let_async_scopeDate: 19th November 2024
Author: Anthony Williams anthony@justsoftwaresolutions.co.uk
Audience: SG1, LEWG
This is intended to address concerns raised in LEWG about ensuring that a counting_scope (see P3149R3) is joined: the scope provided by let_async_scope is always joined, irrespective of how the nested work completes, and whether or not the provided function throws an exception after spawning work.
Code with explicit counting_scope:
some_data_type scoped_data = make_scoped_data();
counting_scope scope;
spawn(on(exec, [&] {
spawn(on(exec, [&] {
if (need_more_work(scoped_data)) {
spawn(on(exec, [&] { do_more_work(scoped_data); }), scope);
spawn(on(exec, [&] { do_more_other_work(scoped_data); }), scope);
}
}),scope);
spawn(on(exec, [&] { do_something_else_with(scoped_data); }), scope);
}),scope);
maybe_throw();
this_thread::sync_wait(scope.join());Here, if maybe_throw throws an exception, then the scope is not joined, and the nested tasks can continue executing asynchronously, potentially accessing both the scope and scoped_data objects out of lifetime.
Using let_async_scope addresses this by encapsulating the scope object and the result of the previous sender. The returned sender does not complete until all tasks nested on the scope complete, even if the function passed to let_async_scope exits via an exception:
auto scope_sender = just(make_scoped_data()) | let_async_scope([](auto scope_token,
auto& scoped_data) {
spawn(on(exec, [scope_token, &scoped_data] {
spawn(on(exec, [scope_token, &scoped_data] {
if (need_more_work(scoped_data)) {
spawn(on(exec, [&scoped_data] { do_more_work(scoped_data); }),scope_token);
spawn(on(exec, [&scoped_data] { do_more_other_work(scoped_data); }),scope_token);
}
}),scope_token);
spawn(on(exec, [&scoped_data] { do_something_else_with(scoped_data); }),scope_token);
}),scope_token);
maybe_throw();
});
this_thread::sync_wait(scope_sender);Here, even if maybe_throw throws an exception, then scope_sender doesn’t complete until all the nested tasks have completed. This prevents out-of-lifetime access to the scoped_data or the scope itself, unless references to the data or scope_token are stored outside the sender tree.
Stop requests are propagated to all senders nested in the async scope, but does not prevent those senders adding additional work to the scope. This allows senders to respond to stop requests by scheduling additional work to perform the necessary cleanup for cancellation.
If either the function passed to let_async_scope throws an exception, or any of the senders associated with the async scope complete with an error, then that exception or error completion is used as the completion of the sender returned from let_async_scope. This applies even if the function passed to let_async_scope returns normally; in that case the return value is discarded in favour of the error return. If multiple errors are raised, one of them is chosen to be used as the completion of the sender returned from let_async_scope; all other errors are discarded.
Given:
auto scope_sender = just(make_scoped_data()) | let_async_scope([](auto scope_token,
auto& scoped_data) {
spawn(just_error(foo{}),scope_token);
spawn(just_error(bar{}),scope_token);
});
this_thread::sync_wait(scope_sender);Then the sync_wait will throw either an exception of type foo or an exception of type bar, but it is not specified which.
In order to allow the error propagation, all senders associated with the scope must have a compatible error signature. The default error signature is set_error_t(std::exception_ptr), and all raised errors are wrapped with AS-EXCEPT-PTR (See [exec.general/8). An explicit error signature can be specified by calling let_async_scope_with_error, in which case errors are not converted unless std::exception_ptr is the only permitted error type.
“Noexcept” error signature scopes:
auto scope_sender = just(make_scoped_data()) | let_async_scope_with_error<>([](auto scope_token,
auto& scoped_data) noexcept {
spawn(just_error(foo{}),scope_token); // error, sender may fail in "noexcept" scope
});
this_thread::sync_wait(scope_sender);“Default” error signature scopes:
auto scope_sender = just(make_scoped_data()) | let_async_scope([](auto scope_token,
auto& scoped_data) noexcept(false) {
spawn(just_error(foo{}),scope_token); // Coerced with AS-EXCEPT-PTR
});
this_thread::sync_wait(scope_sender); // throws foo{}“Explicit error signature” scopes:
auto scope_sender = just(make_scoped_data()) |
let_async_scope_with_error<foo,bar>(
[](auto scope_token, auto& scoped_data) noexcept {
spawn(just_error(foo{}),scope_token); // OK
spawn(just_error(bar{}),scope_token); // OK
spawn(just_error(baz{}),scope_token); // error
});
this_thread::sync_wait(scope_sender | upon_error([](auto e){
static_assert(is_same<decltype(e),foo> || is_same<decltype(e),bar>,
"Error must be foo or bar");
}));let_async_scope provides a means of creating an async scope (see P3149R6), which is associated with a set of tasks, and ensuring that they are all complete before the async scope sender completes.The previous sender’s result is passed to a user-specified invocable, along with an async scope token, which returns a new sender that is connected and started.
The sender returned by let_async_scope completes with the result of the completion of the sender returned from the supplied invocable. It does not complete until all tasks nested on the scope_token passed to the invocable have completed. Additional tasks may be nested on copies of the scope_token, even if the initial sender returned from the invocable has completed. The returned scope_sender will not complete while there are any nested tasks that have not completed.
If the callable supplied to let_async_scope does not return a sender, it must return void. The sender returned from let_async_scope will then have a void value completion.
Stop requests are propagated to all senders nested in the async scope.
The environment from the receiver connected to the sender returned from let_async_scope is propagated via the internal receiver to all senders spawned using the supplied scope token.
If the callable passed to let_async_scope throws an exception, or any of the senders associated with the scope complete with an error, then a stop request is propagated to all outstanding senders spawned using the scope token.
If either the function passed to let_async_scope throws an exception, or any of the senders associated with the async scope complete with an error, then that exception or error completion is used as the completion of the sender returned from let_async_scope. This applies even if the function passed to let_async_scope returns normally; in that case the return value is discarded in favour of the error return. If multiple errors are raised, one of them is chosen to be used as the completion of the sender returned from let_async_scope; all other errors are discarded.
If the error list specified for let_async_scope_with_error does not specify std::exception_ptr, then the function supplied must be declared noexcept, otherwise the program is ill-formed.
If the possible error completions of senders passed to spawn with a token from let_async_scope_with_error are not compatible with the error signatures then the program is ill-formed.
Please note: this wording is incomplete, and needs review.
execution::let_async_scopelet_async_scope transforms a sender’s value completions into a new child asynchronous operation associated with an async scope, by passing the sender’s result datums to a user-specified callable, which returns a new sender that is connected and started.
For a subexpression sndr, let let-async-scope-env(sndr) be expression-equivalent to the first well-formed expression below:
SCHED-ENV(get_completion_scheduler<decayed-typeof<set_value>>(get_env(sndr)))
MAKE-ENV(get_domain, get_domain(get_env(sndr)))
empty_env{}
The expression let_async_scope(sndr, f) is expression-equivalent to:
let_async_scope_with_error<std::exception_ptr>(sndr, f);The expression let_async_scope_with_error<Errors...>(sndr, f) is expression-equivalent to:
transform_sender(
get-domain-early(sndr),
make-sender(let_async_scope_with_error<Errors...>, f, sndr));Where Errors is the list of possible error completion types.
The exposition-only class template impls-for ([exec.snd.general]) is specialized for let_async_scope as follows:
namespace std::execution {
template<class Errors, class State, class Rcvr, class... Args>
void let-async-scope-bind(State& state, Rcvr& rcvr, Args&&... args); // exposition only
template<typename ... Errors>
struct impls-for<decayed-typeof<let_async_scope_with_error<Errors...>>> : default-impls {
static constexpr auto get-state = see below;
static constexpr auto complete = see below;
};
}Let receiver2 denote the following exposition-only class template:
namespace std::execution {
template<class Rcvr, class Env>
struct receiver2 : Rcvr {
explicit receiver2(Rcvr rcvr, Env env)
: Rcvr(std::move(rcvr)), env(std::move(env)) {}
auto get_env() const noexcept {
const Rcvr& rcvr = *this;
return JOIN-ENV(env, FWD-ENV(execution::get_env(rcvr)));
}
Env env; // exposition only
};
}impls-for<decayed-typeof<let_async_scope_with_error<Errors...>>>::get-state is is initialized with a callable object equivalent to the following:
[]<class Sndr, class Rcvr>(Sndr&& sndr, Rcvr& rcvr) requires see below {
auto&& [tag, data, child] = std::forward<Sndr>(sndr);
return [&]<class Fn, class Env>(Fn fn, Env env) {
using args-variant-type = see below;
using ops2-variant-type = see below;
using scope-type = see below;
struct state-type {
Fn fn;
Env env;
scope-type scope;
args-variant-type args;
ops2-variant-type ops2;
mutex error_mutex;
optional<error-variant-type> error;
};
return state-type{std::move(fn), std::move(env), {}, {}};
}(std::forward_like<Sndr>(data), let-async-scope-env(child));
}scope-type is a type that satisfies the asynchronous-scope concept.
scope-type maintains a count of the number of active senders passed to nest on that scope object or an async-scope-token obtained from that scope object.nest completes.join() on an instance of the scope type will complete when the number of senders reaches zero.scope-token-type is the type of the async-scope-token associated with this invocation of let_async_scope_with_error.
Let Sigs be a pack of the arguments to the completion_signatures specialization named by completion_signatures_of_t<child-type<Sndr>, env_of_t<Rcvr>>. Let LetSigs be a pack of those types in Sigs with a return type of decayed-typeof<set_value>. Let as-tuple be an alias template such that as-tuple<Tag(Args...)> denotes the type decayed-tuple<Args...>. Then args-variant-type denotes the type variant<monostate, as-tuple<LetSigs>...>.
Let as-sndr2 be an alias template such that as-sndr2<Tag(Args...)> denotes the type call-result-t<Fn, scope-token-type, remove_cvref_t<Args>&...>. Then ops2-variant-type denotes the type variant<monostate, connect_result_t<as-sndr2<LetSigs>, receiver2<Rcvr, Env>>...>.
The requires-clause constraining the above lambda is satisfied if and only if the types args-variant-type and ops2-variant-type are well-formed.
error-variant-type is a variant<E...>, where the types E... are the corresponding E types from the Errors... parameter of the let_async_scope_with_error<Errors...> invocation.
scope-token-type shall be a unique type, such that invoking spawn(snd, token, env) where token is an instance of scope-token-type invokes a distinct overload of spawn. Such an invocation is ill-formed if the completion signatures of snd include error completions that are not compatible with the Errors... list of the let_async_scope_with_error<Errors...> invocation. If the error list is compatible, then such an invocation of spawn is equivalent to
spawn(snd | upon_error(
[&state](auto&& error){
{
lock_guard guard(state.error_mutex);
state.errors.emplace(TRANSFORM-ERROR(error));
}
state.scope.request_stop();
}), state.scope.get_token(), env);Where TRANSFORM-ERROR is AS-EXCEPT-PTR(error) if Errors... is std::exception_ptr, and std::forward<decltype(error)>(error) otherwise.
The exposition-only function template let-async-scope-bind is equal to:
auto& args = state.args.emplace<decayed-tuple<scope-token-type, Args...>>(
create-scope-token(), std::forward<Args>(args)...);
try {
auto sndr2 = nest(apply(std::move(state.fn), args), state.scope);
auto join_sender = state.scope.join();
auto result_sender = when_all_with_variant(std::move(sndr2), std::move(join_sender)) |
then([](auto& result, auto&) { return result; });
auto rcvr2 = receiver2{std::move(rcvr), std::move(state.env)};
auto mkop2 = [&] { return connect(std::move(result_sender), std::move(rcvr2)); };
auto& op2 = state.ops2.emplace<decltype(mkop2())>(emplace-from{mkop2});
start(op2);
} catch (...) {
state.scope.request_stop();
auto result_sender = when_all(just_error(std::current_exception()), state.scope.join());
auto rcvr2 = receiver2{std::move(rcvr), std::move(state.env)};
auto mkop2 = [&] { return connect(std::move(result_sender), std::move(rcvr2)); };
auto& op2 = state.ops2.emplace<decltype(mkop2())>(emplace-from{mkop2});
start(op2);
}where create-scope-token() creates an instance of the scope-token-type associated with the state for this invocation of let_async_scope_with_error.
impls-for<decayed-typeof<let_async_scope_with_error<Errors...>>>::complete is is initialized with a callable object equivalent to the following:
[]<class Tag, class... Args>
(auto, auto& state, auto& rcvr, Tag, Args&&... args) noexcept -> void {
if constexpr (same_as<Tag, decayed-typeof<set_value>>) {
TRY-EVAL(std::move(rcvr), let-async-scope-bind(state, rcvr, std::forward<Args>(args)...));
} else {
Tag()(std::move(rcvr), std::forward<Args>(args)...);
}
}Let sndr and env be subexpressions, and let Sndr be decltype((sndr)). If sender-for<Sndr, decayed-typeof<let_async_scope_with_error<Errors...>>> is false, then the expression let_async_scope_with_error<Errors...>.transform_env(sndr, env) is ill-formed. Otherwise, it is equal to JOIN-ENV(let-env(sndr), FWD-ENV(env)).
Let the subexpression out_sndr denote the result of the invocation let_async_scope_with_error<Errors...>(sndr, f) or an object copied or moved from such, and let the subexpression rcvr denote a receiver such that the expression connect(out_sndr, rcvr) is well-formed. The expression connect(out_sndr, rcvr) has undefined behavior unless it creates an asynchronous operation ([async.ops]) that, when started:
invokes f when set_value is called with sndr’s result datums,
makes its completion dependent on the completion of a sender returned by f, and
propagates the other completion operations sent by sndr.
spawn is expected.Thanks to Ian Petersen, Lewis Baker, Inbal Levi, Kirk Shoop, Eric Niebler, Ruslan Arutyunyan, Maikel Nadolski, Lucian Radu Teodorescu, Robert Leahy, Dmitry Prokoptsev, and everyone else who contributed to discussions leading to this paper, and commented on early drafts.