Package javolution.context

Provides real-time Context to facilitate separation of concerns and achieve higher level of performance and code predictability.

See: Description

Package javolution.context Description

Provides real-time Context to facilitate separation of concerns and achieve higher level of performance and code predictability.

Separation of Concerns

Separation of concerns is an important design principle greatly misunderstood. Most developers think it is limited to modularity and encapsulation or it requires special programming tools (e.g. Aspect programming).

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.

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! Why low-level code need to know which logging facility is being used (e.g. standard logging, Log4J library or anything else)? Furthermore, why logging should have to be based upon the class hierarchy (standard logging)?

Separation of concerns can be addressed through "Aspect Programming", but there is a rather simpler solution "Context Programming"!

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:

       void myMethod() {
           ...
           LogContext.info("Don't know where this is going to be logged to");
           ...
       }

Used properly Javolution's contexts greatly facilitate the separation of concerns. Contexts are complemented by others classes such as for example the Configurable class to reduce dependency between configuration and application code.

Predefined Contexts:

This package provides few predefined contexts:

FAQ:

  1. 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 StackContext?

    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 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 stack allocations.

  2. Can you explain a little how objects can be "stack" allocated?

    It all depends upon the StackContext default implementation. The default implementation use thread-local queues (no synchronization required); but if you run on a RTSJ virtual machine entering a StackContext could mean using ScopedMemory.

  3. 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)

    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).

    In practice very few methods need to enter a StackContext, only the one generating a significant number of temporary objects (these methods are made "cleaner" and faster through stack allocation). For example:

            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.
                    }
                }
                ...
            }

Copyright © 2005-2012 Javolution. All Rights Reserved.