ruby-concurrency/concurrent-ruby

View on GitHub
docs-source/cancellation.in.md

Summary

Maintainability
Test Coverage

## Examples

**Run async task until cancelled**

Create cancellation and then run work in a background thread until it is cancelled. 

```ruby
cancellation, origin = Concurrent::Cancellation.new
# - origin is used for cancelling, resolve it to cancel 
# - cancellation is passed down to tasks for cooperative cancellation
async_task = Concurrent::Promises.future(cancellation) do |cancellation|
  # Do work repeatedly until it is cancelled
  do_stuff until cancellation.canceled?
  :stopped_gracefully
end

sleep 0.01
# Wait a bit then stop the thread by resolving the origin of the cancellation
origin.resolve 
async_task.value!
```

Or let it raise an error.

```ruby
cancellation, origin = Concurrent::Cancellation.new
async_task = Concurrent::Promises.future(cancellation) do |cancellation|
  # Do work repeatedly until it is cancelled
  while true
    cancellation.check!     
    do_stuff 
  end
end

sleep 0.01
# Wait a bit then stop the thread by resolving the origin of the cancellation
origin.resolve 
async_task.result
```

**Run additional tasks on Cancellation** 

Cancellation can also be used to log or plan re-execution.
  
```ruby
cancellation.origin.chain do
  # This block is executed after the Cancellation is cancelled  
  # It can then log cancellation or e.g. plan new re-execution
end
```

**Run only for limited time – Timeout replacement**

Execute task for a given time then finish. 
Instead of letting Cancellation crate its own origin, it can be passed in as argument.
The passed in origin is scheduled to be resolved in given time which then cancels the Cancellation.

```ruby
timeout = Concurrent::Cancellation.new Concurrent::Promises.schedule(0.02)
# or using shortcut helper method
timeout = Concurrent::Cancellation.timeout 0.02 
count   = Concurrent::AtomicFixnum.new
Concurrent.global_io_executor.post(timeout) do |timeout|
  # do stuff until cancelled  
  count.increment until timeout.canceled?
end #

timeout.origin.wait
count.value                              # => 177576
```

**Parallel background processing with single cancellation**

Each task tries to count to 1000 but there is a randomly failing test. The
tasks share single cancellation, when one of them fails it cancels the others.
The failing tasks ends with an error, the other tasks are gracefully cancelled.

```ruby
cancellation, origin = Concurrent::Cancellation.new
tasks = 4.times.map do |i|
  Concurrent::Promises.future(cancellation, origin, i) do |cancellation, origin, i|
    count = 0
    100.times do
      break count = :cancelled if cancellation.canceled?
      count += 1
      sleep 0.001
      if rand > 0.95
        origin.resolve # cancel
        raise 'random error'
      end
      count
    end
  end
end
Concurrent::Promises.zip(*tasks).result #
# => [false,
#     [:cancelled, nil, :cancelled, :cancelled],
#     [nil, #<RuntimeError: random error>, nil, nil]]
```

Without the randomly failing part it produces following.

```ruby
cancellation, origin = Concurrent::Cancellation.new
tasks = 4.times.map do |i|
  Concurrent::Promises.future(cancellation, origin, i) do |cancellation, origin, i|
    count = 0
    100.times do
      break count = :cancelled if cancellation.canceled?
      count += 1
      sleep 0.001
      # if rand > 0.95
      #   origin.resolve
      #   raise 'random error'
      # end
      count
    end
  end
end
Concurrent::Promises.zip(*tasks).result
```

**Combine cancellations**

The combination created by joining two cancellations cancels when the first or the other does.

```ruby
cancellation_a, origin_a = Concurrent::Cancellation.new
cancellation_b, origin_b = Concurrent::Cancellation.new
combined_cancellation    = cancellation_a.join(cancellation_b)

origin_a.resolve

cancellation_a.canceled?
cancellation_b.canceled?
combined_cancellation.canceled?
```

If a different rule for joining is needed, the source can be combined manually.
The manually created cancellation cancels when both the first and the other cancels.

```ruby
cancellation_a, origin_a = Concurrent::Cancellation.new
cancellation_b, origin_b = Concurrent::Cancellation.new
# cancels only when both a and b is cancelled
combined_cancellation    = Concurrent::Cancellation.new origin_a & origin_b

origin_a.resolve

cancellation_a.canceled?        #=> true
cancellation_b.canceled?        #=> false
combined_cancellation.canceled? #=> false

origin_b.resolve
combined_cancellation.canceled? #=> true
```