Wednesday, 24 September 2014

Working Effectively with Legacy Code

Recently i attended Working Effectively with Legacy Code course by Michael Feathers

It was very good learning experience , he talks about how to work with code that does not have unit test or less unit test. He shared techniques can be used to improve legacy code and get better understanding of application.

This post is to share some of them before i forget:-)

Sprout Method/Class
This is pretty common technique but did't know that it has name. Adding new method/class sounds much easier than changing some existing code for new feature, so we should use this approach for any new feature that we want to introduce.

This approach can be used on existing method also to make it testable.
Work of caution that you don't want over do it !

Poor Man dependency injection
Everybody knows what dependency injection is, apart from some of the benefit it can also be used to make code testable, so for unit test you can have sample/dummy implementation that can be injected to your application code to make is testable. 

One of the problem with this technique is that it will result in method signature changes and that might mean that all the code in that call stack might need to change.
Just remember that you don't have to use spring to do this!

Extract and override
This is interesting one looks like silver bullet or Brahmastra for many problem.This pattern is used to have control on dependency that are hard to fake.
Assume there is function that makes some database/socket call and then does some calculation and you want to write unit test for calculation logic.

To make function testable you can do below changes
  • Extract database part of logic and put that in new function
  • Make that function protected. Thanks to OOPs , finally some good usecase for protected.
  • Write new class that extends original class and only override database interaction function(i.e was protected)
  • Use new class for your testing and you are done!

This is very powerful technique because you don't have to go through pain of changing  constructor/method parameter, so call stack remains intact.
Since it is based on method overriding, so you need to have discipline in team to make sure that fake class is only used for testing. 

Instance delegate
Used to test static function class. Create normal function that will delegate call to static function and during test create another class that will override new function that was created to add fake call .

Singleton make life very difficult from testing perspective, way to make it testable is allow injection of new singleton implementation and use that implementation for testing. 

Create interface to break big class
Although adding new method/class is much simpler but still lot of code is added to existing class/method and over period of time it becomes big i mean really big.

Approach to simplify class
  • Creation of interface without any method 
  • Class that you want to simplify should implements new interface
  • All the function that was using class should now use new interface and compiler will gives you errors about missing method and you can start moving methods to interface to fix the issue.

Main advantage for this approach is that you don't have prerequisite of unit test , you can take benefit of compiler/IDE feature. 

Method & variable dependency graph
 When class grows overtime and it does not adhere to Single responsibility principle, working out what functions goes where could be tedious.
You can draw dependency graph of variable & method to workout how things are related. This can help to come up with new classes will will adhere to single responsibility principle.

Identifying seam plays important role in working out hidden dependency. 
Definition of seam from dictionary is 
A line of junction formed by sewing together two pieces of material along their margins.

Definition in context of re-factoring is - part of code will enable testing.

for e.g functions performs some I/O operation and then some calculation, so if you want to test calculation without doing any major changes to core logic then you can "Extract & Override" approach to solve this.

Scratch Refactoring
I am sure you might have seen one big class/methods with 100s or 1000s of lines and you have to scratch your head to understand what this does.
To make things more interesting that part of code is very critical to company and you have to do some changes.
Scratch Refactoring is very useful for such type of situation

  • Take monster code that you want to simplify/understand
  • This type refactoring is only focused on understanding code, so it should be done in palin text editor so that you are not worried about compilation error that IDE generates.
  • Break big method in small and single concern method, simplify condition , delete some unused code

Main benefit of this approach is that you don't have to commit this in Svn/Git only purpose of this work is understand as much as possible by creating small code blocks.

While you are doing this you will get better understanding and code is simplified to extent that doing real re-factoring will be not be that difficult.

I must say it was very useful session, lot of learning and fresh perspective.
I have got copy of Working Effectively Legacy book and will read & practice it to get more better understanding.

Wednesday, 17 September 2014

TDD - Roman Numerals kata -

Interesting experiment on TDD, i knew it is better but did't expect that proving it could be this simple .

Reference Blog :

Wednesday, 6 August 2014

Compact String List

Whenever application memory profile is analyzed string is one of the most common object that comes right on top .

Java has String pool that solves problem to some extent and lot of interesting optimization has been done in string pool for JDK 6/7/8

Whatever is given by string pool can be easily implemented by WeakHashmap or Concurrent hash map but JVM implementation is very good,so no point in reinventing it. One of the overhead associated with string object is header of char array, each array has basic header cost and extra 4 byte for size of array

On 32 bit for char array - 8(header) + 4(length) = 12
On 64 bit for char array - 12(header) + 4(length) = 16

For each string value 12 to 16 bits is wasted.
Quick memory optimization that can be done is to allocate one big array and store values of multiple string in that array.

Just to visualized how it will look

With above approach we save array header cost but another overhead is introduced that we need another sets of variable to know which part of array belong to value1 or value 2.
Int array can be used to maintain index of different value in big char array,so we save 12 bits per string and that is significant saving when you have lots of string.

In this blog i will share experiment with such approach.
Lets get into code

First thing is storage of multiple string values in single char array.

Pretty straight forward code two array is required one char[] and one int[]
Add function will expand char & int array if required and add values to it.

Iteration over element is another tricky thing that needs to be handled for such compact structure, trove style foreach looks good fit for this.

Iteration code looks like


Memory Usage
Compact list trades off add speed for memory/search, lets have look at memory gain.
In this test text from ALICE'S ADVENTURES IN WONDERLAND url is split by space and it is added to ArrayList<String> and CompactStringList.

"Alice Adventure" book has 32.5K words.

For memory test i used Heinz Kabutz Determining Memory Usage in Java approach and it gave me consistent output so i stick with it for this test.

ArrayList takes around 1755 KB, CompactList takes around 355 KB.
So for this particular example CompactList takes around 80% less memory, gain is very significant.

Detail memory usage
Lets have look at detail memory usage. I used jmap to get top contributor for this test.

This gives better understanding of gain.
Char[] in compactlist takes around 60% less memory and String object is like almost nothing with minor overhead for int[].

So it seems good trade off for memory!

What next's
- One usage is building string pool using compactlist
- CompactList is append only structure any changes done for existing element like delete/update will require rebuilding CompactList
 - Iteration using traditional style will result in garbage creation because it has to build CharSequence, but that can be overcome by using foreach approach that gives access to chars of element.

Code is available @ github

Saturday, 19 July 2014

Saturday, 28 June 2014

MethodHandle returns back in java 8

Longtime back i wrote blog to compare performance of reflection/methodhandle etc.
Methodhandle was introduced in JDK7 and its performance was not that good as compared to reflection.

Recap of test result from JDK 7

MethodHandle is right in bottom as a result it was was not of much use in JDK7.
MethodHandle is used in java 8 lambda and i think due to that lot of improvement was done in it.

Lets have look at JDK8 numbers

MethodHandle implementation of JDK8 is back.
Open JDK mailing conversation has some details about type of optimization that is required in method handle. 

Code used for this blog is available @ github

Update -
One of the reader asked for raw numbers, numbers are in Mili Second

Friday, 14 March 2014

Off Heap concurrent counter

Concurrent counter are part of almost every system, it is used to collect data, thread synchronization etc. 
Java has good support of heap based counter.

There could be use case when you need counter that can be shared between processor.

How to build inter process counters 
 - Database 
This is first option that comes to mind, database sequence is counter that can be used by multiple process. All concurrency is handled by database. It is good option for starter but we know types of overhead(Network,Locks etc) you get with database. Only Larry Elision will be happy about it, not you!
 - Some Server
You could develop some server/middleware that provides such type of service. This option will still have network latency,marshal/unmarshal overhead.

 - Memory Mapped file
You could use memory mapped file to do this. I got idea from looking at thread-safe-interprocess-shared-memory-in-java presentation from PeterLawrey.

Challenges involved in multi process counter.
- Data visibility 
    Changes done by one process should be visible to all the process. This problem can be solved by using memory mapped file. Operating System gives guarantee about it and java memory model supports it to make is possible. 

- Thread safety 
Counters is all about multiple writers , so thread safety becomes big problem. Compare-and-swap is one option to handles multiple writers.
Is it possible to use CAS for off heap operation ? yes it is possible to do that , welcome to Unsafe.
By using using Memorymapped & Unsafe it is possible to use CAS for Off heap operation.

In this blog i will share my experiment of Memory mapped using CAS.

How ?
- How to get memory address
MappedByteBuffer is using DirectByteBuffer, which is off heap memory. So it is possible to get virtual address of memory and use unsafe to perform CAS operation. Lets look at the code.

Above code create memory mapped file of 8 bytes and get the virtual address. This address can be be used to read/write content of memory mapped file.

- How to Write/read in thread safe manner

Important function to look are readVolatile and increment
readVolatile reads directly from memory and increment is using unsafe to perform CAS on the address obtained from MemoryByteBuffer.

Some performance numbers from my system. Each thread increments counter 1 Million times.

Performance of counter is decent , as number of threads are increased CAS failures starts to happen and performance starts to degrade.
Performance of these counter can be improved by having multiple segment to reduce write contention.
I will write about it in next blog.

 - Memory mapped file is very powerful, it can be used to developed lot of things like off heap collections, IPC, Off heap thread coordination etc.
  - Memory mapped file opens gates for GC less programming.

All the code used in this blog is available on github.

Monday, 17 February 2014

AtomicInteger Java 7 vs Java 8

Atomic Integer is interesting class, it is used for building many lock free algorithm. Infact JDK locks are also build using ideas from Atomic datatypes.

As name suggest it is used for doing atomic increment/decremented, so you don't have to use locks, it will use processor level instruction to do so.
It is based on Compare-and-swap instruction.

Issue with CAS
CAS works on optimistic approach, it expects some failure, so it will retry operation, so in theory if there is no contention then it should work pretty fast.

There is another alternate way of doing same thing using Fetch-and-add.
Fetch-and-add is very different from CAS, it is not based on re-try loops.

Dave Dice compares CAS vs Fetch-and-add in atomic_fetch_and_add_vs blog, i can't explain better than this, so i will copy content from his blog

  1. CAS is "optimistic" and admits failure, whereas XADD does not. With XADD there's no explicit window of vulnerability to remote interference, and thus no need for a retry loop. Arguably, XADD has better progress properties, assuming the underlying XADD implementation doesn't have an implicit loop, but even in that case the window would be narrower than with Load;Φ;CAS.
  2. Lets say you were trying to increment a variable with the usual Load;INC;CAS loop. When the CAS starts failing with sufficient frequency you can find that the branch to exit the loop (normally taken under no or light contention) starts to predict toward the failure path. So when the CAS ultimately succeeds, you'll incur a branch mispredict, which can be quite painful on processors with deep pipelines and lots of out-of-order speculative machinery. Typically, this is in a piece of code where you don't want a long stall. There's no loop and no such issues with XADD.
Since fetch-and-add has predictable progress properties, so it is used for developing waiting free algorithms.
Unfortunately JDK 7 does not have support for fetch-and-add, one more reason why C++ people will be happy that C++ is great!

As we all know things do change & java community decided to added support for fetch-and-add in JDK8, one more good reason to migrate to JDK8.

In this blog i will compare performance of AtomicInteger from JDK7 & 8

Atomic Integer - JDK 7 vs JDK 8

In this test i increment counter 10 Million times with different number of threads. Thread numbers are increased to see how counter performs under contention.

X Axis - No of Threads
Y Axis - Ops/Second - Higher is better

JDK8 counter is winner in this case, best performance is when there is no contention , for JDK7 it is 80 MOPS but for JDK8 it close to 130 MOPS.
For single thread difference is not much , JDK8 is around 0.5 times faster but as contention increases performance JDK7 counter starts falling.

I will put another graph by removing 1 thread number, so that we can clearly see how these counter performs.

This gives better idea of how slow JDK7 atomic integer is, for 8 threads JDK8 counter is around 3.5X times faster.

Dive Into Code

JDK 8 - AtomicInteger
public final int getAndIncrement() {
        return unsafe.getAndAddInt(this, valueOffset, 1);

JDK7 - AtomicInteger
 public final int getAndIncrement() {
        for (;;) {
            int current = get();
            int next = current + 1;
            if (compareAndSet(current, next))
                return current;

JDK8 is using new function(getAndAddInt) from unsafe to do the magic. Unsafe has become more useful!

Dive in Assembly
To just confirm that all performance again is coming from fetch-and-add i had look at assembly generated.

JDK 8 
0x0000000002cf49c7: mov    %rbp,0x10(%rsp)
  0x0000000002cf49cc: mov    $0x1,%eax
  0x0000000002cf49d1: lock xadd %eax,0xc(%rdx)  ;*invokevirtual getAndAddInt
                                                ; - java.util.concurrent.atomic.AtomicInteger::incrementAndGet@8 (line 186)


0x0000000002c207f5: lock cmpxchg %r8d,0xc(%rdx)
  0x0000000002c207fb: sete   %r11b
  0x0000000002c207ff: movzbl %r11b,%r11d        ;*invokevirtual compareAndSwapInt
                                                ; - java.util.concurrent.atomic.AtomicInteger::compareAndSet@9 (line 135)
                                                ; - java.util.concurrent.atomic.AtomicInteger::incrementAndGet@12 (line 206)

Introduction of fetch-and-add type of feature in java will make it more suitable for high performance computing, we will see more wait free algorithm in java

Code used for testing is available @ AtomicCounterTest
Just compile for jdk7/8 and execute it.