On this page
Maintenance Ruby 3.4 gets fixes, but no new features. Switch to Ruby 4.0 →
Ractor
Ractor is an Actor-model abstraction for Ruby that provides thread-safe parallel execution.
Ractor.new makes a new Ractor, which can run in parallel.
# The simplest ractor r = Ractor.new {puts "I am in Ractor!"} r.take # wait for it to finish # Here, "I am in Ractor!" is printed
Ractors do not share all objects with each other. There are two main benefits to this: across ractors, thread-safety concerns such as data-races and race-conditions are not possible. The other benefit is parallelism.
To achieve this, object sharing is limited across ractors. For example, unlike in threads, ractors can’t access all the objects available in other ractors. Even objects normally available through variables in the outer scope are prohibited from being used across ractors.
a = 1 r = Ractor.new {puts "I am in Ractor! a=#{a}"} # fails immediately with # ArgumentError (can not isolate a Proc because it accesses outer variables (a).)
The object must be explicitly shared:
a = 1 r = Ractor.new(a) { |a1| puts "I am in Ractor! a=#{a1}"}
On CRuby (the default implementation), Global Virtual Machine Lock (GVL) is held per ractor, so ractors can perform in parallel without locking each other. This is unlike the situation with threads on CRuby.
Instead of accessing shared state, objects should be passed to and from ractors by sending and receiving them as messages.
a = 1 r = Ractor.new do a_in_ractor = receive # receive blocks until somebody passes a message puts "I am in Ractor! a=#{a_in_ractor}" end r.send(a) # pass it r.take # Here, "I am in Ractor! a=1" is printed
There are two pairs of methods for sending/receiving messages:
-
Ractor#sendandRactor.receivefor when the sender knows the receiver (push); -
Ractor.yieldandRactor#takefor when the receiver knows the sender (pull);
In addition to that, any arguments passed to Ractor.new are passed to the block and available there as if received by Ractor.receive, and the last block value is sent outside of the ractor as if sent by Ractor.yield.
A little demonstration of a classic ping-pong:
server = Ractor.new(name: "server") do puts "Server starts: #{self.inspect}" puts "Server sends: ping" Ractor.yield 'ping' # The server doesn't know the receiver and sends to whoever interested received = Ractor.receive # The server doesn't know the sender and receives from whoever sent puts "Server received: #{received}" end client = Ractor.new(server) do |srv| # The server is sent to the client, and available as srv puts "Client starts: #{self.inspect}" received = srv.take # The client takes a message from the server puts "Client received from " \ "#{srv.inspect}: #{received}" puts "Client sends to " \ "#{srv.inspect}: pong" srv.send 'pong' # The client sends a message to the server end [client, server].each(&:take) # Wait until they both finish
This will output something like:
Server starts: #<Ractor:#2 server test.rb:1 running> Server sends: ping Client starts: #<Ractor:#3 test.rb:8 running> Client received from #<Ractor:#2 server test.rb:1 blocking>: ping Client sends to #<Ractor:#2 server test.rb:1 blocking>: pong Server received: pong
Ractors receive their messages via the incoming port, and send them to the outgoing port. Either one can be disabled with Ractor#close_incoming and Ractor#close_outgoing, respectively. When a ractor terminates, its ports are closed automatically.
Shareable and unshareable objects
When an object is sent to and from a ractor, it’s important to understand whether the object is shareable or unshareable. Most Ruby objects are unshareable objects. Even frozen objects can be unshareable if they contain (through their instance variables) unfrozen objects.
Shareable objects are those which can be used by several threads without compromising thread-safety, for example numbers, true and false. Ractor.shareable? allows you to check this, and Ractor.make_shareable tries to make the object shareable if it’s not already, and gives an error if it can’t do it.
Ractor.shareable?(1) #=> true -- numbers and other immutable basic values are shareable Ractor.shareable?('foo') #=> false, unless the string is frozen due to # frozen_string_literal: true Ractor.shareable?('foo'.freeze) #=> true Ractor.shareable?([Object.new].freeze) #=> false, inner object is unfrozen ary = ['hello', 'world'] ary.frozen? #=> false ary[0].frozen? #=> false Ractor.make_shareable(ary) ary.frozen? #=> true ary[0].frozen? #=> true ary[1].frozen? #=> true
When a shareable object is sent (via send or Ractor.yield), no additional processing occurs on it. It just becomes usable by both ractors. When an unshareable object is sent, it can be either copied or moved. The first is the default, and it copies the object fully by deep cloning (Object#clone) the non-shareable parts of its structure.
data = ['foo', 'bar'.freeze] r = Ractor.new do data2 = Ractor.receive puts "In ractor: #{data2.object_id}, #{data2[0].object_id}, #{data2[1].object_id}" end r.send(data) r.take puts "Outside : #{data.object_id}, #{data[0].object_id}, #{data[1].object_id}"
This will output something like:
In ractor: 340, 360, 320 Outside : 380, 400, 320
Note that the object ids of the array and the non-frozen string inside the array have changed in the ractor because they are different objects. The second array’s element, which is a shareable frozen string, is the same object.
Deep cloning of objects may be slow, and sometimes impossible. Alternatively, move: true may be used during sending. This will move the unshareable object to the receiving ractor, making it inaccessible to the sending ractor.
data = ['foo', 'bar'] r = Ractor.new do data_in_ractor = Ractor.receive puts "In ractor: #{data_in_ractor.object_id}, #{data_in_ractor[0].object_id}" end r.send(data, move: true) r.take puts "Outside: moved? #{Ractor::MovedObject === data}" puts "Outside: #{data.inspect}"
This will output:
In ractor: 100, 120 Outside: moved? true test.rb:9:in `method_missing': can not send any methods to a moved object (Ractor::MovedError)
Notice that even inspect (and more basic methods like __id__) is inaccessible on a moved object.
Class and Module objects are shareable so the class/module definitions are shared between ractors. Ractor objects are also shareable. All operations on shareable objects are thread-safe, so the thread-safety property will be kept. We can not define mutable shareable objects in Ruby, but C extensions can introduce them.
It is prohibited to access (get) instance variables of shareable objects in other ractors if the values of the variables aren’t shareable. This can occur because modules/classes are shareable, but they can have instance variables whose values are not. In non-main ractors, it’s also prohibited to set instance variables on classes/modules (even if the value is shareable).
class C class << self attr_accessor :tricky end end C.tricky = "unshareable".dup r = Ractor.new(C) do |cls| puts "I see #{cls}" puts "I can't see #{cls.tricky}" cls.tricky = true # doesn't get here, but this would also raise an error end r.take # I see C # can not access instance variables of classes/modules from non-main Ractors (RuntimeError)
Ractors can access constants if they are shareable. The main Ractor is the only one that can access non-shareable constants.
GOOD = 'good'.freeze BAD = 'bad'.dup r = Ractor.new do puts "GOOD=#{GOOD}" puts "BAD=#{BAD}" end r.take # GOOD=good # can not access non-shareable objects in constant Object::BAD by non-main Ractor. (NameError) # Consider the same C class from above r = Ractor.new do puts "I see #{C}" puts "I can't see #{C.tricky}" end r.take # I see C # can not access instance variables of classes/modules from non-main Ractors (RuntimeError)
See also the description of # shareable_constant_value pragma in Comments syntax explanation.
Ractors vs threads
Each ractor has its own main Thread. New threads can be created from inside ractors (and, on CRuby, they share the GVL with other threads of this ractor).
r = Ractor.new do a = 1 Thread.new {puts "Thread in ractor: a=#{a}"}.join end r.take # Here "Thread in ractor: a=1" will be printed
Note on code examples
In the examples below, sometimes we use the following method to wait for ractors that are not currently blocked to finish (or to make progress).
def wait sleep(0.1) end
It is **only for demonstration purposes** and shouldn’t be used in a real code. Most of the time, take is used to wait for ractors to finish.
Reference
See Ractor design doc for more details.
Class methods 15
Instance methods 13
Ruby version changes 5
-
3.4
-
requireinRactoris allowed. The requiring process will be run on the mainRactor.Ractor._require(feature)is added to run requiring process on the mainRactor. [Feature #20627] -
Ractor.main?is added. [Feature #20627] -
Ractor.[]andRactor.[]=are added to access the ractor local storage of the currentRactor. [Feature #20715] -
Ractor.store_if_absent(key){ init }is added to initialize ractor local variables in thread-safety. [Feature #20875]
-
-
3.0
-
New class added to enable parallel execution. See ractor.md for more details.
-