Monday, November 27, 2006

Unique Role of Java in Programming Language Innovation

I was lucky to attend IBM IRL workshop on Next Generation Programming languages and Environments on 20-21st of Nov at India Internatinal Centre. Writing below some details of the first talk. Though, there were many talks in the workshop. I am able to recall only a few so wont be able to write about all the talks in my posts.

Unique Role of Java in Programming Language Innovation

Dr Bob Blainey of IBM Software group being a distinguished engineer and chief java technologist was no doubt the right person to talk about role of java in driving next generation programming languages and environments. He said that at least three mejor extension of java coming up to aid different programming need.


X10 is a java extension which is designed to address the need for
concurent programming. One can not ignore the major shift towards
exploiting parallelism as much as possible. Next generation
programming languages can not ignore this shift and they have to
provide programmers construct which make use of underlying

There are a few constructs added in java, some of which I am listing

Place :- This is a logical construct which can be mapped to physical
place in a sense that a place could be mapped to different processor
or a different machine altogether. The X10 framework assumes
distributed memory model and gives control to program in a GALS (
Globally Asynchronous Locally Synchronous ) environment.

Activity :- Activity resembles classical java threads. An activity can
access or modify only those objects which resides in the local place (
place in which the activity is residing ). So if an activity wants to
access a remote object, it has to first create an activity at the
remote place and then have to communicate with that newly created
activity to get the desired result. BadPlaceException is thrown in
case an activity tries to access a remote object.

Clock :- Those who are familiar with typical multi threading/multi
programming environments would say that this object acts as a barrier.
Different activities can register themselves with clocks and use it
for the purpose of synchronization. A place can have multiple clocks
and different activities can register for more than one clocks as
well. There is no global clock in the framework which makes it
globally asynchronous.

A container (e.g. array ) can be distributed across places. For
example programmer can decide to distribute array in such a way so
that an array of size 10 ( 0..9) its elements A[0] A[5] A[6] belong to place1
A[2] A[8] A[9] belong to place2. This distribution makes it difficult
to construct a simple for loop which will go through all the array
elements and update it because there is a possibility of a
BadPlaceException. To take care of this situation a projection
operator is introduced. For example,

for( point [i] : A | here )
where "here" refers to the place associated with the given scope just
like "this" refers to the object. This way one can iterate through all
the points ( elements ) of array A only residing within current place.

async( place ) { //some code } - This construct allows programmer to
execute a certain task at a remote place. It is not like a remote
procedure call in a sense that the calling module will not be blocked.
In fact, the code will run as an independent thread which may or may
not communicate back to the calling module after the invocation. In
other words, an asynchronous thread of execution is created at a
remote place to execute certain piece of code.

Many more constructs were described but unfortunately I am not able
to remember other details :-(.


XJ is an augmentation of Java is designed to address need for XML &
Web services. It extends Java syntax as well as semantics. One example
may look like :

Article A = new ( <>
<> getTitle() < /article_title >
< /article > );

I am not able to recall the exact syntax but it looked much like the
code I mentioned above.

Java with a real-time flavour. Major feature is handling garbage
collection with real-time constraints. I really wonder what kind of
algorithms would it take to perform garbage collection with such
constraints. What happens when the GC is halfway through marking the
live nodes and it has to relinquish control to another thread to
satisfy the constraints? Or what is GC is halfway through the
compaction ? I would really like to know the underlying details which
handles these intricacies.

I would like to end this post by a quote which Dr Bob Blainey
mentioned during his talk.

"Claiming that Java is easier than C++ is like saying that K2 is
shorter than everest" -- Larry O' Brian