2 * Copyright 2014 Facebook, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
19 #include <folly/wangle/Future.h>
20 #include <folly/wangle/Deprecated.h>
22 namespace folly { namespace wangle {
25 The ThreadGate strategy encapsulates a bidirectional gate via two Executors,
26 kind of like a stargate for wangle Future chains. Its implementation is
27 slightly more efficient then using Future::via in both directions, and if
28 the pattern is common it can be more convenient (although the overhead of
29 setting up a ThreadGate is less convenient in most situations).
31 // Using a ThreadGate (which has two executors xe and xw)
39 If you're not sure whether you want a ThreadGate, you don't. Use via.
41 There are two actors, the east thread which does the asynchronous operation
42 (the server) and the west thread that wants the asynchronous operation done
45 The client calls gate<T>(fn), which returns a Future<T>. Practically speaking
46 the returned Future<T> is the same as the Future<T> returned by fn. But
47 there are actually two futures involved - the original Future which will be
48 generated by fn (called the east Future), and the Future actually returned
49 by gate<T>(fn) (called the west Future).
51 These two futures are decoupled, and although the fulfilment of the east
52 Future eventually causes fulfilment of the west Future, those fulfilments
53 happen in their own threads.
55 In order to make and use a ThreadGate, you need to provide a strategy for
56 executing code in the east and west threads. These strategies may be
57 different. The only requirement is a threadsafe method
58 `void add(function<void()>&&)`.
60 In order for your ThreadGate to do anything, you need to drive those
61 executors somehow. An event loop is a natural fit. A thread pool might be
62 made to work. You could use a busy loop to make a very expensive space
63 heater. 0MQ would be pleasant.
65 Another pattern supported by the ThreadGate is the single-thread pattern. In
66 this pattern, non-blocking I/O drives the asynchronous operation, and
67 futures are fulfilled in an event loop callback. In this scenario,
68 ThreadGate is largely superfluous, and the executors would likely just
69 execute code immediately and inline (and therefore not need to be driven, or
70 threadsafe). But a Waiter strategy that makes progress by driving the event
71 loop one iteration would allow for gate-and-wait code which is agnostic to
72 the small detail that everything happens in one thread. It would also make
73 Future change toward a multithreaded architecture easier, as you need only
74 change the components of the ThreadGate which your client code is already
77 // DEPRECATED. Just use Future::via() to accomplish the same thing. If it's
78 // not obvious how, feel free to reach out.
79 class DEPRECATED ThreadGate {
81 virtual ~ThreadGate() {}
84 Returns a Future that will be fulfilled after the Future that will be
85 returned by fn() has been fulfilled, with the same value or exception
88 There's a lot of nuance in that sentence. Let's break it down.
90 fn kicks off the asynchronous operation (makes the east Promise), and must
91 be executed in the east thread because the east thread is where the east
92 Promise will be fulfilled. Since gate is being called from the west
93 thread, we must gate fn using the east executor. fn is not executed
94 immediately, it is queued up and will be executed by the east thread as it
97 We create the west Promise and return its Future.
99 When the east thread executes its task, fn is called and the resulting
100 Future gets a callback that will gate another task back to the west.
102 Sometime later, the asynchronous operation completes and the east Promise
103 is fulfilled. Then the east Future executes its callback, which adds a
104 task to the west executor that task is to fulfil the west Promise with the
105 same Try<T>, and it will execute in the west thread.
107 At this point, the west Future is still unfulfilled, even though the east
108 Future has been fulfilled and its callback has finished executing. Only
109 when the west executor is driven to execute that task, the west Future
110 will be completed and its callbacks called.
112 In summary, both east and west need to have plans to drive their
113 executors, or nothing will actually happen. When the executors are driven,
114 then everything flows. */
116 Future<T> gate(std::function<Future<T>()>&& fn) {
118 Future<T> fWest = pWest.getFuture();
120 gate(std::move(fn), std::move(pWest));
125 * This version of gate is to support use cases where the calling thread is
126 * not the west thread. Here is an example use case.
129 * Future<T> fWest = pWest.getFuture();
131 * // Set up callbacks for west from a thread that is not west.
132 * fWest.then(...).then(...);
134 * threadGate.gate(..., std::move(pWest));
136 * This function assumes that it is safe to call addEast from a thread that is
137 * not the west thread.
140 void gate(std::function<Future<T>()>&& fn,
142 folly::MoveWrapper<Promise<T>> pWest(std::move(p));
143 folly::MoveWrapper<std::function<Future<T>()>> fnm(std::move(fn));
144 this->addEast([pWest, fnm, this]() mutable {
145 (*fnm)().then([pWest, this](Try<T>&& t) mutable {
146 folly::MoveWrapper<Try<T>> tm(std::move(t));
147 this->addWest([pWest, tm]() mutable {
148 pWest->fulfilTry(std::move(*tm));
155 If your workflow calls for synchronizing with a
156 west Future, then you may call waitFor, but if your west thread is
157 event-driven you will probably not need to call waitFor.
159 In order for waitFor to behave properly, you must ensure that the Waiter's
160 makeProgress method causes some progress to be made on the west thread,
161 i.e. drives the west executor either directly or indirectly.
163 (Naturally, progress needs to be made on the east thread as well. i.e. the
164 east executor is driven, the asynchronous operation happens, and its
165 Promise is fulfilled. It is likely that none of this concerns the consumer
168 This is the only function that uses the Waiter. It is never called
169 internally. Therefore, if you never use waitFor you can safely provide a
173 void waitFor(Future<T> const& f) {
174 while (!f.isReady()) {
175 this->makeProgress();
180 typename std::add_lvalue_reference<T>::type
181 value(Future<T>& f) {
187 typename std::add_lvalue_reference<const T>::type
188 value(Future<T> const& f) {
193 virtual void addEast(std::function<void()>&&) = 0;
194 virtual void addWest(std::function<void()>&&) = 0;
195 virtual void makeProgress();