Class AtomicCounter

java.lang.Object
org.agrona.concurrent.status.AtomicCounter
All Implemented Interfaces:
AutoCloseable

public class AtomicCounter extends Object implements AutoCloseable
Atomic counter that is backed by an AtomicBuffer that can be read across threads and processes.

In most cases you want to pair the appropriate methods for ordering. E.g.

  1. an increment() (which has volatile semantics) should be combined with a get().
  2. an incrementRelease() with a getAcquire().
  3. an incrementOpaque() with a getOpaque().
  4. an incrementPlain() with a getPlain().
  • Field Details

    • isClosed

      private boolean isClosed
    • id

      private final int id
    • addressOffset

      private final long addressOffset
    • byteArray

      private final byte[] byteArray
    • countersManager

      private CountersManager countersManager
    • byteBuffer

      private final ByteBuffer byteBuffer
  • Constructor Details

    • AtomicCounter

      public AtomicCounter(AtomicBuffer buffer, int counterId)
      Map a counter over a buffer. This version will NOT free the counter on close.
      Parameters:
      buffer - containing the counter.
      counterId - identifier of the counter.
    • AtomicCounter

      public AtomicCounter(AtomicBuffer buffer, int counterId, CountersManager countersManager)
      Map a counter over a buffer. This version will free the counter on close.
      Parameters:
      buffer - containing the counter.
      counterId - identifier for the counter.
      countersManager - to be called to free the counter on close.
  • Method Details

    • id

      public int id()
      Identity for the counter within the CountersManager.
      Returns:
      identity for the counter within the CountersManager.
    • disconnectCountersManager

      public void disconnectCountersManager()
      Disconnect from CountersManager if allocated, so it can be closed without freeing the slot.
    • close

      public void close()
      Close counter and free the counter slot for reuse of connected to CountersManager.
      Specified by:
      close in interface AutoCloseable
    • isClosed

      public boolean isClosed()
      Has this counter been closed?
      Returns:
      true if this counter has already been closed.
    • label

      public String label()
      Return the label for the counter within the CountersManager.
      Returns:
      the label for the counter within the CountersManager.
    • updateLabel

      public void updateLabel(String newLabel)
      Update the label for the counter constructed with a CountersManager.
      Parameters:
      newLabel - for the counter with a CountersManager.
      Throws:
      IllegalStateException - is not constructed CountersManager.
    • appendToLabel

      public AtomicCounter appendToLabel(String suffix)
      Append to the label for a counter constructed with a CountersManager.
      Parameters:
      suffix - for the counter within a CountersManager.
      Returns:
      this for a fluent API.
      Throws:
      IllegalStateException - is not constructed CountersManager.
    • updateKey

      public void updateKey(Consumer<MutableDirectBuffer> keyFunc)
      Update the key for a counter constructed with a CountersManager.
      Parameters:
      keyFunc - callback to use to update the counter's key
      Throws:
      IllegalStateException - is not constructed CountersManager.
    • updateKey

      public void updateKey(DirectBuffer keyBuffer, int offset, int length)
      Update the key for a counter constructed with a CountersManager.
      Parameters:
      keyBuffer - contains key data to be copied into the counter.
      offset - start of the key data within the keyBuffer
      length - length of the data within the keyBuffer (must be <= CountersReader.MAX_KEY_LENGTH)
      Throws:
      IllegalStateException - is not constructed CountersManager.
    • increment

      public long increment()
      Perform an atomic increment that will not lose updates across threads.
      Returns:
      the previous value of the counter
    • incrementOrdered

      public long incrementOrdered()
      Perform an atomic increment that is not safe across threads.

      This method is identical to incrementRelease() and that method should be used instead.

      Returns:
      the previous value of the counter
    • incrementRelease

      public long incrementRelease()
      Perform a non-atomic increment with release semantics.

      It can result into lost updates due to race condition when called concurrently.

      The load has plain memory semantics and the store has release memory semantics.

      The typical use-case is when there is a single writer thread and one or more reader threads and causality needs to be preserved using the getAcquire().

      This method is likely to outperform the increment(). So if there is just a single mutator thread, and one or more reader threads, then it is likely you will prefer this method.

      If no memory ordering is needed, have a look at the incrementOpaque().

      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • incrementOpaque

      public long incrementOpaque()
      Perform a non-atomic increment using opaque semantics.

      It can result into lost updates due to race condition when called concurrently.

      The load has plain memory semantics and the store has opaque memory semantics.

      The typical use-case is when there is a single writer thread and one or more reader threads and surrounding loads/stores don't need to be ordered.

      This method should be at least fast as incrementRelease() since it has weaker memory semantics. So if there is just a single mutator thread, and one or more reader threads, then it is likely you will prefer this method.

      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • incrementPlain

      public long incrementPlain()
      Increment the counter.

      This method is not atomic and this can lead to lost-updates due to race conditions. This load and store have plain memory semantics.

      The typical use-case for this method is when writer and reader are the same thread.

      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • decrement

      public long decrement()
      Perform an atomic decrement that will not lose updates across threads.

      The loads and store have volatile memory semantics.

      Returns:
      the previous value of the counter
    • decrementOrdered

      public long decrementOrdered()
      Perform an atomic decrement that is not safe across threads.
      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • decrementRelease

      public long decrementRelease()
      Decrements the counter non-atomically with release semantics.

      It can result into lost updates to race condition when called concurrently.

      The load has plain memory semantics and the store has release memory semantics.

      The typical use-case is when there is one mutator thread, that calls this method, and one or more reader threads.

      This method is likely to outperform the increment() and probably will be a better alternative.

      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • decrementOpaque

      public long decrementOpaque()
      Perform a non-atomic decrement using opaque semantics.

      It can result into lost updates due to race condition when called concurrently.

      The load has plain memory semantics and the store has opaque memory semantics.

      The typical use-case is when there is a single writer thread and one or more reader threads and surrounding loads and stores don't need to be ordered.

      This method should be at least fast as incrementRelease() since it has weaker memory semantics. So if there is just a single mutator thread, and one or more reader threads, then it is likely you will prefer this method.

      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • decrementPlain

      public long decrementPlain()
      Decrements the counter.

      This method is not atomic and this can lead to lost-updates due to race conditions. This load and store have plain memory semantics.

      The typical use-case for this method is when writer and reader are the same thread.

      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • set

      public void set(long value)
      Set the counter with volatile semantics.
      Parameters:
      value - to be set with volatile semantics.
    • setOrdered

      public void setOrdered(long value)
      Set the counter with ordered semantics.

      This method is identical to setRelease(long) and that method should be used instead.

      Parameters:
      value - to be set with ordered semantics.
    • setRelease

      public void setRelease(long value)
      Set the counter value atomically.

      The store has release memory semantics.

      Parameters:
      value - to be set
      Since:
      2.1.0
    • setOpaque

      public void setOpaque(long value)
      Set the counter value atomically.

      The store has opaque memory semantics.

      Parameters:
      value - to be set
      Since:
      2.1.0
    • setWeak

      public void setWeak(long value)
      Set the counter with normal semantics.

      This method is identical to setPlain(long) and that method should be used instead.

      Parameters:
      value - to be set with normal semantics.
    • setPlain

      public void setPlain(long value)
      Set the counter value with plain memory semantics.
      Parameters:
      value - to be set with normal semantics.
      Since:
      2.1.0
    • getAndAdd

      public long getAndAdd(long increment)
      Add an increment to the counter that will not lose updates across threads.
      Parameters:
      increment - to be added.
      Returns:
      the previous value of the counter
    • getAndAddOrdered

      public long getAndAddOrdered(long increment)
      Add an increment to the counter with ordered store semantics.

      This method is identical to getAndAddRelease(long) and that method should be used instead.

      Parameters:
      increment - to be added with ordered store semantics.
      Returns:
      the previous value of the counter
    • getAndAddRelease

      public long getAndAddRelease(long increment)
      Adds an increment to the counter non-atomically.

      This method is not atomic; it can suffer from lost-updates due to race conditions.

      The load has plain memory semantics and the store has release memory semantics.

      The typical use-case is when there is one mutator thread, that calls this method, and one or more reader threads. Typically, this method is combined with the getAcquire() to read the value.

      Parameters:
      increment - to be added
      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • getAndAddOpaque

      public long getAndAddOpaque(long increment)
      Adds an increment to the counter non-atomically.

      This method is not atomic; it can suffer from lost-updates due to race conditions.

      The load has plain memory semantics and the store has opaque memory semantics.

      The typical use-case is when there is one mutator thread, that calls this method, and one or more reader threads. Typically, this method is combined with a getOpaque() to read the value.

      If ordering of surrounding loads/stores isn't important, then this method is likely to be faster than getAndAddRelease(long) because it has less strict memory ordering requirements.

      Parameters:
      increment - to be added
      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • getAndAddPlain

      public long getAndAddPlain(long increment)
      Adds an increment to the counter non-atomically.

      This method is not atomic; it can suffer from lost-updates due to race conditions.

      The load has plain memory semantics and the store has plain memory semantics.

      The typical use-case for this method is when writer and reader are the same thread.

      Parameters:
      increment - to be added
      Returns:
      the previous value of the counter
      Since:
      2.1.0
    • getAndSet

      public long getAndSet(long value)
      Get the current value of a counter and atomically set it to a new value.
      Parameters:
      value - to be set.
      Returns:
      the previous value of the counter
    • compareAndSet

      public boolean compareAndSet(long expectedValue, long updateValue)
      Compare the current value to expected and if true then set to the update value atomically.
      Parameters:
      expectedValue - for the counter.
      updateValue - for the counter.
      Returns:
      true if successful otherwise false.
    • get

      public long get()
      Get the value for the counter with volatile semantics.
      Returns:
      the value for the counter.
    • getAcquire

      public long getAcquire()
      Get the value for the counter with acquire semantics.
      Returns:
      the value for the counter.
      Since:
      2.1.0
    • getOpaque

      public long getOpaque()
      Get the value for the counter with opaque semantics.
      Returns:
      the value for the counter.
      Since:
      2.1.0
    • getWeak

      public long getWeak()
      Get the value of the counter using weak ordering semantics. This is the same a standard read of a field.

      This call is identical to getPlain() and that method is preferred.

      Returns:
      the value for the counter.
    • getPlain

      public long getPlain()
      Get the value of the counter using plain memory semantics. This is the same a standard read of a field.
      Returns:
      the value for the counter.
      Since:
      2.1.0
    • proposeMax

      public boolean proposeMax(long proposedValue)
      Set the value to a new proposedValue if greater than the current value with plain memory semantics.
      Parameters:
      proposedValue - for the new max.
      Returns:
      true if a new max as been set otherwise false.
    • proposeMaxOrdered

      public boolean proposeMaxOrdered(long proposedValue)
      Set the value to a new proposedValue if greater than the current value with memory ordering semantics.

      This method is identical to proposeMaxRelease(long) and that method should be used instead.

      Parameters:
      proposedValue - for the new max.
      Returns:
      true if a new max as been set otherwise false.
    • proposeMaxRelease

      public boolean proposeMaxRelease(long proposedValue)
      Set the value to a new proposedValue if greater than the current value.

      This call is not atomic and can suffer from lost updates to race conditions.

      The load has plain memory semantics and the store has release memory semantics.

      The typical use-case is when there is one mutator thread, that calls this method, and one or more reader threads.

      Parameters:
      proposedValue - for the new max.
      Returns:
      true if a new max as been set otherwise false.
      Since:
      2.1.0
    • proposeMaxOpaque

      public boolean proposeMaxOpaque(long proposedValue)
      Set the value to a new proposedValue if greater than the current value.

      This call is not atomic and can suffer from lost updates to race conditions.

      The load has plain memory semantics and the store has opaque memory semantics.

      The typical use-case is when there is one mutator thread, that calls this method, and one or more reader threads.

      This method is likely to outperform proposeMaxRelease(long) since this method has less memory ordering requirements.

      Parameters:
      proposedValue - for the new max.
      Returns:
      true if a new max as been set otherwise false.
      Since:
      2.1.0
    • toString

      public String toString()
      Overrides:
      toString in class Object