Ractor
is a Actor-model abstraction for Ruby that provides thread-safe parallel execution.
Ractor.new
can make a new Ractor
, and it will 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!" would be printed
Ractors do not share usual objects, so the same kinds of thread-safety concerns such as data-race, race-conditions are not available on multi-ractor programming.
To achieve this, ractors severely limit object sharing between different ractors. For example, unlike threads, ractors can’t access each other’s objects, nor any objects through variables of the outer scope.
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).)
On CRuby (the default implementation), Global Virtual Machine Lock (GVL) is held per ractor, so ractors are performed in parallel without locking each other.
Instead of accessing the shared state, the objects should be passed to and from ractors via sending and receiving objects as messages.
a = 1 r = Ractor.new do a_in_ractor = receive # receive blocks till somebody will pass 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" would be printed
There are two pairs of methods for sending/receiving messages:
-
Ractor#send
andRactor.receive
for when the sender knows the receiver (push); -
Ractor.yield
andRactor#take
for when the receiver knows the sender (pull);
In addition to that, an argument to Ractor.new
would be passed to block and available there as if received by Ractor.receive
, and the last block value would be sent outside of the ractor as if sent by Ractor.yield
.
A little demonstration on a classic ping-pong:
server = Ractor.new 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 inside client, and available as srv puts "Client starts: #{self.inspect}" received = srv.take # The Client takes a message specifically 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 specifically to the server end [client, server].each(&:take) # Wait till they both finish
This will output:
Server starts: #<Ractor:#2 test.rb:1 running> Server sends: ping Client starts: #<Ractor:#3 test.rb:8 running> Client received from #<Ractor:#2 rac.rb:1 blocking>: ping Client sends to #<Ractor:#2 rac.rb:1 blocking>: pong Server received: pong
It is said that Ractor
receives messages via the incoming port, and sends them to the outgoing port. Either one can be disabled with Ractor#close_incoming
and Ractor#close_outgoing
respectively. If a ractor terminated, its ports will be closed automatically.
Shareable and unshareable objects
When the object is sent to and from the ractor, it is important to understand whether the object is shareable or unshareable. Most of objects are unshareable objects.
Shareable objects are basically those which can be used by several threads without compromising thread-safety; e.g. immutable ones. Ractor.shareable?
allows to check this, and Ractor.make_shareable
tries to make object shareable if it is not.
Ractor.shareable?(1) #=> true -- numbers and other immutable basic values are Ractor.shareable?('foo') #=> false, unless the string is frozen due to # freeze_string_literals: true Ractor.shareable?('foo'.freeze) #=> true 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 happens, and 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 makes the object’s full copy by deep cloning of 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:
In ractor: 340, 360, 320 Outside : 380, 400, 320
(Note that object id of both array and non-frozen string inside array have changed inside the ractor, showing it is different objects. But the second array’s element, which is a shareable frozen string, has the same object_id.)
Deep cloning of the objects may be slow, and sometimes impossible. Alternatively, move: true
may be used on sending. This will move the object to the receiving ractor, making it inaccessible for a 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.
Besides frozen objects, there are shareable objects. Class
and Module
objects are shareable so the Class/Module definitions are shared between ractors. Ractor
objects are also shareable objects. All operations for the shareable mutable 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 instance variables of mutable shareable objects (especially Modules and classes) from ractors other than main:
class C class << self attr_accessor :tricky end end C.tricky = 'test' r = Ractor.new(C) do |cls| puts "I see #{cls}" puts "I can't see #{cls.tricky}" 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' 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 creates its own thread. New threads can be created from inside ractor (and, on CRuby, sharing 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 examples below, sometimes we use the following method to wait till ractors that are not currently blocked will finish (or process till next blocking) method.
def wait sleep(0.1) end
It is **only for demonstration purposes** and shouldn’t be used in a real code. Most of the times, just take
is used to wait till ractor will finish.
Reference
See Ractor design doc for more details.
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 291
def self.count
__builtin_cexpr! %q{
ULONG2NUM(GET_VM()->ractor.cnt);
}
end
Returns total count of Ractors currently running.
Ractor.count #=> 1 r = Ractor.new(name: 'example') { Ractor.yield(1) } Ractor.count #=> 2 (main + example ractor) r.take # wait for Ractor.yield(1) r.take # wait till r will finish Ractor.count #=> 1
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 277
def self.current
__builtin_cexpr! %q{
rb_ractor_self(rb_ec_ractor_ptr(ec));
}
end
Returns the currently executing Ractor
.
Ractor.current #=> #<Ractor:#1 running>
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 837
def self.main
__builtin_cexpr! %q{
rb_ractor_self(GET_VM()->ractor.main_ractor);
}
end
returns main ractor
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 814
def self.make_shareable obj, copy: false
if copy
__builtin_cexpr! %q{
rb_ractor_make_shareable_copy(obj);
}
else
__builtin_cexpr! %q{
rb_ractor_make_shareable(obj);
}
end
end
Make obj
shareable between ractors.
obj
and all the objects it refers to will be frozen, unless they are already shareable.
If copy
keyword is true
, the method will copy objects before freezing them This is safer option but it can take be slower.
Note that the specification and implementation of this method are not mature and may be changed in the future.
obj = ['test'] Ractor.shareable?(obj) #=> false Ractor.make_shareable(obj) #=> ["test"] Ractor.shareable?(obj) #=> true obj.frozen? #=> true obj[0].frozen? #=> true # Copy vs non-copy versions: obj1 = ['test'] obj1s = Ractor.make_shareable(obj1) obj1.frozen? #=> true obj1s.object_id == obj1.object_id #=> true obj2 = ['test'] obj2s = Ractor.make_shareable(obj2, copy: true) obj2.frozen? #=> false obj2s.frozen? #=> true obj2s.object_id == obj2.object_id #=> false obj2s[0].object_id == obj2[0].object_id #=> false
See also the “Shareable and unshareable objects” section in the Ractor
class docs.
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 262
def self.new(*args, name: nil, &block)
b = block # TODO: builtin bug
raise ArgumentError, "must be called with a block" unless block
if __builtin_cexpr!("RBOOL(ruby_single_main_ractor)")
warn("Ractor is experimental, and the behavior may change in future versions of Ruby! " \
"Also there are many implementation issues.", uplevel: 0, category: :experimental)
end
loc = caller_locations(1, 1).first
loc = "#{loc.path}:#{loc.lineno}"
__builtin_ractor_create(loc, name, args, b)
end
Create a new Ractor
with args and a block.
A block (Proc
) will be isolated (can’t access to outer variables). self
inside the block will refer to the current Ractor
.
r = Ractor.new { puts "Hi, I am #{self.inspect}" } r.take # Prints "Hi, I am #<Ractor:#2 test.rb:1 running>"
args
passed to the method would be propagated to block args by the same rules as objects passed through send
/Ractor.receive: if args
are not shareable, they will be copied (via deep cloning, which might be inefficient).
arg = [1, 2, 3] puts "Passing: #{arg} (##{arg.object_id})" r = Ractor.new(arg) {|received_arg| puts "Received: #{received_arg} (##{received_arg.object_id})" } r.take # Prints: # Passing: [1, 2, 3] (#280) # Received: [1, 2, 3] (#300)
Ractor’s name
can be set for debugging purposes:
r = Ractor.new(name: 'my ractor') {} p r #=> #<Ractor:#3 my ractor test.rb:1 terminated>
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 419
def self.receive
__builtin_cexpr! %q{
ractor_receive(ec, rb_ec_ractor_ptr(ec))
}
end
Receive an incoming message from the current Ractor’s incoming port’s queue, which was sent there by send
.
r = Ractor.new do v1 = Ractor.receive puts "Received: #{v1}" end r.send('message1') r.take # Here will be printed: "Received: message1"
Alternatively, private instance method receive
may be used:
r = Ractor.new do v1 = receive puts "Received: #{v1}" end r.send('message1') r.take # Here will be printed: "Received: message1"
The method blocks if the queue is empty.
r = Ractor.new do puts "Before first receive" v1 = Ractor.receive puts "Received: #{v1}" v2 = Ractor.receive puts "Received: #{v2}" end wait puts "Still not received" r.send('message1') wait puts "Still received only one" r.send('message2') r.take
Output:
Before first receive Still not received Received: message1 Still received only one Received: message2
If close_incoming
was called on the ractor, the method raises Ractor::ClosedError
if there are no more messages in incoming queue:
Ractor.new do close_incoming receive end wait # in `receive': The incoming port is already closed => #<Ractor:#2 test.rb:1 running> (Ractor::ClosedError)
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 497
def self.receive_if &b
Primitive.ractor_receive_if b
end
Receive only a specific message.
Instead of Ractor.receive
, Ractor.receive_if
can provide a pattern by a block and you can choose the receiving message.
r = Ractor.new do p Ractor.receive_if{|msg| msg.match?(/foo/)} #=> "foo3" p Ractor.receive_if{|msg| msg.match?(/bar/)} #=> "bar1" p Ractor.receive_if{|msg| msg.match?(/baz/)} #=> "baz2" end r << "bar1" r << "baz2" r << "foo3" r.take
This will output:
foo3 bar1 baz2
If the block returns a truthy value, the message will be removed from the incoming queue and returned. Otherwise, the message remains in the incoming queue and the following received messages are checked by the given block.
If there are no messages left in the incoming queue, the method will block until new messages arrive.
If the block is escaped by break/return/exception/throw, the message is removed from the incoming queue as if a truthy value had been returned.
r = Ractor.new do val = Ractor.receive_if{|msg| msg.is_a?(Array)} puts "Received successfully: #{val}" end r.send(1) r.send('test') wait puts "2 non-matching sent, nothing received" r.send([1, 2, 3]) wait
Prints:
2 non-matching sent, nothing received Received successfully: [1, 2, 3]
Note that you can not call receive/receive_if in the given block recursively. It means that you should not do any tasks in the block.
Ractor.current << true Ractor.receive_if{|msg| Ractor.receive} #=> `receive': can not call receive/receive_if recursively (Ractor::Error)
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 346
def self.select(*ractors, yield_value: yield_unspecified = true, move: false)
raise ArgumentError, 'specify at least one ractor or `yield_value`' if yield_unspecified && ractors.empty?
__builtin_cstmt! %q{
const VALUE *rs = RARRAY_CONST_PTR_TRANSIENT(ractors);
VALUE rv;
VALUE v = ractor_select(ec, rs, RARRAY_LENINT(ractors),
yield_unspecified == Qtrue ? Qundef : yield_value,
(bool)RTEST(move) ? true : false, &rv);
return rb_ary_new_from_args(2, rv, v);
}
end
Waits for the first ractor to have something in its outgoing port, reads from this ractor, and returns that ractor and the object received.
r1 = Ractor.new {Ractor.yield 'from 1'} r2 = Ractor.new {Ractor.yield 'from 2'} r, obj = Ractor.select(r1, r2) puts "received #{obj.inspect} from #{r.inspect}" # Prints: received "from 1" from #<Ractor:#2 test.rb:1 running>
If one of the given ractors is the current ractor, and it would be selected, r
will contain :receive
symbol instead of the ractor object.
r1 = Ractor.new(Ractor.current) do |main| main.send 'to main' Ractor.yield 'from 1' end r2 = Ractor.new do Ractor.yield 'from 2' end r, obj = Ractor.select(r1, r2, Ractor.current) puts "received #{obj.inspect} from #{r.inspect}" # Prints: received "to main" from :receive
If yield_value
is provided, that value may be yielded if another Ractor
is calling take
. In this case, the pair [:yield, nil]
would be returned:
r1 = Ractor.new(Ractor.current) do |main| puts "Received from main: #{main.take}" end puts "Trying to select" r, obj = Ractor.select(r1, Ractor.current, yield_value: 123) wait puts "Received #{obj.inspect} from #{r.inspect}"
This will print:
Trying to select Received from main: 123 Received nil from :yield
move
boolean flag defines whether yielded value should be copied (default) or moved.
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 773
def self.shareable? obj
__builtin_cexpr! %q{
RBOOL(rb_ractor_shareable_p(obj));
}
end
Checks if the object is shareable by ractors.
Ractor.shareable?(1) #=> true -- numbers and other immutable basic values are frozen Ractor.shareable?('foo') #=> false, unless the string is frozen due to # freeze_string_literals: true Ractor.shareable?('foo'.freeze) #=> true
See also the “Shareable and unshareable objects” section in the Ractor
class docs.
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 630
def self.yield(obj, move: false)
__builtin_cexpr! %q{
ractor_yield(ec, rb_ec_ractor_ptr(ec), obj, move)
}
end
Send a message to the current ractor’s outgoing port to be consumed by take
.
r = Ractor.new {Ractor.yield 'Hello from ractor'} puts r.take # Prints: "Hello from ractor"
The method is blocking, and will return only when somebody consumes the sent message.
r = Ractor.new do Ractor.yield 'Hello from ractor' puts "Ractor: after yield" end wait puts "Still not taken" puts r.take
This will print:
Still not taken Hello from ractor Ractor: after yield
If the outgoing port was closed with close_outgoing
, the method will raise:
r = Ractor.new do close_outgoing Ractor.yield 'Hello from ractor' end wait # `yield': The outgoing-port is already closed (Ractor::ClosedError)
The meaning of move
argument is the same as for send
.
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 827
def [](sym)
Primitive.ractor_local_value(sym)
end
get a value from ractor-local storage
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 832
def []=(sym, val)
Primitive.ractor_local_value_set(sym, val)
end
set a value in ractor-local storage
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 737
def close_incoming
__builtin_cexpr! %q{
ractor_close_incoming(ec, RACTOR_PTR(self));
}
end
Closes the incoming port and returns its previous state. All further attempts to Ractor.receive
in the ractor, and send
to the ractor will fail with Ractor::ClosedError
.
r = Ractor.new {sleep(500)} r.close_incoming #=> false r.close_incoming #=> true r.send('test') # Ractor::ClosedError (The incoming-port is already closed)
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 756
def close_outgoing
__builtin_cexpr! %q{
ractor_close_outgoing(ec, RACTOR_PTR(self));
}
end
Closes the outgoing port and returns its previous state. All further attempts to Ractor.yield
in the ractor, and take
from the ractor will fail with Ractor::ClosedError
.
r = Ractor.new {sleep(500)} r.close_outgoing #=> false r.close_outgoing #=> true r.take # Ractor::ClosedError (The outgoing-port is already closed)
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 703
def inspect
loc = __builtin_cexpr! %q{ RACTOR_PTR(self)->loc }
name = __builtin_cexpr! %q{ RACTOR_PTR(self)->name }
id = __builtin_cexpr! %q{ INT2FIX(rb_ractor_id(RACTOR_PTR(self))) }
status = __builtin_cexpr! %q{
rb_str_new2(ractor_status_str(RACTOR_PTR(self)->status_))
}
"#<Ractor:##{id}#{name ? ' '+name : ''}#{loc ? " " + loc : ''} #{status}>"
end
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 716
def name
__builtin_cexpr! %q{RACTOR_PTR(self)->name}
end
The name set in Ractor.new
, or nil
.
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 430
def receive
__builtin_cexpr! %q{
ractor_receive(ec, rb_ec_ractor_ptr(ec))
}
end
same as Ractor.receive
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 501
def receive_if &b
Primitive.ractor_receive_if b
end
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 586
def send(obj, move: false)
__builtin_cexpr! %q{
ractor_send(ec, RACTOR_PTR(self), obj, move)
}
end
Send a message to a Ractor’s incoming queue to be consumed by Ractor.receive
.
r = Ractor.new do value = Ractor.receive puts "Received #{value}" end r.send 'message' # Prints: "Received: message"
The method is non-blocking (will return immediately even if the ractor is not ready to receive anything):
r = Ractor.new {sleep(5)} r.send('test') puts "Sent successfully" # Prints: "Sent successfully" immediately
Attempt to send to ractor which already finished its execution will raise Ractor::ClosedError
.
r = Ractor.new {} r.take p r # "#<Ractor:#6 (irb):23 terminated>" r.send('test') # Ractor::ClosedError (The incoming-port is already closed)
If close_incoming
was called on the ractor, the method also raises Ractor::ClosedError
.
r = Ractor.new do sleep(500) receive end r.close_incoming r.send('test') # Ractor::ClosedError (The incoming-port is already closed) # The error would be raised immediately, not when ractor will try to receive
If the obj
is unshareable, by default it would be copied into ractor by deep cloning. If the move: true
is passed, object is moved into ractor and becomes inaccessible to sender.
r = Ractor.new {puts "Received: #{receive}"} msg = 'message' r.send(msg, move: true) r.take p msg
This prints:
Received: message in `p': undefined method `inspect' for #<Ractor::MovedObject:0x000055c99b9b69b8>
All references to the object and its parts will become invalid in sender.
r = Ractor.new {puts "Received: #{receive}"} s = 'message' ary = [s] copy = ary.dup r.send(ary, move: true) s.inspect # Ractor::MovedError (can not send any methods to a moved object) ary.class # Ractor::MovedError (can not send any methods to a moved object) copy.class # => Array, it is different object copy[0].inspect # Ractor::MovedError (can not send any methods to a moved object) # ...but its item was still a reference to `s`, which was moved
If the object was shareable, move: true
has no effect on it:
r = Ractor.new {puts "Received: #{receive}"} s = 'message'.freeze r.send(s, move: true) s.inspect #=> "message", still available
# File tmp/rubies/ruby-3.1.3/ractor.rb, line 697
def take
__builtin_cexpr! %q{
ractor_take(ec, RACTOR_PTR(self))
}
end
Take a message from ractor’s outgoing port, which was put there by Ractor.yield
or at ractor’s finalization.
r = Ractor.new do Ractor.yield 'explicit yield' 'last value' end puts r.take #=> 'explicit yield' puts r.take #=> 'last value' puts r.take # Ractor::ClosedError (The outgoing-port is already closed)
The fact that the last value is also put to outgoing port means that take
can be used as some analog of Thread#join
(“just wait till ractor finishes”), but don’t forget it will raise if somebody had already consumed everything ractor have produced.
If the outgoing port was closed with close_outgoing
, the method will raise Ractor::ClosedError
.
r = Ractor.new do sleep(500) Ractor.yield 'Hello from ractor' end r.close_outgoing r.take # Ractor::ClosedError (The outgoing-port is already closed) # The error would be raised immediately, not when ractor will try to receive
If an uncaught exception is raised in the Ractor
, it is propagated on take as a Ractor::RemoteError
.
r = Ractor.new {raise "Something weird happened"} begin r.take rescue => e p e # => #<Ractor::RemoteError: thrown by remote Ractor.> p e.ractor == r # => true p e.cause # => #<RuntimeError: Something weird happened> end
Ractor::ClosedError
is a descendant of StopIteration
, so the closing of the ractor will break the loops without propagating the error:
r = Ractor.new do 3.times {|i| Ractor.yield "message #{i}"} "finishing" end loop {puts "Received: " + r.take} puts "Continue successfully"
This will print:
Received: message 0 Received: message 1 Received: message 2 Received: finishing Continue successfully