get arabic initial medial final characters unicode number

As the notion of initial/isolated/final forms are specific just to Perso-Arabic script (I'm not sure, maybe to some historically related scripts), they are not universal, so I don't know that there is a ready-to-use methods of doing what you want; I doubt it. (Please see: http://en.wikipedia.org/wiki/Perso-Arabic_script[^].) You can easily do it by yourself.

Something like this:

enum ArabicCharacterClass {
    Digit,
    Letter,
    Punctuation,
    Symbol, //?
    //something else?
}

enum ArabicContextualForm {
   None, //?
   End,
   Middle, 
   Beginning,
   Isolated
}

struct ArabicCharacterDescriptor {
   public ArabicCharacterDescriptor(char codePoint) {
       CodePoint = codePoint;
       //can throw exception if the code point is not Arabic (Perso-Arabic)
       //calculate the other members using the dictionaries (see below) if required
   }
   public char CodePoint { get; private set; }
   public ArabicCharacterClass CharacterClass { get; private set; }
   public ArabicContextualForm ContextualForm { get; private set; }
   public string Name { get; private set; }
   public string Din31635 { get; private set; }
   public string IPA { get; private set; }
   //something else?
   public override string ToString() {
       return //... calculate from other members:
       // whatever you want as a default ASCII string representation; could be your notation
   }
}

Find out the Perso-Arabic subset of the of the Unicode code points (there are many, take patience):
http://unicode.org/[^],
http://www.unicode.org/charts/PDF/U0750.pdf[^],
http://www.unicode.org/charts/PDF/U08A0.pdf[^],
http://www.unicode.org/charts/PDF/UFB50.pdf[^],
http://www.unicode.org/charts/PDF/UFE70.pdf[^].

Traverse all this data and create an instance of ArabicCharacterDescriptor of each character, classify them all and put some collection. Its the best to put then all in a collection based on key-value pair; for fast search. You many need to have two or more container with the same values (say, the values will be the instances of ArabicCharacterDescriptor), but indexed with different keys, for a fast search by code point, name, IPA or whatever else.

Please see: http://msdn.microsoft.com/en-us/library/5tbh8a42.aspx[^].

The collections based on key-value pairs are:
http://msdn.microsoft.com/en-us/library/f7fta44c.aspx[^],
http://msdn.microsoft.com/en-us/library/ms132319.aspx[^],
http://msdn.microsoft.com/en-us/library/xfhwa508.aspx[^].

The one most usually used is System.Collections.Generic.Dictionary<TKey, TValue>.

Basically, that's it.

Good luck,

—SA