A Misleading Name of Computer Science Concept: Dynamic Programming

Have you heard of these words before?
Probably yes!
First let's make a clear (well, not so clear) difference, clearer. Dynamic Programming is not about Dynamic Typing. Dynamic typing is a property of  a particular programming language. Languages like Lisp, Python, and Ruby are all dynamically typed languages. Unlike those, languages like C/C++, Java, and C# are statically typed language. The difference relies on whether the compiler checks certain things (types, overload resolution, etc.)  before running the program or not.
Dynamic Programming on the other hand is something very different. It's an ancient method for solving problems by basically dividing these problems into smaller, and easier to solve problems.

It's mainly focused on addressing these two issues:

1. Overlapping subproblems 
2. Optimal Substructures

I think the best way to explain these two fuzzy concepts is by using examples. To start let's see an example of using a famous dynamic programming technique called  Memorization. Assume that we need to implement a function that calculates a Fibonacci sequence. I know this is easy, but let's look at this first implementation:

static int FibClassic(int n, ref int numberOfStepsTaken)
        {
            numberOfStepsTaken += 1;
            if (n <= 1)
                return 1;
            Console.WriteLine("FibClassic called with: {0}", n);
            return FibClassic(n - 1, ref numberOfStepsTaken) + FibClassic(n - 2, ref numberOfStepsTaken);
        }

Note: on the above code I used two counter variables to count the number of times the method executed. I also printed the input on which the method is called every time, just to give you a hint of how dividing a problem can cause the same subproblem to be computed more than once --Subproblem Overlapping, remember! Now let's run the code given the number 6 as input and see what happens:

int z = 0;
            int x = FibClassic(6, ref z);
            Console.WriteLine("{0}: {1}", x, z);

And here's the result of this call:

As you can see the method has been called first with input 6 which is the initial input we passed in, then 5, 4, 3,  2, 2, ... oops!! Can you spot it? the method is being called on the same input more than once! Think about this for a while, pretty logical, yeah!? Our strategy is based on dividing the problem into simpler subproblems. For example to get the 6th item in the Fibonacci sequence we divide the problem into two smaller problems, getting the 5th item, and 4th item and adding them together, then to get the 5th, you should get the 4th and 3rd, etc. The next figure shows how is this working.

As you notice the overlapping happens when solving one part of the problem includes solving another part of the problem, in such a case we can take advantage of this and simply memorize the solution for the overlapped problem and each time we need that result, we don't have to compute it again, we just supply it from wherever we stored it. Here's the modified method to do it:

static int FibFast(int n, ref int numberOfStepsTaken, Dictionary store)
        {
            numberOfStepsTaken += 1;
            if (n <= 1)
                return 1;
            if (!store.ContainsKey(n))
                store[n] = FibFast(n - 1, ref numberOfStepsTaken, store) + FibFast(n - 2, ref numberOfStepsTaken, store);
            return store[n];
        }

If you run that same method on the same input (6) you should get 13 as the result (the same old result) but the number of iterations would be 11 which is about half the number of iterations the first method take. This doesn't seem to be a very huge enhancement, but let's see how the two methods act for bigger numbers. Running the first method of input equals to 30 we get the result  1346269 and number of iterations  2692537. Now running the second method on the same input (30) we get the result 1346269 -which is the same result- and number of iterations 59! That's a HUGE difference!

Now back to core, what does this have to do with the term Dynamic Programming??
Actually, it's a very misleading term, historically it was invented by a mathematician called Bellmen and he was at the time being paid by the US defense department to work on something else. He didn't want them to know what he was doing, so he made up  a name that he was sure it has no clue what it actually meant. Now we have to live with this forever :)

The Zen Of Python

I found these fantastic guiding principles of the design of Python published here by Tim Peters, and  once I saw them, I was like, Man! I ought to share these, so look them up on the site, or if you don't want to leave programming for COWARDS just yet, I'm quoting them here for you ;)  :

The Zen of Python


Beautiful is better than ugly.
Explicit is better than implicit.
Simple is better than complex.
Complex is better than complicated.
Flat is better than nested.
Sparse is better than dense.
Readability counts.
Special cases aren't special enough to break the rules.
Although practicality beats purity.
Errors should never pass silently.
Unless explicitly silenced.
In the face of ambiguity, refuse the temptation to guess.
There should be one-- and preferably only one --obvious way to do it.
Although that way may not be obvious at first unless you're Dutch.
Now is better than never.
Although never is often better than *right* now.
If the implementation is hard to explain, it's a bad idea.
If the implementation is easy to explain, it may be a good idea.
Namespaces are one honking great idea -- let's do more of those!

Those quotes are pure gold. One could think of writing them down on a board and keep that board hanged in front of his eyes, right above the computer screen!

Awesome, awesome stuff!

Into Git and Loving It

As it is gaining popularity day by day,and as everyone I know likes it, I thought .. well, let's give it a try.
Git is a powerful source control system. It's free, and open source. It's pretty simple to use, it's intended to handle both small and large projects, and it's so freakingly FAST.

Unlike SVN and TFS, Git is not centeralized. This means that every Git clone is a fully fledged repository with all versions and history information available.

Git is written in C (well, mostly) and is available for many platforms. If you are a linux guy you can download Git from here and if you are a Windows fellow then download msygit.

This post is aimed to be a tutorial introduction to Git. Together we will walk through the process of creating the repository, adding initial files to the repository, committing those files, changing them, creating branches, viewing differences, resolving conflicts, merging and finally pushing to the centeralized repository.

Now to get started let's assume that you want to create a simple project which is basically an html file, a bunch of javascript files and a css file. Pick whatever directory you wish to create your initial files. After you're done right click the directory containing your files and select "Git Bash Here" (assuming that you already installed msygit). A command window will show up, to initialize your repository enter:

git init-db

This will create the repository for you (and will also create a default "master" branch). Now enter:

git add -a

to add all the files in this directory to the repository. You can select specific files by specifying the required file name like so:

git add filename.css

The selected files are now tracked by Git to your repsoitory, to commit those files to the repository (the LOCAL repository, remember?) :

git commit -a

This will open vim for you to enter a commit message. Enter your message, and quite vi, then Git will commit your changes to the DB.

Note: 
If you're not familiar with vim or vi, to quite the editor saving the changes type ":wq" -that's colon wq- or ":q" to quit without saving. Or you can specify -m to the git commit command followed by the message you want in double quotes, like so: 

git commit -a -m "my initial commit"

Now try editing some files and then, to view the changes, type:

git diff

This will highlight the changes between your uncommitted version and the last committed version. If you want to see the history of your changes at any time, user:

git whatchanged

or use:

git whatchanged -p

to see the complete differences at each change.

The very cool thing about Git is how easily it enables you to create new branches. For example, if you want to create an expreimental function inside a class file you're working on or inside a javascript file but you want it away from the master branch, you can easily create a new branch, checkout this branch, edit your file, and all the changes will be completely isolated from the master branch. To create a new branch:

git branch testBranch
git branch

The first command will create testBranch for you and the second command will list all the available branches on your repository (by far there should be only testBranch and the master branch "master").

Note: The selected branch has an * displayed before it. 


Now enter :

git  checkout testBranch

to switch to the newly created branch. Try editing somefiles, and show the diffs.

git diff

You should now see the changes you made, commit those changes. And switch to the master branch:

git commit -a
git checkout master

If you look at the files you just edited when you were on testBranch -after switching to the master branch- you would notice that the changes you made while you were on testBranch are gone. However, if you switch to the testBranch again you will see your changes there. If you're happy with the changes you made on testBranch, you can merge those changes to the master branch by using the following command:

git merge testBranch

If there is no confilcts, you're allset. If there are conflicts you will have to resolve them manually and then commit the file. Git will show you which files have conflicts.
Now, if you are done with the test branch and want to delte it, use:

git branch -d testBranch

If you used to be a TFS and VSS guy like myself, take a deep sigh of relief and enjoy Git.
I will try to blog more on Git on the next posts, stay tuned!

The Selected ORM and Isolation Framework

A while ago I posted two little polls about the favorited isolation framework, and ORM. The results of the two posts were as follows

Which ORM?
1- NHibernate 49%(44 votes)
2- LLBLGEN 34%(31 votes)
3- Entity Framework 6%(6 votes)
4- LINQ to SQL 6%(6 votes)
5- Subsonice 2%(2 votes)

Which Isolation Framework?
1- Typemock Isolator 50%(10 votes)
2- Moq 45%(9 votes)
3- NMock2 5%(1 votes)
4- RhinoMocks 0%(0 votes)
5- Stubs 0%(0 votes)

For myself I voted for NHibernate and Moq. I know all of those tools are purely awesome, though.

So, dear reader, do you have any other suggestions for an ORM or an Isolation Framework?

A Quick Tip: The Different Classes Of Algorithms

Algorithms are the heart of computer science. They are the thoughts, the ideas, and the most fun part of this industry. Scientists categorize the various known algorithms into 4 classes: Logarithmic, Linear, Quadratic and Exponential. Let's look at those briefly:
1- Logarithmic Algorithms:
   This type is the fastest of those 4 classes. It has a run curve that looks something like this:

Where the x here is the number of items to be processed by the algorithm, and y is the time takes by the algorithm. As you can see from the figure, the time taken increases slowly when the number of items to be processed increase. Binary Search is a perfect example for a logarithmic algorithm. If you recall, binary search, divides the array into two halves and excludes one half each time it does a search. Here's a code example of binary search implementation in C#:

bool BSearch(int[] list, int item, int first, int last)
{
 if(last - first < 2)
             return list[first] == item || list[last] == item;
             
        int mid = (first + last)/2;
        if(list[mid] == item)
             return true;
             
        if (list[mid] > item)
                return BSearch(list, item, first, mid - 1);
 
 return BSearch(list, item, mid + 1, last);
}

2- Linear Algorithms:
  Runs in time, linear to the input items. It's curve looks like the following:

A linear search is a typical example of that, where one would traverse the array or, whatever data structure, item by item. An implementation of linear search looks like this:

bool LinearSearch(int[] list, int item)
{
     for(int i = 0; i < list.Length; i++)
          if(list[i] == item)
              return true;
        return false;
}

3- Quadratic Algorithm:    This one runs time grows to to the power of 2, with each increase in the input sizes, which means while processing 2 items the algorithm will do 4 steps, 3 items will take 9 steps, 4 items will take 16 steps, etc. The next figure shows how a quadratic curve might look like:

A famous example of an algorithm with quadratic growth is Selection Sort:

void SelectionSrot(int[] list)
{
            int i, j;
            int min, temp;

            for (i = 0; i < list.Length - 1; i++)
            {
                min = i;

                for (j = i + 1; j < list.Length; j++)
                {
                    if (list[j] < list[min])
                    {
                        min = j;
                    }
                }

                temp = list[i];
                list[i] = list[min];
                list[min] = temp;
            }
}

4- Exponential Algorithm:
This is the super slow of this list of four. It grows exponentially (that is, for 2 items it takes 2 ^ 2, for 3 items it takes 2^3, for 4 items it takes 2 ^ 4, etc.
Again algorithms are the key to computer science, the most fun part of programmer's  job. Choose your algorithms carefully and always try to improve.

Hope this helps.