RWRunnableServer


RWRunnableServer  RWThread

Data Type and Member Function Indexes
Package
Header File
Description
Example
Public Constructors
Public Member Operator
Public Member Functions
Protected Constructor
See Also

Data Type and Member Function Indexes
(exclusive of constructors and destructors)

Member Functions

enqueue()
getCapacity()

make()
operator=()

setCapacity()
stop()

Package

Threading

Header File

#include <rw/thread/RWRunnableServer.h>

Description

The runnable server may be used alone, or within a class for an active object class implementation. The runnable server object, when started, waits for other threads to enqueue runnable objects that they would like the server thread to execute for them. Each runnable object that the server finds in its internal queue is dequeued and started. The runnable server does not join with the runnables it starts. The runnable server continues to dequeue runnables and execute them until stopped, interrupted, or canceled.

Runnables may be enqueued with a guard functor, or a priority value, or both.

The server uses the guard functor to determine whether the associated runnable is currently eligible for retrieval and execution. A guarded runnable is not retrieved by the server until that runnable is the first runnable in the queue whose guard evaluates to TRUE. Using guards to block runnable execution is important when implementing active objects; guards can be used to insure that the active object maintains consistent state and preserves any necessary invariants.

The priority value is used during write operations to determine a runnable's insertion point within the server's input queue, such that the set of un-processed and eligible runnables will be retrieved in priority order by the server.

The RWRunnable::requestCancellation() function should be used if the server thread is to stop execution as soon as possible without dequeuing and dispatching any additional runnables.

The RWRunnable::requestInterrupt() function can be used to temporarily suspend execution of the server thread.

An active object implementation using this class defines a client interface with member functions that simply produce "runnable functor" objects (see RWRunnableFunction.h and RWTRunnableIOUFunction.h) which, when invoked, execute some corresponding private or local member function within the active object instance. The public members pass the runnable functors to the server thread by calling the enqueue() member, which stores the runnables in a queue. The client interface member function then returns to its caller.

Runnables that use IOUs can be used to represent a future return value or result of the asynchronous active object operation. In this situation, the client interface would create the appropriate IOU-capable runnable object, enqueue the runnable for execution, retrieve the runnable's IOU result, and return that IOU to the caller. The caller could then redeem that IOU at some point in the future.

Example

#include <rw/thread/RWRunnableServer.h>
#include <rw/thread/RWRunnable.h>
#include <rw/thread/rwtMakeRunnableFunction.h>

RWRunnableServer runnableServer(RWRunnableServer::make());
RWRunnable runnable = rwtMakeRunnableFunction(function);
runnableServer.start();
runnableServer.enqueue(runnable); 
// ...
runnableServer.stop();
runnableServer.join();

The following is a skeleton of an active object class implemented using an RWRunnableServer. For a complete active object example, refer to activobj.cpp in the threxam directory.

      class ActiveObject {
      public:
         // Starts runnable server.
         ActiveObject();
       
         // Stops runnable server.
         ~ActiveObject(void);
       
         // A function with no return. Creates and enqueues an 
         // RWRunnableFunction, wrapping _func1();
         void func1(arg1,arg2,arg3);
      
         // A function with a return value. Creates and enqueues 
         // an RWTRunnableIOUFunction, wrapping _func2();
         RWTIOUResult<ReturnType> func2(arg1,arg2,arg3);
       
      private:
         // func1()'s better half 
         void _func1(arg1,arg2,arg3);
       
         // func2()'s better half
         ReturnType _func2(arg1,arg2,arg3);
       
         // Guard functions called by RWRunnableGuard objects
         RWBoolean guard1Func();
         RWBoolean guard2Func();
      
         // Runnable execution guards enqueued on  
         // RWRunnableServer along with RWRunnable objects.
         RWRunnableGuard guard1_;
         RWRunnableGuard guard2_;
       
         // The internal runnable server instance.
         RWRunnableServer runnableServer_;
      };

Public Constructors

RWRunnableServer(void);

Constructs an empty, invalid, handle instance.

RWRunnableServer(const RWRunnableServer& second);

Constructs an external interface handle to the thread object pointed to by a second handle, if any.

Public Member Operator

RWRunnableServer&
operator=(const RWRunnableServer& second);

Binds this external interface handle to the thread object pointed to by a second handle, if any.

Public Member Functions

void
enqueue(const RWRunnable& runnable);

Passes a runnable instance to the server for execution. All runnable are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, in priority order, and executes them to completion.

If the server is running and its input queue has sufficient capacity to hold another runnable, this function inserts the runnable and returns. If the server is running but its input queue is full, this function blocks the calling thread until the input queue is no longer full (because the server has removed a sufficient number of runnables, or another thread has changed the maximum capacity of the input queue) or until the server is stopped.

If the server is stopped while the calling thread is blocked and waiting within this function for space in the input queue, the calling thread is unblocked and the function returns by throwing an RWTHRClosedException. If the server is already closed when this function is called, the function immediately returns by throwing an RWTHRClosedException.

This function takes the parameter runnable, which is a const reference to an RWRunnable instance of the type used to instantiate this template class.

Runnables enqueued with this method are not guarded and are assigned a priority value of zero.

RWWaitStatus
enqueue(const RWRunnable& runnable,
        unsigned long milliseconds);

Passes a runnable instance to the server for execution. All runnables are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, in priority order, and executes them to completion.

If the server is running and its input queue has sufficient capacity to hold another runnable, this function inserts the runnable and returns RW_THR_COMPLETED. If the server is running but its input queue is full, this function blocks the calling thread until the input queue is no longer full (because the server has removed a sufficient number of runnables, or another thread has changed the maximum capacity of the input queue), or until the server is stopped, or until the specified amount of time passes. If the runnable cannot inserted in the input queue within the specified amount of time, this function returns a value of RW_THR_TIMEOUT.

This function takes two parameters:

The parameter runnable is a const reference to an RWRunnable instance.
The parameter milliseconds is an unsigned long value that specifies the maximum number of milliseconds to wait for the operation to complete.

Runnables enqueued with this method are not guarded and are assigned a priority value of zero.

void
enqueue(const RWRunnable& runnable, 
        const RWRunnableGuard& guard);

Passes a guarded runnable instance to the server for execution. All runnable are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, in priority order, and executes them to completion.

This function takes two parameters:

The parameter runnable is a const reference to an RWRunnable instance.
The parameter guard is a reference to an RWTFunctorR0<RWBoolean> functor instance that is invoked by the server thread to determine whether the associated runnable is currently eligible for retrieval from the input queue and execution. The server will retrieve and execute a runnable if its guard functor evaluates to TRUE, and will skip any runnable whose guard functor evaluates to FALSE, until a runnable is found that can be executed. An empty functor handle indicates that the entry is always eligible for execution. The functor must not attempt to enqueue other runnables on the same server instance, as such access results in deadlock.

Runnables enqueued with this method are assigned a priority value of zero.

RWWaitStatus
enqueue(const RWRunnable& runnable,
        const RWRunnableGuard& guard
        unsigned long milliseconds);

Passes a guarded runnable instance to the server for execution. All runnables are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, in priority order, and executes them to completion.

If the server is running and its input queue has sufficient capacity to hold another runnable, this function inserts the runnable and returns RW_THR_COMPLETED. If the server is running but its input queue is full, this function blocks the calling thread until the input queue is no longer full (because the server has removed a sufficient number of runnables, or another thread has changed the maximum capacity of the input queue), or until the server is stopped, or until the specified amount of time passes. If the runnable cannot be inserted in the input queue within the specified amount of time, this function returns a value of RW_THR_TIMEOUT.

This function takes three parameters.

The parameter runnable is a const reference to an RWRunnable instance.
The parameter guard is a reference to an RWTFunctorR0<RWBoolean> functor instance that is invoked by the server thread to determine whether the associated runnable is currently eligible for retrieval from the input queue and execution. The server will retrieve and execute a runnable if its guard functor evaluates to TRUE, and will skip any runnable whose guard functor evaluates to FALSE, until a runnable is found that can be executed. An empty functor handle indicates that the entry is always eligible for execution. The functor must not attempt to enqueue other runnables on the same server instance, as such access results in deadlock.
The parameter milliseconds is an unsigned long value that specifies the maximum number of milliseconds to wait for the operation to complete.

Runnables enqueued with this method are assigned a priority value of zero.

void
enqueue(long priority,
        const RWRunnable& runnable);

Passes a prioritized runnable instance to the server for execution. All runnable are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, in priority order, and executes them to completion.

This function takes two parameters:

The parameter priority is a long that is used to determine the insertion point of the runnable within the server's input queue. A runnable with a greater priority value is positioned to be read before a runnable of lower priority. Runnables of equal priority are positioned such that older runnables are executed before newer runnables.
The parameter runnable is a const reference to an RWRunnable instance.

Runnables enqueued with this method are not guarded.

RWWaitStatus
enqueue(long priority,
        const RWRunnable& runnable,
        unsigned long milliseconds);

Passes a prioritized runnable instance to the server for execution. All runnables are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, and executes them to completion.

This function takes three parameters:

The parameter priority is a long that is used to determine the insertion point of the runnable within the server's input queue. A runnable with a greater priority value is positioned to be read before a runnable of lower priority. Runnables of equal priority are positioned such that older runnables are executed before newer runnables.
The parameter runnable is a const reference to an RWRunnable instance.
The parameter milliseconds is an unsigned long value that specifies the maximum number of milliseconds to wait for the operation to complete.

Runnables enqueued with this method are not guarded.

void
enqueue(long priority,
        const RWRunnable& runnable, 
        const RWRunnableGuard& guard);

Passes a guarded and prioritized runnable instance to the server for execution. All runnable are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, and executes them to completion.

This function takes three parameters:

The parameter priority is a long that is used to determine the insertion point of the runnable within the server's input queue. A runnable with a greater priority value is positioned to be read before a runnable of lower priority. Runnables of equal priority are positioned such that older runnables are executed before newer runnables.
The parameter runnable is a const reference to an RWRunnable instance.
The parameter guard is a reference to an RWTFunctorR0<RWBoolean> functor instance that is invoked by the server thread to determine whether the associated runnable is currently eligible for retrieval from the input queue and execution. The server will retrieve and execute a runnable if its guard functor evaluates to TRUE, and will skip any runnable whose guard functor evaluates to FALSE, until a runnable is found that can be executed. An empty functor handle indicates that the entry is always eligible for execution. The functor must not attempt to enqueue other runnables on the same server instance, as such access results in deadlock.

RWWaitStatus
enqueue(long priority,
        const RWRunnable& runnable,
        const RWRunnableGuard& guard
        unsigned long milliseconds);

Passes a guarded and prioritized runnable instance to the server for execution. All runnables are inserted into an internal producer-consumer queue, where the server retrieves them, one at a time, in priority order, and executes them to completion.

This function takes four parameters.

The parameter priority is a long that is used to determine the insertion point of the runnable within the server's input queue. A runnable with a greater priority value is positioned to be read before a runnable of lower priority. Runnables of equal priority are positioned such that older runnables are executed before newer runnables.
The parameter runnable is a const reference to an RWRunnable instance.
The parameter guard is a reference to an RWTFunctorR0<RWBoolean> functor instance that is invoked by the server thread to determine whether the associated runnable is currently eligible for retrieval from the input queue and execution. The server will retrieve and execute a runnable if whose guard functor evaluates to TRUE, and will skip any runnable whose guard functor evaluates to FALSE, until a runnable is found that can be executed. An empty functor handle indicates that the entry is always eligible for execution. The functor must not attempt to enqueue other runnables on the same server instance, as such access results in deadlock.
The parameter milliseconds is an unsigned long value that specifies the maximum number of milliseconds to wait for the operation to complete.

size_t
getCapacity(void) const;

Returns the maximum capacity of the server's input queue. A value of zero indicates that the queue has no capacity limit.

static RWRunnableServer
make(void);

Makes a runnable server.

static RWRunnableServer
make(size_t maxCapacity);

Makes a runnable server whose input queue has the specified maximum capacity.

static RWRunnableServer
make(const RWThreadAttribute& serverThreadAttr);

Makes a runnable server instance with the specified thread attributes.

static RWRunnableServer
make(const RWThreadAttribute& serverThreadAttr,
     size_t maxCapacity);

Makes a runnable server instance with the specified thread attributes whose input queue has the specified maximum capacity.

size_t
setCapacity(size_t maxCapacity) const;

Sets the maximum capacity of the server's input queue. The maximum capacity of the input queue determines the maximum number of un-processed entries allowed to accumulate within the input queue. Once the number of entries equals or exceeds this number, any thread attempting to enqueue another runnable is blocked until the server removes a sufficient number of runnables from the queue, or until another thread increases the capacity. A value of zero indicates that the queue has no capacity limit. This function returns the maximum capacity value that was in use before the call.

void
stop(void)

Requests that the server stop execution after it executes the runnables that are currently enqueued. Any attempt to enqueue a runnable after the server has been stopped, or before it has been started, results in an RWTHRClosedException.

Protected Constructor

RWRunnableServer(RWRunnableServerImp* threadImpP)

Constructs an external interface handle to an RWRunnableServerImp instance.