Merge pull request #336 from ianks/atomic-ref-try

AtomicReference: update try_update API (resolves #330)

Author: jdantonio

CHANGELOG.md

@@ -1,5 +1,10 @@
 ### Next Release v0.9.0 (Target Date: 7 June 2015)
 
+
+* Updated `AtomicReference`
+  - `AtomicReference#try_update` now simply returns instead of raising exception
+  - `AtomicReference#try_update!` was added to raise exceptions if an update
+    fails. Note: this is the same behavior as the old `try_update`
 * Pure Java implementations of
   - `AtomicBoolean`
   - `AtomicFixnum`
@@ -58,7 +63,7 @@
   - `Channel`
   - `Exchanger`
   - `LazyRegister`
-  - **new Future Framework** <http://ruby-concurrency.github.io/concurrent-ruby/Concurrent/Edge.html> - unified 
+  - **new Future Framework** <http://ruby-concurrency.github.io/concurrent-ruby/Concurrent/Edge.html> - unified
     implementation of Futures and Promises which combines Features of previous `Future`,
     `Promise`, `IVar`, `Event`, `Probe`, `dataflow`, `Delay`, `TimerTask` into single framework. It uses extensively
     new synchronization layer to make all the paths **lock-free** with exception of blocking threads on `#wait`.
@@ -80,7 +85,7 @@
   - Add AbstractContext#default_executor to be able to override executor class wide
   - Add basic IO example
   - Documentation somewhat improved
-  - All messages should have same priority. It's now possible to send `actor << job1 << job2 << :terminate!` and 
+  - All messages should have same priority. It's now possible to send `actor << job1 << job2 << :terminate!` and
     be sure that both jobs are processed first.
 * Refactored `Channel` to use newer synchronization objects
 * Added `#reset` and `#cancel` methods to `TimerSet`
@@ -162,7 +167,7 @@ Please see the [roadmap](https://github.com/ruby-concurrency/concurrent-ruby/iss
   - `SerializedExecutionDelegator` for serializing *any* executor
 * Updated `Async` with serialized execution
 * Updated `ImmediateExecutor` and `PerThreadExecutor` with full executor service lifecycle
-* Added a `Delay` to root `Actress` initialization 
+* Added a `Delay` to root `Actress` initialization
 * Minor bug fixes to thread pools
 * Refactored many intermittently failing specs
 * Removed Java interop warning `executor.rb:148 warning: ambiguous Java methods found, using submit(java.lang.Runnable)`

lib/concurrent/atomic_reference/direct_update.rb

@@ -13,7 +13,7 @@ module AtomicDirectUpdate
     # Pass the current value to the given block, replacing it
     # with the block's result. May retry if the value changes
     # during the block's execution.
-    # 
+    #
     # @yield [Object] Calculate a new value for the atomic reference using
     #   given (old) value
     # @yieldparam [Object] old_value the starting value of the atomic reference
@@ -27,17 +27,45 @@ def update
     # @!macro [attach] atomic_reference_method_try_update
     #
     # Pass the current value to the given block, replacing it
+    # with the block's result. Return nil if the update fails.
+    #
+    # @yield [Object] Calculate a new value for the atomic reference using
+    #   given (old) value
+    # @yieldparam [Object] old_value the starting value of the atomic reference
+    #
+    # @note This method was altered to avoid raising an exception by default.
+    # Instead, this method now returns `nil` in case of failure. For more info,
+    # please see: https://github.com/ruby-concurrency/concurrent-ruby/pull/336
+    #
+    # @return [Object] the new value, or nil if update failed
+    def try_update
+      old_value = get
+      new_value = yield old_value
+
+      return unless compare_and_set old_value, new_value
+
+      new_value
+    end
+
+    # @!macro [attach] atomic_reference_method_try_update!
+    #
+    # Pass the current value to the given block, replacing it
     # with the block's result. Raise an exception if the update
     # fails.
-    # 
+    #
     # @yield [Object] Calculate a new value for the atomic reference using
     #   given (old) value
     # @yieldparam [Object] old_value the starting value of the atomic reference
     #
+    # @note This behavior mimics the behavior of the original
+    # `AtomicReference#try_update` API. The reason this was changed was to
+    # avoid raising exceptions (which are inherently slow) by default. For more
+    # info: https://github.com/ruby-concurrency/concurrent-ruby/pull/336
+    #
     # @return [Object] the new value
     #
     # @raise [Concurrent::ConcurrentUpdateError] if the update fails
-    def try_update
+    def try_update!
       old_value = get
       new_value = yield old_value
       unless compare_and_set(old_value, new_value)

spec/concurrent/atomic/atomic_reference_spec.rb

@@ -33,6 +33,15 @@
     expect(res).to eq 1001
   end
 
+  specify :test_try_update_bang do
+    # use a number outside JRuby's fixnum cache range, to ensure identity is preserved
+    atomic = described_class.new(1000)
+    res = atomic.try_update! {|v| v + 1}
+
+    expect(atomic.value).to eq 1001
+    expect(res).to eq 1001
+  end
+
   specify :test_swap do
     atomic = described_class.new(1000)
     res = atomic.swap(1001)
@@ -42,12 +51,21 @@
   end
 
   specify :test_try_update_fails do
+    # use a number outside JRuby's fixnum cache range, to ensure identity is preserved
+    atomic = described_class.new(1000)
+    expect(
+      # assigning within block exploits implementation detail for test
+      atomic.try_update {|v| atomic.value = 1001 ; v + 1}
+    ).to be_falsey
+  end
+
+  specify :test_try_update_bang_fails do
     # use a number outside JRuby's fixnum cache range, to ensure identity is preserved
     atomic = described_class.new(1000)
     expect {
       # assigning within block exploits implementation detail for test
-      atomic.try_update{|v| atomic.value = 1001 ; v + 1}
-    }.to raise_error(Concurrent::ConcurrentUpdateError)
+      atomic.try_update! {|v| atomic.value = 1001 ; v + 1}
+    }.to raise_error Concurrent::ConcurrentUpdateError
   end
 
   specify :test_update_retries do