Tuesday 21 April 2015

Find intersecting node between 2 Linked List

Kind of data structure questions which are normal in C language because C has pointers, so we have address of each memory location. But when we take java into consideration we don't have any pointer to judge whether nodes are same or not except double equal check( i.e. reference check). And then java.util.collections package comes to rescue us.

There are a number of ways by which you can solve the problem but important things are time and space complexity. It matters a lot when you are designing a highly scalable system where performance matters a lot and other side of story is "space comes at premium" which cannot be ignored.

Now let me put the problem statement into diagrammatic representation so that it would be easy to understand.

So here intersecting node is 45 which is first common element in both the linked list.

Time to start with my version of solution which is not very good from space complexity point of view but takes linear time complexity.

Before we start a very good point about IdentityHashMap is that it does reference matching i.e. double equals check not .equals() matching. So we are going to take this advantage in our algorithm.

Alogrithm steps:
1. Create an IndentityHashMap.
2. iterate through list 1 and put all nodes as key into IdentityHashMap.
3. iterate through list 2 and do containsKey check for all nodes in list2. Which ever containsKey returns true is the intersecting node between 2 linked list.

Time Complexity: O(n + m) linear order where n= elements in list1 and m= elements in list2
Space Complexity: O(n) linear order where n = elements in list1.

So far we are done with our algorithm design and need to jump start with code:

1. Below is pojo class which represents individual node in linked list.

 public class Node<T> {  
      private T value;  
      private Node<T> next;  
      public T getValue() {  
           return value;  
      }  
      public void setValue(T value) {  
           this.value = value;  
      }  
      public Node<T> getNext() {  
           return next;  
      }  
      public void setNext(Node<T> next) {  
           this.next = next;  
      }       
 }  

2. Main class where intersection logic written:

 import java.util.IdentityHashMap;  
 import java.util.Map;  
 public class IntersectionExample {  
      public static void main(String[] args) {  
           Node<Integer> list1 = new Node<Integer>();  
           list1.setValue(20);  
           Node<Integer> node1 = new Node<Integer>();  
           node1.setValue(60);  
           Node<Integer> node2 = new Node<Integer>();  
           node2.setValue(90);  
           Node<Integer> node3 = new Node<Integer>();  
           node3.setValue(45);  
           Node<Integer> node4 = new Node<Integer>();  
           node4.setValue(23);  
           node4.setNext(null);  
           list1.setNext(node1);  
           node1.setNext(node2);  
           node2.setNext(node3);  
           node3.setNext(node4);  
           Node<Integer> list2 = new Node<Integer>();  
           list2.setValue(9);  
           Node<Integer> point1 = new Node<Integer>();  
           point1.setValue(5);  
           Node<Integer> point2 = new Node<Integer>();  
           point2.setValue(8);  
           list2.setNext(point1);  
           point1.setNext(point2);  
           point2.setNext(node3);       
           getIntersection(list1, list2);  
      }  
      public static void getIntersection(Node<Integer> list1, Node<Integer> list2){  
           final Integer constant = 1;  
           Map<Node<Integer>, Integer> map = new IdentityHashMap<Node<Integer>, Integer>();  
           while(list1 != null){  
                Node<Integer> current = list1;  
                map.put(current, constant);  
                list1 = list1.getNext();  
           }  
           while(list2 != null){  
                Node<Integer> current = list2;  
                if(map.containsKey(current)){  
                     System.out.println("Intesecting Node, value: "+current.getValue());  
                 break;  
                }  
                list2 = list2.getNext();  
           }  
      }  
 }  

Output:

 Intesecting Node, value: 45  

That's all from my end on this post. As there can be a number of ways by which you can solve this problem, looking forward for your contribution. Thanks.

Monday 20 April 2015

Difference between new Random() and new Random(long seed) constructor behavior

Once again encountered with an interesting thing which I think need to be shared. This is regarding java.util.Random class to generate random numbers in code base and why 2 different constructors provided in Random class, what behavioral changes we are going to achieve using below 2 constructor:
1. new Random();
2. new Random(long seed); 

So we will start our discussion with code examples which covers above constructors.

Example 1 using new Random() constructor:

 import java.util.Random;  
 public class Example7 {  
      public static void main(String[] args) {  
           Random rand = new Random(); //Random instance without seed  
           for(int i=0; i<10; i++){  
                int randomNumber = rand.nextInt(2);  
                System.out.print(randomNumber+", ");  
           }  
      }  
 }  

Output:
 1, 0, 0, 1, 1, 0, 1, 1, 0, 1,   

(Please note output varies from system to system and execution to execution)

Example 2 using new Random(long seed):


 import java.util.Random;  
 public class Example7 {  
      public static void main(String[] args) {  
           Random rand = new Random(2); //Random instance with seed  
           for(int i=0; i<10; i++){  
                int randomNumber = rand.nextInt(2);  
                System.out.print(randomNumber+", ");  
           }  
      }  
 }  

Output:

 1, 0, 1, 0, 0, 1, 1, 0, 1, 1,   

Now with example 2 you will see that you will get same set of result irrespective of how many time you execute example 2.

But at the same time with example 1 where we used new Random() constructor we will get different set of number on different executions.

Till this we can say that definitely it is "seed" parameter which is responsible for giving same set of random numbers every time.

Time to dig to down to "long seed" parameter to understand the behaviour.

If you check the implementation of default constructor of Random class it uses seed value which will be different for every Random class object creation, where as for parametrized constructor of Random, seed value would be same for each object creation with same seed value.

So no matter how many random objects you create with same seed value, there sequence would be same. Also above statement is true for individual execution also.

If you want go to further into deep, better place would be to check code under Random class. It uses "linear congruential pseudo random number generator" to produce random number..to be honest still trying to understand the algorithm behind it :-)

That's all from my end on Random class. Let me know if you can add something more into it, definitely I am looking forward to update the post.

Friday 17 April 2015

Smart Java 7 JIT Complier

I don't know about other guys but definitely I used System.currentTimeMillis() to calculate method execution time while doing performance analysis using junits.

This post will cover how things can go completely wrong with such kind of performance testing.

Before we deep dive into the issue let me put a code snippet to explain the issue in better manner:

 public class APSum {  
      public static void main(String[] args) {  
           long then = System.currentTimeMillis();  
           ArithmeticProgression.sum(500000000);  
           long now = System.currentTimeMillis();  
           System.out.println("Elapsed time: " + (now - then));  
      }  
 }  
 class ArithmeticProgression{  
      public static double sum(int i){  
           double sum=0;  
           for(int index=0; index<=i; index++){  
                sum = sum + (double)index;  
           }  
           return sum;  
      }  
 }  

 As we can see in above code, system is calculating sum of numbers starting from zero to provided input number and returning the result. Also we are calculating execution time of method sum using  System.currentTimeMillis().

Now by going through code we can say that total execution time calculation depends on for loop which executed 500000000 time and "sum = sum + (double)index" calculation.

Here our thinking can go wrong with respect to JIT compiler and reason to say that JIT compiler is too much smart.

Let us understand sample of how JIT optimization works:
1. As you can see in main method, we haven't used variable returned from ArithmeticProgression sum method.
2. So when JIT see such kind of code which doesn't mean anything to caller, it just run empty for loop without doing any calculation of "sum = sum + (double)index".

So our assumption of calculating execution time can go completely wrong with the case.

Bottom line is JIT compiler uses a lot of background algorithm to optimize our code while running.

Note:
1. JIT compiler optimization is highly platform dependent so you cannot expect same behaviour across different platform.
2. Scenario is hard to reproduce on local system eclipse as JIT works on compilationThreshold flag of JVM so JIT comes into picture only when compilation reaches over compilationThreshold value.

Sunday 8 March 2015

Prefer Arrays.asList where you don't need grow able nature of list

Some interesting performance gain which I encountered while writing some test code and thought of sharing with you people.

May be you encountered situation like you have an array in your hand and some method need to be called which takes java.util.List as input parameter. In that case you need to convert that array into list and then call the method. 

Here I am assuming that target method doesn't want to change anything in input list. List content will be used only for method processing.

Now this can be achieved in 2 ways:
1. Create one arrayList, iterate through array and populate arrayList using add method,
2. Using Arrays.asList method to create a fixed size arrayList.

Here second way is far better than first way in terms of performance.

We will take one example to know performance gain achieved using second way.

 package com.example.performance;  
 import java.util.ArrayList;  
 import java.util.Arrays;  
 import java.util.List;  
 public class ListExample {  
      public static void main(String[] args) {  
           int maxNumber = 10;  
           long startTime = System.nanoTime();  
           List<Integer> numbers = new ArrayList<Integer>();  
           for(int i=1; i<=maxNumber; i++){  
                numbers.add(i);  
           }  
           long endTime = System.nanoTime();  
           long timeWithArrayList = endTime - startTime;  
           System.out.println("Time taken to populate array list using add method: "+timeWithArrayList);  
           startTime = System.nanoTime();  
           Integer[] backingArray = new Integer[maxNumber];  
           for(int i=1; i<=maxNumber; i++){  
                backingArray[i-1]=i;  
           }  
           numbers = Arrays.asList(backingArray);  
           endTime = System.nanoTime();  
           long timeWithArraysAsList = endTime - startTime;  
           System.out.println("Time taken to populate array list using Arrays asList method : "+timeWithArraysAsList);  
      }  
 }  

Let us look at result of this example:

 Time taken to populate array list using add method: 786449  
 Time taken to populate array list using Arrays asList method : 199285  
  
As you can see Arrays.asList took 1/3rd time when compared against list.add method.

(Note: processing time will vary, but asList method will always take less time when compared to add method.)

It is time to know the reason why add method takes more time if compared against Arrays.asList method.

If you check implementation of ArrayList add method, for each add method call array list checks for capacity using ensureCapacity method. So this is an overhead if you want fixed size list for some specific purpose.

That's all from my end on this topic. Looking forward for your suggestions and comments.


Sunday 22 February 2015

Java Serialization Part 1

I would not be shy away to say that I feel difficulty while questions or coding from java IO package comes in front of me. May be the reason is I haven't got that much coding exposure of IO operations. So things become confusing when we go into depth of java IO package.

Java serialization is also a small bit of that Java IO package, where our understanding is less.

I remember 2-3 years before one of my friend asked for real time usage of serialization and I told him that we can persist objects in a flat file instead of database. May be that was somehow correct answer but he was not convinced with that.

Today I am pretty much comfortable to answer his questions with example of EJB remote beans or high speed trading application where things cannot work without serialization.

Also mostly people get stuck when they have been asked to write a code which can serialize an object. 

So here I am going to share some real time example of serialization usage and code example related to serialization.

As I told earlier here are some real time example where serialization comes into picture:

1. May be few of you have worked on EJB and used it's remote bean. Now by saying any java class as remote bean, we mean that the class is available in some other JVM and client which want to invoke method of that remote bean exists in some other JVM. 
      Suppose if that method invocation need some object as parameter then somehow the object need to be passed from one JVM to other JVM via network. As network doesn't understand anything else 0 and 1, so this is done using serialization. There is complete cycle of java object serialization and de-serialization while invoking EJB remote method.

2. My next example will be stock exchange high speed trading application. As we know share price changes in fraction of seconds, so delay of millisecond in placing order will affect purchase cost tremendously. Trading application cannot go for relational database to store live trading transactions because of latency associated with it. Such kind of applications heavily rely on memory mapped file where serialization comes into picture.

Below picture can better explain definition of Serialization rather than words.


   


Code example for Serialization:

Here is my version of Serialization utility which can serialize and de-serialize an object:

 import java.io.FileInputStream;  
 import java.io.FileNotFoundException;  
 import java.io.FileOutputStream;  
 import java.io.IOException;  
 import java.io.ObjectInputStream;  
 import java.io.ObjectOutputStream;  
 import java.util.Arrays;  
 import java.util.logging.Level;  
 import java.util.logging.Logger;  
 public class SerializationUtils<E> {  
      Logger logger = Logger.getLogger("SerializationUtils.class");  
      private String filePath;  
      public SerializationUtils(String filePath){  
           this.filePath = filePath;  
      }  
      public void writeObject(E object){  
           FileOutputStream fos=null;   
           ObjectOutputStream oos= null;   
           try{  
                fos = new FileOutputStream(this.filePath);  
                oos = new ObjectOutputStream(fos);  
                oos.writeObject(object);  
           }catch(FileNotFoundException fnfe){  
                logger.log(Level.INFO, filePath+" not found");  
           }catch(IOException ioe){  
                logger.log(Level.SEVERE, Arrays.toString(ioe.getStackTrace()));  
           }finally{  
                try{if(null != oos)oos.close();}catch(IOException ioe)  
                     {logger.log(Level.SEVERE, Arrays.toString(ioe.getStackTrace()));}  
           }  
      }  
      @SuppressWarnings("unchecked")  
      public E readObject(){  
           FileInputStream fis = null;  
           ObjectInputStream ois = null;  
           E e = null;  
           try{  
                fis = new FileInputStream(this.filePath);  
                ois = new ObjectInputStream(fis);  
                e = (E)ois.readObject();  
           }catch(ClassNotFoundException cnfe){  
                logger.log(Level.INFO, Arrays.toString(cnfe.getStackTrace()));  
           }catch(FileNotFoundException fnfe){  
                logger.log(Level.INFO, filePath+" not found");  
           }catch(IOException ioe){  
                logger.log(Level.SEVERE, Arrays.toString(ioe.getStackTrace()));  
           }finally{  
                try{if(null != ois)ois.close();}catch(IOException ioe)  
                {logger.log(Level.SEVERE, Arrays.toString(ioe.getStackTrace()));}  
           }  
           return e;  
      }  
 }  

Now let us use this utility to serialize an object of employee class object.

Employee class(must implement Serializable interface):

 import java.io.Serializable;  
 public class Employee implements Serializable{  
      private static final long serialVersionUID = 1L;  
      private String employeeId;  
      private String name;  
      public String getEmployeeId() {  
           return employeeId;  
      }  
      public void setEmployeeId(String employeeId) {  
           this.employeeId = employeeId;  
      }  
      public String getName() {  
           return name;  
      }  
      public void setName(String name) {  
           this.name = name;  
      }  
      @Override  
      public String toString() {  
           return "Employee [employeeId=" + employeeId + ", name=" + name + "]";  
      }  
 }  

Main class using serialization for object of employee class:

 public class SerializationTest {  
      public static void main(String[] args){  
           SerializationUtils<Employee> utils = new SerializationUtils<Employee>("E:\\WrkSpace\\Test\\serialized_object.txt");  
           Employee emp = new Employee();  
           emp.setEmployeeId("123");  
           emp.setName("shashi");  
           utils.writeObject(emp);  
           System.out.println("done with serialization");  
           Employee deserializedEmployee = utils.readObject();  
           System.out.println(deserializedEmployee.toString());  
      }  
 }  

Here is the output which comes in console as expected:

 done with serialization  
 Employee [employeeId=123, name=shashi]  

If you are curious about what have been written in flat file, here it is:


That's all from my end in this part of serialization. We will cover more of serialization in second part where we will override standard writeObject and readObject method of ObjectOutputStream and ObjectInputStream respectively.



Wednesday 18 February 2015

Find second highest number in an array

Nowadays technical interview involves writing small small code to judge whether candidate is aware of java APIs or not and how he/she using APIs in different scenarios.

For any such scenario, there can be 10 ways in which you can code a given problem but among those only 2-3 ways will solve the problem efficiently.

Take an example to "find out second highest number in a given array."

My version of solving this problem statement would be using stack with size 2. It will hold highest and second highest number by traversing on given array.

Pseudo Code:
1. Create a stack with size limit 2.
2. Start traversing on given array arr[n] from i=0 to n-1, step 1.
3. If arr[i] is greater than HEAD element of stack then push arr[i] into stack, so that HEAD element       will go one level down and arr[i] will become HEAD.
    Else if arr[i] is greater than TAIL element of stack then replace TAIL element with arr[i].

This diagram will be helpful to understand stack push operation:



As we know that Stack class of Java has no size limitation so somehow we have to limit the size to 2. Therefore I created customized stack to store elements. As you can see I overridden push method and if stack size becomes equal to max size(in our case its 2), then I am removing bottom element before doing any push operation to stack.

 import java.util.Stack;  
 @SuppressWarnings("serial")  
 public class SizedStack<E> extends Stack<E> {  
      private int maxSize;  
      public SizedStack(int maxSize){  
           super();  
           this.maxSize = maxSize;  
      }  
      @Override  
      public E push(E item){  
           if(this.size() == maxSize){  
                this.remove(0);  
           }  
           return super.push(item);  
      }  
 } 


Class which is implementing discussed pseudo code:

 public class FindSecondLargest {  
      public static void main(String[] args) {  
           int[] arr = {4,3,2,9,4,5,3,7};  
           System.out.println(getSecondLargest(arr));  
      }  
      public static int getSecondLargest(int[] arr){  
           SizedStack<Integer> stack = new SizedStack<Integer>(2);  
           int head = stack.push(0);  
           int tail = stack.get(0);  
           for(int i=0; i<arr.length; i++){  
                if(arr[i] > head){  
                     head = stack.push(arr[i]);  
                }else if(arr[i] > tail){  
                     tail = stack.set(0, arr[i]);  
                }  
           }  
           return stack.get(0);  
      }  
 }  

As you can see that each element in the array will be compared against 2 elements at most which resides in stack.

Since this operation is repeated exactly n times(size of array), hence we can say that time complexity of this algorithm will be O(n).

At the same time I cannot say that my version is best approach for given problem statement, but one of efficient way to solve the problem.

If you can give some solution and time complexity in terms of some logarithmic function then definitely your solution would be more efficient that mine version.


Saturday 14 February 2015

Prefer Primitive over Wrapper classes

I was going through "Effective Java" book by Joshua Bloch and encountered a very interesting fact discussed on performance. So I thought, I must share this with you people.

It is about how you can optimize your code tremendously by just using primitive variable.

Here we are going to do the same thing in 2 different program, one using Wrapper Long class and one using primitive long.

We will take an example of doing sum from 1 to Integer.MAX_VALUE and analyze total execution time.

Example 1: Sum using wrapper Long class.

 import java.util.concurrent.TimeUnit;  
 public class LongTest {  
      public static void main(String[] args) {  
           Long sum= new Long(0);  
           long startTime = System.nanoTime();  
           for(long i=0; i<Integer.MAX_VALUE;i++){  
                sum+=i;  
           }  
           long endTime= System.nanoTime();  
           System.out.println(sum+" in "+TimeUnit.SECONDS.convert(endTime-startTime, TimeUnit.NANOSECONDS)+" seconds");  
      }  
 }  

If you execute above code snippet, total execution time will come around approximately 7 second.


Example 2: Sum using primitive long

 import java.util.concurrent.TimeUnit;  
 public class LongTest {  
      public static void main(String[] args) {  
           long sum=0;  
           long startTime = System.nanoTime();  
           for(long i=0; i<Integer.MAX_VALUE;i++){  
                sum+=i;  
           }  
           long endTime= System.nanoTime();  
           System.out.println(sum+" in "+TimeUnit.SECONDS.convert(endTime-startTime, TimeUnit.NANOSECONDS)+" seconds");  
      }  
 }  

If you execute above code snippet, total execution time will come around approximately 1 second.

Example 2 differs from example 1 just by variable declaration. It uses primitive long instead of Wrapper Long class.

So bottom line is by just using primitives, we saw huge performance gain. Here it is 6 seconds which is huge difference.

Conclusion: Always try to use primitive as much as possible over Wrapper class, because that reduces unnecessary overhead of auto boxing and un-boxing from JVM end.