Class: Async::Reactor
- Extended by:
- Forwardable
- Defined in:
- lib/async/reactor.rb
Overview
An asynchronous, cooperatively scheduled event reactor.
Instance Attribute Summary
Attributes inherited from Node
#annotation, #children, #head, #parent, #tail
Class Method Summary collapse
-
.run(*arguments, **options, &block) ⇒ Object
The preferred method to invoke asynchronous behavior at the top level.
- .selector ⇒ Object
Instance Method Summary collapse
-
#<<(fiber) ⇒ Object
Schedule a fiber (or equivalent object) to be resumed on the next loop through the reactor.
-
#async(*arguments, **options) {|Task| ... } ⇒ Task
Start an asynchronous task within the specified reactor.
-
#close ⇒ void
Stop each of the children tasks and close the selector.
-
#closed? ⇒ Boolean
Check if the selector has been closed.
- #finished? ⇒ Boolean
-
#initialize(parent = nil, selector: self.class.selector, logger: nil) ⇒ Reactor
constructor
A new instance of Reactor.
-
#interrupt ⇒ Object
Interrupt the reactor at the earliest convenience.
- #logger ⇒ Object
- #register(io, interest, value = Fiber.current) ⇒ Object
-
#run(*arguments, **options, &block) ⇒ Object
Run the reactor until all tasks are finished.
-
#run_once(timeout = nil) ⇒ Boolean
Run one iteration of the event loop.
-
#sleep(duration) ⇒ Object
Put the calling fiber to sleep for a given ammount of time.
- #stop(later = true) ⇒ Object
- #stopped? ⇒ Boolean
- #to_s ⇒ Object
-
#with_timeout(timeout, exception = TimeoutError) ⇒ Object
Invoke the block, but after the specified timeout, raise TimeoutError in any currenly blocking operation.
-
#yield(fiber = Fiber.current) ⇒ Object
Yield the current fiber and resume it on the next iteration of the event loop.
Methods inherited from Node
#annotate, #backtrace, #children?, #consume, #description, #print_hierarchy, #transient?, #traverse
Constructor Details
#initialize(parent = nil, selector: self.class.selector, logger: nil) ⇒ Reactor
Returns a new instance of Reactor.
75 76 77 78 79 80 81 82 83 84 85 86 87 |
# File 'lib/async/reactor.rb', line 75 def initialize(parent = nil, selector: self.class.selector, logger: nil) super(parent) @selector = selector @timers = Timers::Group.new @logger = logger @ready = [] @running = [] @interrupted = false @guard = Mutex.new end |
Class Method Details
.run(*arguments, **options, &block) ⇒ Object
The preferred method to invoke asynchronous behavior at the top level.
-
When invoked within an existing reactor task, it will run the given block
asynchronously. Will return the task once it has been scheduled.
-
When invoked at the top level, will create and run a reactor, and invoke
the block as an asynchronous task. Will block until the reactor finishes running.
47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 |
# File 'lib/async/reactor.rb', line 47 def self.run(*arguments, **, &block) if current = Task.current? reactor = current.reactor return reactor.async(*arguments, **, &block) else reactor = self.new begin return reactor.run(*arguments, **, &block) ensure reactor.close end end end |
.selector ⇒ Object
63 64 65 66 67 68 69 70 71 72 73 |
# File 'lib/async/reactor.rb', line 63 def self.selector if backend = ENV['ASYNC_BACKEND']&.to_sym if NIO::Selector.backends.include?(backend) return NIO::Selector.new(backend) else warn "Could not find ASYNC_BACKEND=#{backend}!" end end return NIO::Selector.new end |
Instance Method Details
#<<(fiber) ⇒ Object
Schedule a fiber (or equivalent object) to be resumed on the next loop through the reactor.
144 145 146 |
# File 'lib/async/reactor.rb', line 144 def << fiber @ready << fiber end |
#async(*arguments, **options) {|Task| ... } ⇒ Task
Start an asynchronous task within the specified reactor. The task will be executed until the first blocking call, at which point it will yield and and this method will return.
This is the main entry point for scheduling asynchronus tasks.
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 |
# File 'lib/async/reactor.rb', line 109 def async(*arguments, **, &block) task = Task.new(self, **, &block) # I want to take a moment to explain the logic of this. # When calling an async block, we deterministically execute it until the # first blocking operation. We don't *have* to do this - we could schedule # it for later execution, but it's useful to: # - Fail at the point of the method call where possible. # - Execute determinstically where possible. # - Avoid scheduler overhead if no blocking operation is performed. task.run(*arguments) # logger.debug "Initial execution of task #{fiber} complete (#{result} -> #{fiber.alive?})..." return task end |
#close ⇒ void
This method returns an undefined value.
Stop each of the children tasks and close the selector.
247 248 249 250 251 252 253 |
# File 'lib/async/reactor.rb', line 247 def close # This is a critical step. Because tasks could be stored as instance variables, and since the reactor is (probably) going out of scope, we need to ensure they are stopped. Otherwise, the tasks will belong to a reactor that will never run again and are not stopped. self.stop(false) @selector.close @selector = nil end |
#closed? ⇒ Boolean
Check if the selector has been closed.
257 258 259 |
# File 'lib/async/reactor.rb', line 257 def closed? @selector.nil? end |
#finished? ⇒ Boolean
155 156 157 158 |
# File 'lib/async/reactor.rb', line 155 def finished? # TODO I'm not sure if checking `@running.empty?` is really required. super && @ready.empty? && @running.empty? end |
#interrupt ⇒ Object
Interrupt the reactor at the earliest convenience. Can be called from a different thread safely.
133 134 135 136 137 138 139 140 |
# File 'lib/async/reactor.rb', line 133 def interrupt @guard.synchronize do unless @interrupted @interrupted = true @selector.wakeup end end end |
#logger ⇒ Object
89 90 91 |
# File 'lib/async/reactor.rb', line 89 def logger @logger || Console.logger end |
#register(io, interest, value = Fiber.current) ⇒ Object
125 126 127 128 129 130 |
# File 'lib/async/reactor.rb', line 125 def register(io, interest, value = Fiber.current) monitor = @selector.register(io, interest) monitor.value = value return monitor end |
#run(*arguments, **options, &block) ⇒ Object
Run the reactor until all tasks are finished. Proxies arguments to #async immediately before entering the loop, if a block is provided.
223 224 225 226 227 228 229 230 231 232 233 234 235 |
# File 'lib/async/reactor.rb', line 223 def run(*arguments, **, &block) raise RuntimeError, 'Reactor has been closed' if @selector.nil? initial_task = self.async(*arguments, **, &block) if block_given? while self.run_once # Round and round we go! end return initial_task ensure logger.debug(self) {"Exiting run-loop because #{$! ? $! : 'finished'}."} end |
#run_once(timeout = nil) ⇒ Boolean
Run one iteration of the event loop.
163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 |
# File 'lib/async/reactor.rb', line 163 def run_once(timeout = nil) # logger.debug(self) {"@ready = #{@ready} @running = #{@running}"} if @ready.any? # running used to correctly answer on `finished?`, and to reuse Array object. @running, @ready = @ready, @running @running.each do |fiber| fiber.resume if fiber.alive? end @running.clear end if @ready.empty? interval = @timers.wait_interval else # if there are tasks ready to execute, don't sleep: interval = 0 end # If we are finished, we stop the task tree and exit: if self.finished? return false end # If there is no interval to wait (thus no timers), and no tasks, we could be done: if interval.nil? # Allow the user to specify a maximum interval if we would otherwise be sleeping indefinitely: interval = timeout elsif interval < 0 # We have timers ready to fire, don't sleep in the selctor: interval = 0 elsif timeout and interval > timeout interval = timeout end # logger.debug(self) {"Selecting with #{@children&.size} children with interval = #{interval ? interval.round(2) : 'infinite'}..."} if monitors = @selector.select(interval) monitors.each do |monitor| monitor.value.resume end end @timers.fire # We check and clear the interrupted flag here: if @interrupted @guard.synchronize do @interrupted = false end return false end # The reactor still has work to do: return true end |
#sleep(duration) ⇒ Object
Put the calling fiber to sleep for a given ammount of time.
263 264 265 266 267 268 269 270 271 272 273 274 275 |
# File 'lib/async/reactor.rb', line 263 def sleep(duration) fiber = Fiber.current timer = @timers.after(duration) do if fiber.alive? fiber.resume end end Task.yield ensure timer.cancel if timer end |
#stop(later = true) ⇒ Object
237 238 239 240 241 242 |
# File 'lib/async/reactor.rb', line 237 def stop(later = true) @children&.each do |child| # We don't want this process to propagate `Async::Stop` exceptions, so we schedule tasks to stop later. child.stop(later) end end |
#stopped? ⇒ Boolean
97 98 99 |
# File 'lib/async/reactor.rb', line 97 def stopped? @children.nil? end |
#to_s ⇒ Object
93 94 95 |
# File 'lib/async/reactor.rb', line 93 def to_s "\#<#{self.description} #{@children&.size || 0} children (#{stopped? ? 'stopped' : 'running'})>" end |
#with_timeout(timeout, exception = TimeoutError) ⇒ Object
Invoke the block, but after the specified timeout, raise TimeoutError in any currenly blocking operation. If the block runs to completion before the timeout occurs or there are no non-blocking operations after the timeout expires, the code will complete without any exception.
280 281 282 283 284 285 286 287 288 289 290 291 292 293 |
# File 'lib/async/reactor.rb', line 280 def with_timeout(timeout, exception = TimeoutError) fiber = Fiber.current timer = @timers.after(timeout) do if fiber.alive? error = exception.new("execution expired") fiber.resume error end end yield timer ensure timer.cancel if timer end |
#yield(fiber = Fiber.current) ⇒ Object
Yield the current fiber and resume it on the next iteration of the event loop.
149 150 151 152 153 |
# File 'lib/async/reactor.rb', line 149 def yield(fiber = Fiber.current) @ready << fiber Fiber.yield end |