TABS Series session on "Agile and Lean Concepts for IT Professionals"

On Friday, Mar 5th, 2010 attended ThoughtWorks TABS (ThoughtWorks Agile Business Series) session on "Agile and Lean Concepts for IT Professionals". It was a great piece of experience. They took all the attendees to the actual Project teams and showed how their teams are agile. Followed by that was the session on "Lean Concepts for IT" by Scott Shaw. It was a really informative session.

Now looking forward for ThoughtWorks Master Class Series talk on "10 Ways to Improve Your Code" by Neal Ford on Wednesday, Mar 10th, 2010.

Intel - Software Developer Supercomputing Conference, 2009

Today I attended Intel's - Software Developer Supercomputing Conference, 2009. It was a really great conference. Lot of ideas were shared by intel people like how to improve application performance using threads, libraries (like MKL-Math Kernel Library), using their new products, etc. etc.

The session was interactive enough and at the last they distributed Intel's T-shirts :)

Some useful pointers:
whatif.intel.com
go-parallel.com

..."sixth sense" - yes, it's coming...

Today I came across bleeding edge research of the year 2009 - Pranav Mistri's "sixth sense". It's a simple, wearable device that enhances the real world with digital information.

Great to see future developments in this area..

You can find more info on his website here.

J2SE 5.0 reached its End of Service Life (EOSL) Notice

On November 3, 2009, J2SE 5.0 reached its End of Service Life (EOSL), which is the date of final publicly available update of version 5.0 (i.e. J2SE 5.0 Update 22).

You can find more information here.
Learn more about EOL policy here.

Ubuntu 9.10 coming...

Code Kata by Dave Thomas

Dave Thomas took the idea of coding practice and made something very well, pragmatic out of it.
There are series of katas, which are small, thought-provoking exercises that programmers can do in the language of their choice.

You can find it here. Really worth giving a try :)

Chaining methods using power of "return this"

In one of my older post [here], I discussed about the power of Fluent Interfaces. We can use those in building method chaining, thus improving the readability of code just english-like sentences.

We can also implement this using "return this" in class methods.

Take a look at below source code:

Bug.java



CreateBugs.java



Output:



The above programs are self explanatory :)

To overcome "Maximum open cursors exceeded" issue

When we execute a query in oracle, a ResultSet is created and stored in the memory. Oracle allows the programmer to access this ResultSet in the memory through cursors.

When we fire SELECT query / DML query from Java code, Oracle returns the reference of this created cursor to the Java program, which we refer as a ResultSet. Following is the hierarchy of creating objects in JDBC:

[For DML queries, we don't require "ResultSet" object]

Now, when we execute this query in loop for multiple records, we can use the same "Connection" object, but we'll have to close ResultSet object and Statement object at the end of each loop.

[Here we're using distributed database connection which means for each request, we create new "Connection" object and after the request is processed, we close this "Connection" object. Now, there might be the case that for one request, we need to fire multiple SELECT / DML queries, so we use the same "Connection" object to process this request.]

Closing ResultSet object and Statement object at the end of each loop will make sure that the cursors opened by executing this query will be closed / released.

[For DML queries, we need to close only "Statement" object].

Finally we close the connection object when we finish executing the queries in loop.

All the resources (including cursors) are released when the "Connection" object is closed.

Following this, we'll never encounter "Maximum open cursors exceeded" issue in any application :)

Maintaining and Managing variables stored in application's session

In one of our project, we were storing frequently used application related data into the session, but now as the project is growing incrementally, storing new data / modifying existing data into session has became very confusing and difficult to manage.

I was just thinking about the workaround for coming out of this havoc, and voila, I got one :). This workaround I've implemented in our project. I would like to share the idea which brought us out of this mess.

Idea (Workaround): We can use one singleton java class, which has one HashMap variable. There will be custom methods for getting, setting and removing (operations same as that of session operations) values from HashMap. Now we can use this implementation for storing data (which was supposed to be stored into the session) into the HashMap present in this class, and then we can put the object of this java class (we'll be having one object for each session) into the session. So, at the session level, there will be only one object present in the session, and now it's lot easy to maintain and manage session variables (now HashMap data). Also, there will be only two session operations in each class, first: getting this java class object from session, second: storing back this java class object back to the session.

Implementing this thing was really fun and gave us lot of flexibility to maintain and manage session variables... :)

-C-O-N-C-E-P-T- Series : Java

Difference between using "" and toString() method for converting object value to String:

Case I - Using "":

-----------------------------------------------------------------
BigDecimal dec = new BigDecimal("1234567");
String str = "" + dec; // converting dec to String
System.out.println("str: "+str);

o/p:
str: 1234567
-----------------------------------------------------------------

-----------------------------------------------------------------
BigDecimal dec = null;
String str = "" + dec; // converting dec to String
System.out.println("str: "+str);

o/p:
str: null
-----------------------------------------------------------------


Case II - Using toString() method:

-----------------------------------------------------------------
BigDecimal dec = new BigDecimal("1234567");
String str = null;

if( dec!=null ) {
str = dec.toString(); // converting dec to String
}

System.out.println("str: "+str);

o/p:
str: 1234567
-----------------------------------------------------------------

-----------------------------------------------------------------
BigDecimal dec = null;
String str = null;

if( dec!=null ) {

str = dec.toString(); // converting dec to String

}

System.out.println("str: "+str);

o/p:
str: null
-----------------------------------------------------------------

In first case, we don't require null check, but doesn't has readability.
In second case, we require null check, but has readability.

Each one has its own ease of use.

Custom Map Implementation in Java

Blogging after a long-long-long-... time :)

When we fetch records from HashMap, it gives records according to hash code, i.e. the records returned are not in the order when they're placed in the Map.

When we fetch records from TreeMap, it gives records in the sorted order of key, i.e. the records returned are not in the order when they're placed in the Map.

What I want was a Map that should return sequential values/objects that are put into when they're fetched. So, why not to make our own custom implementation of Map that does what is needed.

I constructed following Custom Map Implementation. Please find the code below:

CustomMapImpl.java



===
Supporting Class: CustomEntrySetImpl.java



===

Output of above program:

key1 : value1
key2 : value2
key4 : value4
key5 : value5
key6 : value6
key7 : value7

---

It was a very great experience. I also digged into the actual Java Map source code which was an amazing experience.

Using "synchronization" effectively

"Synchronization is anti-performance. So how can we use synchronization and still optimize on performance? So apply synchronization where it is absolutely necessary, or as minimum as possible."


We'll consider following two code snippets:

Scenario: Creating/accessing singleton object

Method level synchronization:



As you can see, the "synchronized" keyword is used at method level. So, now in multi-threaded environment, multiple threads accessing the getInstance() method will be queued-in. Following figure shows the scenario:



As shown in above figure, thread "t1" (the thread that is allowed to enter the getInstance() method first) will create the first object of "DemoClass" class and other threads will just get the reference of that singleton object which was created by thread "t1". Now this will definitely hamper the application's performance as every thread after "t1" will be queued-in, as there is no need after the singleton object is been created. So, we can use "synchronized" at block level which will reduce this performance bottle-neck. How, we'll see it soon :)

Block level synchronization:



As you can see, the "synchronized" keyword is used at block level. So, now in multi-threaded environment, multiple threads accessing the getInstance() method will NOT be queued-in. Following figure shows the scenario:



As shown in above figure, multiple threads can simultaneously access the getInstance() method. But, it can be that more than one threads can simultaneously hit the first "if statement". So, we can use "synchronized" keyword inside "if" such that those number of threads which are arrived parallel-ly can be queued-in. The second "if statement" makes sure that only one object is created as we're dealing with singleton class. Now, multiple threads can parallel-ly access the getInstance() method to get the reference of a singleton object. This removes the performance bottle-neck as mentioned in the above scenario.

Cool concept, isn't it :)

Updating the state of a Singleton Object after specific interval of time

I've created a Singleton Object which holds some values which are fetched from the database table. Following are the class specs:

1) All setter methods are private
2) All getter methods are public
3) One configuration method that sets all the values in the related attributes of class using their private setters

The need was to update the state of this singleton object when there are any changes made in the configuration table in the database. But to achieve this, I need to restart the server each time when any changes has been made in the configuration table.

So I created one WatchDog class in a thread that continuously changes the state of a singleton configuration object after a specific interval of time, & this time interval in-turn is fetched dynamically from the same configuration table from the database :)

Some experience with Log4j

I was doing some basic practice programs which implements Log4j for logging. I was getting following warning message on the console. Here I was using PropertyConfigurator to configure Log4j using the log4j.properties file.



I found the description about the above warning on many of the technical forums which gave a reason that this warning occurs when the log4j.properties file has not been found by the program in the specified path.

But if that was the case, I purposely tried giving wrong file path to the same program, and whoa I got the following exception:



Later I found that there was a small spelling mistake in my log4j.properties file, due to which it was failing to initialize the log4j system. So, when I corrected that mistake & tried to run the program, I got the expected output :)

-C-O-N-C-E-P-T- Series : Java

Synchronizing a static method and its effect in the multi-threading environment:

Since static code can modify only the static data, you will protect a static data with synchronized keyword. We know that static data and methods have only one copy for all the objects of that class. Therefore, you only need one lock per class to synchronize static methods. Object locks are not required for static methods. Static methods use a class lock for synchronization. This special lock is also an object lock.

Consider following java code:


Fig1


Fig2

Both the figure above are self explanatory, but I would like to elaborate more on Fig2. When we synchronize an instance method, the thread accessing that method acquires "Object Lock" on the object such that no other thread can access that method unless it completes its execution. Now for "synchronized static" method, it acquires "Class Lock" on that class. This class lock also implicitly has object lock, it's because as non-static methods can modify static variables, so we need to also put a lock on synchronized instance variables & this is internally done by the class lock.
Hence if one thread is executing a static synchronized method of a class then it holds a lock on all other synchronized methods of that class and in effect no other thread can call any of the synchronized static methods of that class until this current thread releases the lock.

This concept is really very interesting to learn :)

Stay tuned... :)