<body>
<p> Provides real-time {@link javolution.context.Context} to facilitate
separation of concerns and achieve higher level of performance and
code predictability.</p>
<h2><a name="OVERVIEW">Separation of Concerns</a></h2>
<p> Separation of concerns is an important design principle greatly misenderstood.
Most developers think it is limited to modularity and encapsulation or it
requires special programming tools (e.g. Aspect programming).</p>
<p> Separation of concerns is very powerful and easier than it looks.
Basically, it could be summarized as the "pass the buck principle".
If you don't know what to do with some information, just give it to someone
else who might know.</p>
<p> A frequent example is the catching of exceptions too early (with some logging processing)
instead of throwing a checked exception. Unfortunately, they are still plenty of cases
where the separation of concerns is not as good as it could be. For example logging!
Using the standard logging, the code has to know which logger to log to? Why?</p>
<p> Separation of concerns can be adressed through "Aspect Programming",
but there is a rather simpler solution <b>"Context Programming"</b>!</p>
<p> It does not require any particular tool, it basically says that every threads
has a context which can be customized by someone else (the one who knows what to do).
Then, your code looks a lot cleaner and is way more flexible as you don't have
to worry about logging, security, performance etc. in your low level methods.
For example:[code]
void myMethod() {
...
LogContext.info("Don't know where this is going to be logged to");
...
}[/code]
</p>
<p> Used properly <b>J</b>avolution's {@link javolution.context.Context contexts}
greatly facilitate the separation of concerns. Contexts are
complemented by others classes such as for example the
{@link javolution.lang.Configurable Configurable} class to reduce
dependency between configuration and application code.</p>
<h2><a name="PREDEFINED">Predefined Contexts:</a></h2>
<p> This package provides few predefined contexts:<ul>
<li>{@link javolution.context.LocalContext LocalContext} - To define locally
scoped environment settings.</li>
<li>{@link javolution.context.ConcurrentContext ConcurrentContext} - To take advantage of concurrent
algorithms on multi-processors systems.</li>
<li>{@link javolution.context.AllocatorContext AllocatorContext} - To control
object allocation, e.g. {@link javolution.context.StackContext StackContext}
to allocate on the stack (or RTSJ ScopedMemory).</li>
<li>{@link javolution.context.LogContext LogContext} - For thread-based or object-based logging
capability, e.g. {@link javolution.util.StandardLog StandardLog} to leverage standard
logging capabilities.
<i>Note: <code>java.util.logging</code> provides class-based
logging (based upon class hierarchy).</i></li>
<li>{@link javolution.context.PersistentContext PersistentContext} - To achieve persistency across
multiple program execution.</li>
<li>{@link javolution.context.SecurityContext SecurityContext} - To address application-level security
concerns.</li>
<li>{@link javolution.testing.TestContext TestContext} - To address varied aspect of testing such as performance and regression. </li>
</ul>
</p>
<h2><a name="FAQ">FAQ:</a></h2>
<ol>
<a name="FAQ-1"></a>
<li><b>I am writing an application using third party libraries.
I cannot avoid GC unless I get the source and patch it to Javolution.
Can I still make my application real-time using {@link javolution.context.StackContext StackContext}?</b>
<p> You cannot get determinism using "any" library (including Java standard library)
regardless of the garbage collector issue. Array resizing, lazy initialization, map rehashing (...)
would all introduce unexpected delays (this is why Javolution comes with its own
{@link javolution.lang.Realtime real-time} collections implementation).
Still, you may use incremental/real-time collectors (if few milliseconds delays are acceptable).
These collectors work even faster if you limit the amount of garbage produced onto the heap
through {@link javolution.context.StackContext stack allocations}.</p>
</li><p></p>
<a name="FAQ-2"></a>
<li><b>Can you explain a little how objects can be "stack" allocated?</b>
<p> It all depends upon the StackContext {@link javolution.context.StackContext#DEFAULT default} implementation.
The default implementation use thread-local queues (no synchronization required);
but if you run on a <a href="https://rtsj.dev.java.net/">RTSJ</a> virtual machine
entering a {@link javolution.context.StackContext StackContext} could mean using <code>ScopedMemory</code>.</p>
</li><p></p>
<a name="FAQ-3"></a>
<li><b>As a rule, I am skeptical of classes that pool small objects.
At one time (5 years ago) it was a big win. Over time, the advantage has
diminished as garbage collectors improve. Object pools can make it much more
difficult for the garbage collector to do its job efficiently, and can have
adverse effects on footprint. (Joshua Bloch) </b>
<p> Stack allocation is different from object pooling, it is a simple and transparent
way to make your methods "clean" (no garbage generated), it has also the side effect
of making your methods faster and more time-predictable.
If all your methods are "clean" then your whole
application is "clean", faster and more time-predictable (aka real-time).</p>
<p> In practice very few methods need to enter a {@link javolution.context.StackContext StackContext},
only the one generating a significant number of temporary objects (these methods are made "cleaner"
and faster through stack allocation). For example:[code]
public final class DenseVector<F extends Field<F>> extends Vector<F> {
...
public F times(Vector<F> that) {
final int n = this.getDimension();
if (that.getDimension() != n) throw new DimensionException();
StackContext.enter();
try { // Reduces memory allocation / garbage collection.
F sum = this.get(0).times(that.get(0));
for (int i = 1; i < n; i++) {
sum = sum.plus(this.get(i).times(that.get(i)));
}
return StackContext.outerCopy(sum); // Stack object exported through copy.
} finally {
StackContext.exit(); // Resets stack.
}
}
...
}[/code]</p>
</li><p></p>
</ol>
</body>
|