Need help with dataflow?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

124 Stars 11 Forks MIT License 115 Commits 3 Opened issues


Dataflow concurrency for Ruby (inspired by the Oz language)

Services available


Need anything else?

Contributors list

# 149,855
24 commits
# 337,853
1 commit

h1. What's this?

A Ruby library that adds Dataflow variables (inspired by the Oz language). Dataflow variables have the property that they can only be bound/assigned to once, or have an equivalent value as an existing assignment (see "unification").

Dataflow variables must be declared before they are used, and can be passed around as data without actually being bound. If the variable gets used (in this library this means a method call) while being unbound then the currently executing thread will suspend.

h1. What's the point?

Ruby is Object Oriented (with the ability to mutate local, instance, class, and global variables, and even constants), and on top of that it has powerful reflection and meta-programming abilities. While these features are useful for certain problems, they are not within the declarative model. Staying in the declarative model gives one 2 advantages:

It is easy to reason about what the program does

Simple but powerful concurrency is possible

Ruby, like many other OO languages, is facing the hurdles of taking advantage of the increase of processor cores within a simple parallel programming model. This library lets you program Ruby in the declarative concurrent model when you need to take advantage of multiple cores (assuming a Ruby implementation that uses native threads in one way or another).

The trick to this kind of programming is binding variables from other threads. The nice thing is that many existing libraries/classes/methods can still be used, just avoid side-effects. Use regular Ruby threading to create threads, use "local" or "declare" to create new variables, and use "unify" to bind variables.

h1. Install

To install the latest release as a gem:

sudo gem install dataflow

h1. IRC

#dataflow-gem @

h1. Examples

# Local variables
include Dataflow

local do |x, y, z|
  # notice how the order automatically gets resolved { unify y, x + 2 } { unify z, y + 3 } { unify x, 1 }
  z #=> 6
# Module methods version

Dataflow.local do |x, y, z|
  # notice how the order automatically gets resolved { Dataflow.unify y, x + 2 } { Dataflow.unify z, y + 3 } { Dataflow.unify x, 1 }
  z #=> 6

# Note that a gobal Dataflow.declare is not supported
# Instance variables
class AnimalHouse
  include Dataflow
  declare :small_cat, :big_cat

  def fetch_big_cat { unify big_cat, small_cat.upcase }
    unify small_cat, 'cat'
end #=> 'CAT'
# Data-driven concurrency
include Dataflow

local do |stream, doubles, triples, squares|
  unify stream, { } { unify doubles, {|n| n*2 } } { unify triples, {|n| n*3 } } { unify squares, {|n| n**2 } } { stream.each {|x| unify x, rand(100) } }

  puts "original: #{stream.inspect}"
  puts "doubles:  #{doubles.inspect}"
  puts "triples:  #{triples.inspect}"
  puts "squares:  #{squares.inspect}"  
# By-need trigger laziness
include Dataflow

local do |x, y, z| { unify y, by_need { 4 } } { unify z, x + y } { unify x, by_need { 3 } }
  z #=> 7
# Need-later future expressions
include Dataflow

local do |x, y, z|
  unify y, need_later { 4 }
  unify z, need_later { x + y }
  unify x, need_later { 3 }
  z #=> 7
include Dataflow

# flow without parameters
local do |x|
  flow do 
    # other stuff
    unify x, 1337
  x #=> 1337

# flow with an output parameter
local do |x|
  flow(x) do
    # other stuff
  x #=> 1337
# barrier
include Dataflow

local do |lock1, lock2|
  flow { unify lock1, :unlocked }
  flow { unify lock2, :unlocked }
  barrier lock1, lock2
  puts "Barrier broken!"  
# FutureQueue
include Dataflow

local do |queue, first, second|
  unify queue,
  queue.pop first
  queue.push 1
  queue.push 2
  queue.pop second
  first #=> 1
  second #=> 2

h1. Anonymous variables

Sometimes you may want to pack a data structure with variables that do not need to be referenced with labels. For those cases anonymous variables are a good choice, here are some options:

include Dataflow { } { Dataflow.local } { local }
# and technically not anonymous { local {|v| v } }

h1. Debugging

If you are having trouble and need to debug dataflow variables, simply call #inspect.

If the variable has already been bound, it call inspect on its bound value like normal.However, if the variable is not bound yet then you will get a special string that contains the proxies #id that you can use to track down which proxy objects are being passed around to which parts of your program:

include Dataflow
local do |my_var| 
  my_var.inspect # => #<:variable:2637860 unbound> 

h1. Fork method customization

By default both #flow and #need_later use Thread.fork as their fork method. Youc an access the fork method via Dataflow.forker.

If you would like to use a custom forker, simple set it to an object that responds to #call and internally calls a block passed to it (for an example of a synchronous forker, see spec/forker_spec.rb):

Dataflow.forker = MyClass.method(:fork_with_threadpool)

Also note that #flow is used interally by #need_later, in case you want to override that specifically.

h1. Ports using Dataflow

Ports are an extension of the declarative concurrent model to support nondeterministic behavior. They accomplish this through the use of a single state variable. Ports are also inspired by the Oz language.

An Actor class in the style of Erlang message-passing processes is also provided. It makes use of the asynchronous behavior of ports, but otherwise uses no state variables.

h1. Examples using Ports

include Dataflow

local do |port, stream|
  unify port, {port.send 2} {port.send 8} {port.send 1024}
  stream.take(3).sort #=> [2, 8, 1024]

h1. Examples using Actors

include Dataflow

Ping = {
  3.times {
    case receive
    when "Ping"
      puts "Ping"
      Pong.send "Pong"

Pong = {
  3.times {
    case receive
    when "Pong"
      puts "Pong"
      Ping.send "Ping"
} { Ping.send "Ping" }


h1. Equality

Most Ruby implmentations will not use method calls for equality operations in base types/classes. This means equality between dataflow variables and those base types will not behave as expected. Require the following to get equality on base types that uses method calls, while still passing rubyspec:

require "dataflow/equality"

h1. References

The basis of dataflow variables around a language is not common among popular languages and may be confusing to some. For an in-depth introduction to the Oz language and the techniques used in this library (including byneed triggers, port objects, and comparisons to Erlang message passing) see the book "Concepts, Techniques, and Models of Computer Programming":,Techniques,andModelsofComputer_Programming

h1. Contributors

larrytheliquid, amiller

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.