# Died
# in block at -e line 1
=end code
If you wish to avoid this behavior, use C<start> in non-sink context or
catch the exception yourself:
=begin code
# Don't sink it:
my $ = start { die }; sleep ⅓; say "hello"; # OUTPUT: «hello␤»
# Catch yourself:
start { die; CATCH { default { say "caught" } } };
sleep ⅓;
say "hello";
# OUTPUT: «caught␤hello␤»
=end code
This behavior exists only syntactically, by using an alternate C<.sink> method
for L<Promise|/type/Promise> objects created by C<start> blocks in sink context, thus simply
sinking a L<Promise|/type/Promise> object that was created by other means won't trigger this
behavior.
=head2 method in
method in(Promise:U: $seconds, :$scheduler = $*SCHEDULER --> Promise:D)
Creates a new Promise that will be kept in C<$seconds> seconds, or later.
my $proc = Proc::Async.new('perl6', '-e', 'sleep 10; warn "end"');
my $result = await Promise.anyof(
my $promise = $proc.start, # may or may not work in time
Promise.in(5).then: { # fires after 5 seconds no matter what
unless $promise { # don't do anything if we were successful
note 'timeout';
$proc.kill;
}
}
).then: { $promise.result }
# OUTPUT: «timeout␤»
C<$seconds> can be fractional or negative. Negative values are treated as
C<0> (i.e. L<keeping|/routine/keep> the returned L<Promise|/type/Promise> right away).
Please note that situations like these are often more clearly handled with
a L<react and whenever block|/language/concurrency#index-entry-react-react>.
=head2 method at
method at(Promise:U: $at, :$scheduler = $*SCHEDULER --> Promise:D)
Creates a new C<Promise> that will be kept C<$at> the given time—which is
given as an L<Instant|/type/Instant> or equivalent L<Numeric|/type/Numeric>—or as soon as possible after it.
my $p = Promise.at(now + 2).then({ say "2 seconds later" });
# do other stuff here
await $p; # wait here until the 2 seconds are over
If the given time is in the past, it will be treated as L<now|/routine/now> (i.e.
L<keeping|/routine/keep> the returned L<Promise|/type/Promise> right away).
Please note that situations like these are often more clearly handled with
a L<react and whenever block|/language/concurrency#index-entry-react-react>.
=head2 method kept
multi method kept(Promise:U: \result = True --> Promise:D)
Returns a new promise that is already kept, either with the given value,
or with the default value C<True>.
=head2 method broken
multi method broken(Promise:U: --> Promise:D)
multi method broken(Promise:U: \exception --> Promise:D)
Returns a new promise that is already broken, either with the given value,
or with the default value C«X::AdHoc.new(payload => "Died")»
=head2 method allof
method allof(Promise:U: *@promises --> Promise:D)
Returns a new promise that will be kept when all the promises passed as
arguments are kept or broken. The result of the individual Promises is
not reflected in the result of the returned promise: it simply
indicates that all the promises have been completed in some way.
If the results of the individual promises are important then they should
be inspected after the C<allof> promise is kept.
In the following requesting the C<result> of a broken promise will cause the
original Exception to be thrown. (You may need to run it several times to
see the exception.)
my @promises;
for 1..5 -> $t {
push @promises, start {
sleep $t;
};
}
my $all-done = Promise.allof(@promises);
await $all-done;
@promises>>.result;
say "Promises kept so we get to live another day!";
=head2 method anyof
method anyof(Promise:U: *@promises --> Promise:D)
Returns a new promise that will be kept as soon as any of the promises
passed as arguments is kept or broken. The result of the completed
Promise is not reflected in the result of the returned promise which
will always be Kept.
You can use this to wait at most a number of seconds for a promise:
my $timeout = 5;
await Promise.anyof(
Promise.in($timeout),
start {
# do a potentially long-running calculation here
},
);
=head2 method then
method then(Promise:D: &code)
Schedules a piece of code to be run after the invocant has been kept or
broken, and returns a new promise for this computation. In other words,
creates a chained promise.
my $timer = Promise.in(2);
my $after = $timer.then({ say "2 seconds are over!"; 'result' });
say $after.result; # 2 seconds are over
# result
=head2 method keep
multi method keep(Promise:D: \result = True)
Keeps a promise, optionally setting the result. If no result is passed, the
result will be C<True>.
Throws an exception of type C<X::Promise::Vowed> if a vow has already been
taken. See method C<vow> for more information.
my $p = Promise.new;
if Bool.pick {
$p.keep;
}
else {
$p.break;
}
=head2 method break
multi method break(Promise:D: \cause = False)
Breaks a promise, optionally setting the cause. If no cause is passed, the
cause will be C<False>.
Throws an exception of type C<X::Promise::Vowed> if a vow has already been
taken. See method C<vow> for more information.
my $p = Promise.new;
$p.break('sorry');
say $p.status; # OUTPUT: «Broken␤»
say $p.cause; # OUTPUT: «sorry␤»
=head2 method result
method result(Promise:D)
Waits for the promise to be kept or broken. If it is kept, returns the result;
otherwise throws the result as an exception.
=head2 method cause
method cause(Promise:D)
If the promise was broken, returns the result (or exception). Otherwise, throws
an exception of type C<X::Promise::CauseOnlyValidOnBroken>.
=head2 method Bool
multi method Bool(Promise:D:)
Returns C<True> for a kept or broken promise, and C<False> for one in state
C<Planned>.
=head2 method status
method status(Promise:D --> PromiseStatus)
Returns the current state of the promise: C<Kept>, C<Broken> or C<Planned>:
=for code :skip-test<compile time error>
say "promise got Kept" if $promise.status ~~ Kept;
=head2 method scheduler
method scheduler(Promise:D:)
Returns the scheduler that manages the promise.
=head2 method vow
=for code
my class Vow {
has Promise $.promise;
method keep() { ... }
method break() { ... }
}
method vow(Promise:D: --> Vow:D)
Returns an object that holds the sole authority over keeping or breaking a
promise. Calling C<keep> or C<break> on a promise that has vow taken throws an
exception of type C<X::Promise::Vowed>.
my $p = Promise.new;
my $vow = $p.vow;
$vow.keep($p);
say $p.status; # OUTPUT: «Kept␤»
=head2 method Supply
method Supply(Promise:D:)
Returns a L<Supply|/type/Supply> that will emit the C<result> of the L<Promise|/type/Promise> being Kept
or C<quit> with the C<cause> if the L<Promise|/type/Promise> is Broken.
=head2 sub await
multi sub await(Promise:D --> Promise)
multi sub await(*@ --> Array)
Waits until one or more promises are I<all> fulfilled, and then returns their
values. Also works on L<channels|/type/Channel>. Any broken promises will
rethrow their exceptions. If a list is passed it will return a
list containing the results of awaiting each item in turn.
=end pod
# vim: expandtab softtabstop=4 shiftwidth=4 ft=perl6
You can’t perform that action at this time.