ruby/doc/ractor.md

Ractor - Ruby's Actor-like concurrent abstraction

Ractor is designed to provide a parallel execution feature of Ruby without thread-safety concerns.

Summary

Multiple Ractors in an interpreter process

You can make multiple Ractors and they run in parallel.

• Ractor.new{ expr } creates a new Ractor and expr is run in parallel on a parallel computer.
• Interpreter invokes with the first Ractor (called main Ractor).
• If the main Ractor terminates, all other Ractors receive termination requests, similar to how threads behave. (if main thread (first invoked Thread), Ruby interpreter sends all running threads to terminate execution).
• Each Ractor contains one or more Threads.
  • Threads within the same Ractor share a Ractor-wide global lock like GIL (GVL in MRI terminology), so they can't run in parallel (without releasing GVL explicitly in C-level). Threads in different ractors run in parallel.
  • The overhead of creating a Ractor is similar to overhead of one Thread creation.

Limited sharing between multiple ractors

Ractors don't share everything, unlike threads.

• Most objects are Unshareable objects, so you don't need to care about thread-safety problems which are caused by sharing.
• Some objects are Shareable objects.
  • Immutable objects: frozen objects which don't refer to unshareable-objects.
    • i = 123: i is an immutable object.
    • s = "str".freeze: s is an immutable object.
    • a = [1, [2], 3].freeze: a is not an immutable object because a refers unshareable-object [2] (which is not frozen).
    • h = {c: Object}.freeze: h is an immutable object because h refers Symbol :c and shareable Object class object which is not frozen.
  • Class/Module objects
  • Special shareable objects
    • Ractor object itself.
    • And more...

Communication between Ractors with Ractor::Port

Ractors communicate with each other and synchronize the execution by message exchanging between Ractors. Ractor::Port is provided for this communication.

port = Ractor::Port.new

Ractor.new port do |port|
  # Other ractors can send to the port
  port << 42
end

port.receive # get a message to the port. Only the creator Ractor can receive from the port
#=> 42

Ractors have its own default port and Ractor#send, Ractor.receive will use it.

Copy & Move semantics to send messages

To send unshareable objects as messages, objects are copied or moved.

• Copy: use deep-copy.
• Move: move membership.
  • Sender can not access the moved object after moving the object.
  • Guarantee that at least only 1 Ractor can access the object.

Thread-safety

Ractor helps to write a thread-safe concurrent program, but we can make thread-unsafe programs with Ractors.

• GOOD: Sharing limitation
  • Most objects are unshareable, so we can't make data-racy and race-conditional programs.
  • Shareable objects are protected by an interpreter or locking mechanism.
• BAD: Class/Module can violate this assumption
  • To make it compatible with old behavior, classes and modules can introduce data-race and so on.
  • Ruby programmers should take care if they modify class/module objects on multi Ractor programs.
• BAD: Ractor can't solve all thread-safety problems
  • There are several blocking operations (waiting send) so you can make a program which has dead-lock and live-lock issues.
  • Some kind of shareable objects can introduce transactions (STM, for example). However, misusing transactions will generate inconsistent state.

Without Ractor, we need to trace all state-mutations to debug thread-safety issues. With Ractor, you can concentrate on suspicious code which are shared with Ractors.

Creation and termination

Ractor.new

• Ractor.new{ expr } generates another Ractor.

# Ractor.new with a block creates new Ractor
r = Ractor.new do
  # This block will be run in parallel with other ractors
end

# You can name a Ractor with `name:` argument.
r = Ractor.new name: 'test-name' do
end

# and Ractor#name returns its name.
r.name #=> 'test-name'

Given block isolation

The Ractor executes given expr in a given block. Given block will be isolated from outer scope by the Proc#isolate method (not exposed yet for Ruby users). To prevent sharing unshareable objects between ractors, block outer-variables, self and other information are isolated.

Proc#isolate is called at Ractor creation time (when Ractor.new is called). If given Proc object is not able to isolate because of outer variables and so on, an error will be raised.

begin
  a = true
  r = Ractor.new do
    a #=> ArgumentError because this block accesses `a`.
  end
  r.join # see later
rescue ArgumentError
end

• The self of the given block is the Ractor object itself.

r = Ractor.new do
  p self.class #=> Ractor
  self.object_id
end
r.value == self.object_id #=> false

Passed arguments to Ractor.new() becomes block parameters for the given block. However, an interpreter does not pass the parameter object references, but send them as messages (see below for details).

r = Ractor.new 'ok' do |msg|
  msg #=> 'ok'
end
r.value #=> 'ok'

# almost similar to the last example
r = Ractor.new do
  msg = Ractor.receive
  msg
end
r.send 'ok'
r.value #=> 'ok'

An execution result of given block

Return value of the given block becomes an outgoing message (see below for details).

r = Ractor.new do
  'ok'
end
r.value #=> `ok`

Error in the given block will be propagated to the receiver of an outgoing message.

r = Ractor.new do
  raise 'ok' # exception will be transferred to the receiver
end

begin
  r.value
rescue Ractor::RemoteError => e
  e.cause.class   #=> RuntimeError
  e.cause.message #=> 'ok'
  e.ractor        #=> r
end

Communication between Ractors

Communication between Ractors is achieved by sending and receiving messages. There are two ways to communicate with each other.

• (1) Message sending/receiving via Ractor::Port
• (2) Using shareable container objects
  • Ractor::TVar gem (ko1/ractor-tvar)
  • more?

Users can control program execution timing with (1), but should not control with (2) (only manage as critical section).

For message sending and receiving, there are two types of APIs: push type and pull type.

• (1) send/receive via Ractor::Port.
  • Ractor::Port#send(obj) (Ractor::Port#<<(obj) is an alias) send a message to the port. Ports are connected to the infinite size incoming queue so Ractor::Port#send will never block.
  • Ractor::Port#receive dequeue a message from its own incoming queue. If the incoming queue is empty, Ractor::Port#receive calling will block the execution of a thread.
• Ractor.select() can wait for the success of Ractor::Port#receive.
• You can close Ractor::Port by Ractor::Port#close only by the creator Ractor of the port.
  • If the port is closed, you can't send to the port. If Ractor::Port#receive is blocked for the closed port, then it will raise an exception.
  • When a Ractor is terminated, the Ractor's ports are closed.
• There are 3 ways to send an object as a message
  • (1) Send a reference: Sending a shareable object, send only a reference to the object (fast)
  • (2) Copy an object: Sending an unshareable object by copying an object deeply (slow). Note that you can not send an object which does not support deep copy. Some T_DATA objects (objects whose class is defined in a C extension, such as StringIO) are not supported.
  • (3) Move an object: Sending an unshareable object reference with a membership. Sender Ractor can not access moved objects anymore (raise an exception) after moving it. Current implementation makes new object as a moved object for receiver Ractor and copies references of sending object to moved object. T_DATA objects are not supported.
  • You can choose "Copy" and "Move" by the move: keyword, Ractor#send(obj, move: true/false) and Ractor.yield(obj, move: true/false) (default is false (COPY)).

Wait for multiple Ractors with Ractor.select

You can wait multiple Ractor port's receiving. The return value of Ractor.select() is [port, msg] where port is a ready port and msg is received message.

To make convenient, Ractor.select can also accept Ractors to wait the termination of Ractors. The return value of Ractor.select() is [r, msg] where r is a terminated Ractor and msg is the value of Ractor's block.

Wait for a single ractor (same as Ractor#value):

r1 = Ractor.new{'r1'}

r, obj = Ractor.select(r1)
r == r1 and obj == 'r1' #=> true

Waiting for two ractors:

r1 = Ractor.new{'r1'}
r2 = Ractor.new{'r2'}
rs = [r1, r2]
as = []

# Wait for r1 or r2's Ractor.yield
r, obj = Ractor.select(*rs)
rs.delete(r)
as << obj

# Second try (rs only contain not-closed ractors)
r, obj = Ractor.select(*rs)
rs.delete(r)
as << obj
as.sort == ['r1', 'r2'] #=> true

TODO: Current Ractor.select() has the same issue of select(2), so this interface should be refined.

TODO: select syntax of go-language uses round-robin technique to make fair scheduling. Now Ractor.select() doesn't use it.

Closing Ractor's ports

• Ractor::Port#close close the ports (similar to Queue#close).
  • port.send(obj) where port is closed, will raise an exception.
  • When the queue connected to the port is empty and port is closed, Ractor::Port#receive raises an exception. If the queue is not empty, it dequeues an object without exceptions.
• When a Ractor terminates, the ports are closed automatically.

Example (try to get a result from closed Ractor):

r = Ractor.new do
  'finish'
end
r.join # success (wait for the termination)
r.value # success (will return 'finish')

# the first Ractor which success the `Ractor#value` can get the result
Ractor.new r do |r|
  r.value #=> Ractor::Error
end

Example (try to send to closed (terminated) Ractor):

r = Ractor.new do
end

r.join # wait terminate

begin
  r.send(1)
rescue Ractor::ClosedError
  'ok'
else
  'ng'
end

Send a message by copying

Ractor::Port#send(obj) copy obj deeply if obj is an unshareable object.

obj = 'str'.dup
r = Ractor.new obj do |msg|
  # return received msg's object_id
  msg.object_id
end

obj.object_id == r.value #=> false

Some objects are not supported to copy the value, and raise an exception.

obj = Thread.new{}
begin
  Ractor.new obj do |msg|
    msg
  end
rescue TypeError => e
  e.message #=> #<TypeError: allocator undefined for Thread>
else
  'ng' # unreachable here
end

Send a message by moving

Ractor::Port#send(obj, move: true) moves obj to the destination Ractor. If the source Ractor touches the moved object (for example, call the method like obj.foo()), it will be an error.

# move with Ractor#send
r = Ractor.new do
  obj = Ractor.receive
  obj << ' world'
end

str = 'hello'
r.send str, move: true
modified = r.value #=> 'hello world'

# str is moved, and accessing str from this Ractor is prohibited

begin
  # Error because it touches moved str.
  str << ' exception' # raise Ractor::MovedError
rescue Ractor::MovedError
  modified #=> 'hello world'
else
  raise 'unreachable'
end

Some objects are not supported to move, and an exception will be raised.

r = Ractor.new do
  Ractor.receive
end

r.send(Thread.new{}, move: true) #=> allocator undefined for Thread (TypeError)

To achieve the access prohibition for moved objects, class replacement technique is used to implement it.

Shareable objects

The following objects are shareable.

• Immutable objects
  • Small integers, some symbols, true, false, nil (a.k.a. SPECIAL_CONST_P() objects in internal)
  • Frozen native objects
    • Numeric objects: Float, Complex, Rational, big integers (T_BIGNUM in internal)
    • All Symbols.
  • Frozen String and Regexp objects (their instance variables should refer only shareable objects)
• Class, Module objects (T_CLASS, T_MODULE and T_ICLASS in internal)
• Ractor and other special objects which care about synchronization.

Implementation: Now shareable objects (RVALUE) have FL_SHAREABLE flag. This flag can be added lazily.

To make shareable objects, Ractor.make_shareable(obj) method is provided. In this case, try to make shareable by freezing obj and recursively traversable objects. This method accepts copy: keyword (default value is false).Ractor.make_shareable(obj, copy: true) tries to make a deep copy of obj and make the copied object shareable.

Language changes to isolate unshareable objects between Ractors

To isolate unshareable objects between Ractors, we introduced additional language semantics on multi-Ractor Ruby programs.

Note that without using Ractors, these additional semantics is not needed (100% compatible with Ruby 2).

Global variables

Only the main Ractor (a Ractor created at starting of interpreter) can access global variables.

$gv = 1
r = Ractor.new do
  $gv
end

begin
  r.join
rescue Ractor::RemoteError => e
  e.cause.message #=> 'can not access global variables from non-main Ractors'
end

Note that some special global variables, such as $stdin, $stdout and $stderr are Ractor-local. See [Bug #17268] for more details.

Instance variables of shareable objects

Instance variables of classes/modules can be get from non-main Ractors if the referring values are shareable objects.

class C
  @iv = 1
end

p Ractor.new do
  class C
     @iv
  end
end.value #=> 1

Otherwise, only the main Ractor can access instance variables of shareable objects.

class C
  @iv = [] # unshareable object
end

Ractor.new do
  class C
    begin
      p @iv
    rescue Ractor::IsolationError
      p $!.message
      #=> "can not get unshareable values from instance variables of classes/modules from non-main Ractors"
    end

    begin
      @iv = 42
    rescue Ractor::IsolationError
      p $!.message
      #=> "can not set instance variables of classes/modules by non-main Ractors"
    end
  end
end.join

shared = Ractor.new{}
shared.instance_variable_set(:@iv, 'str')

r = Ractor.new shared do |shared|
  p shared.instance_variable_get(:@iv)
end

begin
  r.join
rescue Ractor::RemoteError => e
  e.cause.message #=> can not access instance variables of shareable objects from non-main Ractors (Ractor::IsolationError)
end

Note that instance variables for class/module objects are also prohibited on Ractors.

Class variables

Only the main Ractor can access class variables.

class C
  @@cv = 'str'
end

r = Ractor.new do
  class C
    p @@cv
  end
end


begin
  r.join
rescue => e
  e.class #=> Ractor::IsolationError
end

Constants

Only the main Ractor can read constants which refer to the unshareable object.

class C
  CONST = 'str'
end
r = Ractor.new do
  C::CONST
end
begin
  r.join
rescue => e
  e.class #=> Ractor::IsolationError
end

Only the main Ractor can define constants which refer to the unshareable object.

class C
end
r = Ractor.new do
  C::CONST = 'str'
end
begin
  r.join
rescue => e
  e.class #=> Ractor::IsolationError
end

To make multi-ractor supported library, the constants should only refer shareable objects.

TABLE = {a: 'ko1', b: 'ko2', c: 'ko3'}

In this case, TABLE references an unshareable Hash object. So that other ractors can not refer TABLE constant. To make it shareable, we can use Ractor.make_shareable() like that.

TABLE = Ractor.make_shareable( {a: 'ko1', b: 'ko2', c: 'ko3'} )

To make it easy, Ruby 3.0 introduced new shareable_constant_value Directive.

# shareable_constant_value: literal

TABLE = {a: 'ko1', b: 'ko2', c: 'ko3'}
#=> Same as: TABLE = Ractor.make_shareable( {a: 'ko1', b: 'ko2', c: 'ko3'} )

shareable_constant_value directive accepts the following modes (descriptions use the example: CONST = expr):

• none: Do nothing. Same as: CONST = expr
• literal:
  • if expr consists of literals, replaced to CONST = Ractor.make_shareable(expr).
  • otherwise: replaced to CONST = expr.tap{|o| raise unless Ractor.shareable?(o)}.
• experimental_everything: replaced to CONST = Ractor.make_shareable(expr).
• experimental_copy: replaced to CONST = Ractor.make_shareable(expr, copy: true).

Except the none mode (default), it is guaranteed that the assigned constants refer to only shareable objects.

See doc/syntax/comments.rdoc for more details.

Implementation note

• Each Ractor has its own thread, it means each Ractor has at least 1 native thread.
• Each Ractor has its own ID (rb_ractor_t::pub::id).
  • On debug mode, all unshareable objects are labeled with current Ractor's id, and it is checked to detect unshareable object leak (access an object from different Ractor) in VM.

Examples

Traditional Ring example in Actor-model

RN = 1_000
CR = Ractor.current

r = Ractor.new do
  p Ractor.receive
  CR << :fin
end

RN.times{
  r = Ractor.new r do |next_r|
    next_r << Ractor.receive
  end
}

p :setup_ok
r << 1
p Ractor.receive

Fork-join

def fib n
  if n < 2
    1
  else
    fib(n-2) + fib(n-1)
  end
end

RN = 10
rs = (1..RN).map do |i|
  Ractor.new i do |i|
    [i, fib(i)]
  end
end

until rs.empty?
  r, v = Ractor.select(*rs)
  rs.delete r
  p answer: v
end

Worker pool

(1) One ractor has a pool

require 'prime'

N = 1000
RN = 10

# make RN workers
workers = (1..RN).map do
  Ractor.new do |; result_port|
    loop do
      n, result_port = Ractor.receive
      result_port << [n, n.prime?, Ractor.current]
    end
  end
end

result_port = Ractor::Port.new
results = []

(1..N).each do |i|
  if workers.empty?
    # receive a result
    n, result, w = result_port.receive
    results << [n, result]
  else
    w = workers.pop
  end

  # send a task to the idle worker ractor
  w << [i, result_port]
end

# receive a result
while results.size != N
  n, result, _w = result_port.receive
  results << [n, result]
end

pp results.sort_by{|n, result| n}

Pipeline

# pipeline with send/receive

r3 = Ractor.new Ractor.current do |cr|
  cr.send Ractor.receive + 'r3'
end

r2 = Ractor.new r3 do |r3|
  r3.send Ractor.receive + 'r2'
end

r1 = Ractor.new r2 do |r2|
  r2.send Ractor.receive + 'r1'
end

r1 << 'r0'
p Ractor.receive #=> "r0r1r2r3"

Supervise

# ring example again

r = Ractor.current
(1..10).map{|i|
  r = Ractor.new r, i do |r, i|
    r.send Ractor.receive + "r#{i}"
  end
}

r.send "r0"
p Ractor.receive #=> "r0r10r9r8r7r6r5r4r3r2r1"

# ring example with an error

r = Ractor.current
rs = (1..10).map{|i|
  r = Ractor.new r, i do |r, i|
    loop do
      msg = Ractor.receive
      raise if /e/ =~ msg
      r.send msg + "r#{i}"
    end
  end
}

r.send "r0"
p Ractor.receive #=> "r0r10r9r8r7r6r5r4r3r2r1"
r.send "r0"
p Ractor.select(*rs, Ractor.current) #=> [:receive, "r0r10r9r8r7r6r5r4r3r2r1"]
r.send "e0"
p Ractor.select(*rs, Ractor.current)
#=>
# <Thread:0x000056262de28bd8 run> terminated with exception (report_on_exception is true):
# Traceback (most recent call last):
#         2: from /home/ko1/src/ruby/trunk/test.rb:7:in `block (2 levels) in <main>'
#         1: from /home/ko1/src/ruby/trunk/test.rb:7:in `loop'
# /home/ko1/src/ruby/trunk/test.rb:9:in `block (3 levels) in <main>': unhandled exception
# Traceback (most recent call last):
#         2: from /home/ko1/src/ruby/trunk/test.rb:7:in `block (2 levels) in <main>'
#         1: from /home/ko1/src/ruby/trunk/test.rb:7:in `loop'
# /home/ko1/src/ruby/trunk/test.rb:9:in `block (3 levels) in <main>': unhandled exception
#         1: from /home/ko1/src/ruby/trunk/test.rb:21:in `<main>'
# <internal:ractor>:69:in `select': thrown by remote Ractor. (Ractor::RemoteError)

# resend non-error message

r = Ractor.current
rs = (1..10).map{|i|
  r = Ractor.new r, i do |r, i|
    loop do
      msg = Ractor.receive
      raise if /e/ =~ msg
      r.send msg + "r#{i}"
    end
  end
}

r.send "r0"
p Ractor.receive #=> "r0r10r9r8r7r6r5r4r3r2r1"
r.send "r0"
p Ractor.select(*rs, Ractor.current)
[:receive, "r0r10r9r8r7r6r5r4r3r2r1"]
msg = 'e0'
begin
  r.send msg
  p Ractor.select(*rs, Ractor.current)
rescue Ractor::RemoteError
  msg = 'r0'
  retry
end

#=> <internal:ractor>:100:in `send': The incoming-port is already closed (Ractor::ClosedError)
# because r == r[-1] is terminated.

# ring example with supervisor and re-start

def make_ractor r, i
  Ractor.new r, i do |r, i|
    loop do
      msg = Ractor.receive
      raise if /e/ =~ msg
      r.send msg + "r#{i}"
    end
  end
end

r = Ractor.current
rs = (1..10).map{|i|
  r = make_ractor(r, i)
}

msg = 'e0' # error causing message
begin
  r.send msg
  p Ractor.select(*rs, Ractor.current)
rescue Ractor::RemoteError
  r = rs[-1] = make_ractor(rs[-2], rs.size-1)
  msg = 'x0'
  retry
end

#=> [:receive, "x0r9r9r8r7r6r5r4r3r2r1"]