Object Oriented Design – A Class Diagram Walkthrough

Making a Class Diagram with UML Notation – Explained Step by Step:

 Note: Your feedback , issues are warmly welcomed and I would be happy to assist you with any problem you might be having .Class diagrams are often used in object oriented analysis and design to show the various identified objects , their types (classes) , their data and operations and how the relate , or communicate with other objects or classes. Here are some of the standard UML conventions and a sample class diagram to get you started: 

Identify the classes in the problem domain:

 This is the first and most important step for creating the class diagram. What classes should you include and how to identify them given a problem statement or a scenario? The solution is simple and some time should be spent with the client / system engineers /your teacher in case you are making an academic project or any concerned authority having the required business knowledge.The first step however is creating a problem statement or establishing a set of requirements that the software should fulfill. E.g. you need to create a small library management system. What are some of the requirements for the system? Let’s try to make a problem statement or the scenario for your design. For the sake of simplicity considers the following simple problem Statement. “The library management system would be used by the librarian to keep track of books, library members and the borrowing activity. Not all books can be borrowed by the members. Some books may only be available to review in the library; all such books are termed as Reference books. There are 2 types of library members students and college faculty members. Faculty members can also check out research papers and magazines where as students can only checkout books. The system need to send alerts whenever a book that has to be returned within a specific date is not returned. An email is sent to the librarian and the borrower. The system maintains a catalogue having a description of each book that is available in the library” 

Ways to identify Classes:

 Do the Noun Analysis: Go through the problem statement again and again and try to figure out all the nouns that you come across. In our case some strong contenders for the classes of the library management system would be: Librarian,Book,Member,Catalogue,Student_Member,Faculty_Member,Reference_Book,Issueable_Book,Alert 

Perhaps there might be some nouns that I skipped but why??  I did that intentionally because of either of the two reasons. Either there is no data associated with them. They have no role to play in the system i.e. they don’t have any functions or actions associated with them. So by this we mean that “In order for a noun to be an object or a class it should have some attributes or member data and some actions or member functions.

 Resolving the relationships between classes:

Object oriented analysis and design is all about relationships between objects. Look for a certain type of relationships. These include generalization/inheritance/specialization, composition / containership, aggregation/ collections and associations. All of these are explained briefly below and discussed in terms of our scenario. There are some key questions associated with object oriented design over here.  How data is passed between the classes? What classes should have the objects of another class?What classes create the objects of other classes?What are some of the utility classes that the application requires? All these questions play an important role in the design. Think in terms of hierarchy for generalizations/inheritence:First of all check for hierarchy that may exist between the classes or the objects.This would model the inheritance or generaliztion or the superclass/subclass relationship.In natural language this might resolve to “type of” or “can be”. If you find such words in a problem statement there might be a case of generalization. E.g. In our scenario take the three classes BOOK , REFERENCE_BOOK and ISSUEABLE_BOOK . So a book can be a reference book or an issue able book or you can say that reference books and issue able books are the type of books. They all should ideally share many characteristics which are common to all the three classes.What I mean to say is, whether it is any book (reference or issue able) they all have ISBN number, author a language, a topic e.t.c so what is the difference. The difference is implied by the business rule that reference books cannot be issued but only reviewed. So for reference books we might add a Boolean data member called m_bReference and set it to true or something. Same is the case with Member, Student_Member and Faculty_Member. Inheritance is indicated in the class diagram with a filled black triangle pointing towards the base class and connecters connecting it to all the derived classes. In our scenario the inheritance could be shown as the figure below:

 Think of the Part-Whole Relationships:

Ok so now you know what might be the possible hierarchies in your system and what might be the common data members and the common functions. What are the differences and how the qualify to be another class. Another important type of relationship in terms of Object Oriented Analysis and Design is the “Part of” or “Part Whole” relationship. In Natural language, you might come across “has a” words to quickly identify these kind of relationships. In the list of candidate classes observe which of the objects the part of other objects is. This can also be taken as collection.To give you some examples for common scenarios. A Windows form is a collection of many controls.An automobile is composed of many parts.A shopping cart has many itemsA university is comprised of many campuses Here you want to ask a simple question to resolve the type of relationship and to know whether it is an aggregation or a composition. “Is the whole destroyed when the part is destroyed” or “It doesn’t make a difference to the Whole if the part is destroyed”. Now going through this example list. A form would still exist if any of the control is destroyed; the form serves it purpose properly. This implies an aggregation .An automobile wont function properly if any of its parts are destroyed e.g. it cannot function if the tires, steering, battery is taken out so this implies composition. A university will still exist if a campus is destroyed. So it depends upon the scenario and the business rules.

Returning to our depicted scenario the only part whole relationship noticeable is perhaps the catalogue. A catalogue has list of books available in the system with a short description of each. The catalogue is not destroyed if a book is destroyed so I would use aggregation in the design. In UML aggregation is shown with an empty diamond where as composition is shown with a solid black diamond.

 The Uses relationship – Association:

 Many objects use the methods/functions of other classes for utility. The objects are somehow related might be statically or dynamically but there is not a hierarchy or a part whole relationship. Such relationships are called association and can be mostly identified with “makes use of” or “uses” in the problem statement. E.g. There is a class called “Pen” which exposes a method called “Draw ()”. There is also a “Shape” class having a method called“drawShape ( )” . In the implementation of “drawShape( )” it internally calls / invokes the “Draw ()” method exposed by Pen. In our scenario  A librarian uses the book and student information to issue a book.The catalogue needs a reference of the book to add or update itselfAnd probably some more: 

Here is the sample design which I created with Visio just for the sake of discussion and show how relationships can be expressed in UML in terms of object oriented design.The business rules that I assumed for this discussion were really simple just to make the understanding better.If it helped you in any way or you require some discussion in any area do supply your feedback by leaving comments for this post. Thanks.

       

Database Design Tutorial

A Database Design Tutorial:

There are several different kinds of database representations. An Entity Relationship Diagram or a ER model, an Enhanced Entity Relationship Diagram EER model, Object Relational Model etc. ER Model is the most universally followed database convention at least in my experience with software application development. Any enterprise application today would surely have a database connected to it, so I think that understanding database design and to be able to design databases is a “must have” for Software developers personnel. In this article I would walk through some of the essentials to give you the basic idea by giving you an example. Once you know how it’s done in a “Non Textbook” way I guess reading the text book or some advanced tutorial would be very easy and interesting. 

The First Step in ER modeling: 

The first step in designing the database is to know what are the business objects or so called “Entities”. How to determine what the entities are? The textbook says Read the problem statement or the scenario and try to identify nouns. From a collection of nouns it would be much clearer to determine what the entities are in your system. No doubt it’s a text book approach. I have had very few instances where I had a problem statement and I could do a noun analysis for an object oriented or the database design. So try to think in terms of the business and ask yourself why? This is called the “ROOT CAUSE ANALYSIS”. Ask yourself the question Why the customer wants to develop this system, who would be using the system, what features should be provided and most importantly what data needs to be stored?. 

 E.g. If someone wants to develop an online forum and you have to design the database for them, try to think what a forum does and generate a problem statement. The requirement would be vaguely “A discussion forum has many categories. In each category there are many topics and each topic may have many messages by different users. Old messages (let’s say 10 days old) are deleted automatically by the system. User can post new messages and edit their message” what would the database design be?It is evident with the problem statement that you need to keep track of categories, messages (post) and user. An entity in the system might be a strong or a weak entity. A Weak entity is one which cannot exist on its own in a system and requires a relationship with a strong entity (the one with a primary key) to identify its records. Such an entity is described by a double lined rectangle in the model. What attributes to store? What is a primary key? What is a partial Key? What are candidate keys? What is a super key? How to select them: Simply ask yourself what is the information that the system should store for a particular entity, importantly how that information would be identified at a later time? Is there a way, an attribute which uniquely identifies any record/row/ tuple in the table? What are some of the other important attributes the system has to store?
The answer to these questions would be very important to you in terms of database design. Primarily while resolving the ER model you would be converting all your entities to tables so what are the columns of that table. They are the characteristics of the entity. e.g. In our scenario what are the important attributes of a User (in this example we only require a loginid and a password but in a slightly complex scenario like a university registration system there might be many attributes like address , phone number , date of birth , previous college , social security etc Similarly what is the data about the categories we want to store (I guess the category name , Description , we would like to know the number of messages in the category and the topic with maximum posts and stuff like that but lets not talk about that here)The set of attributes which can together uniquely identify an entity instance constitute the Super key. The minimum set of attributes (subset of super key) which can uniquely identify a record in the database is called the candidate key. There can be more than one candidate key for the entity e.g. A Student number and the Social security number we use only one as the primary key of the table and the candidate key which we use to uniquely identify any record in the system is called the Primary Key ( The programmer assigned candidate key). Note in case of weak entities the identifying attribute is called the partial key instead of the primary key. 

Step 2 : What are relationships and why? What is cardinality and multiplicity?

 Once you have identified the entities, all of them think of the relationships and the cardinalities or multiplicity of the relationship. For the sake of simplicity check each of the entity against all others in order to clarify the design. Start with the User. How is the user related to Categories? A user can create a new category? If yes then we have a relationship. Then there are 2 things that need to be answered.1.       Can a Single User create more than 1 category2.       Can a Single Category be created by more than one user These two questions are the key to determine the cardinalities which would indicate how we resolve the database design in terms of tables.The relationship can have the following cardinalities: 

·          1 to 1·          1 to Many·          Many to 1·          Many to Many 

Similarly then compare if the user and topic has a relationship. That would be pretty similar to the first one. Now also compare the user with all the other entities.Note that relationship may also have attributes like the date/time when the message was posted etc.Relationships are indicated as diamonds in the design. A relationship with one of the participating entities is weak is called a weak relationship and is indicated with a double lined diamond. Resolving Relationship in the form of physical Database design:It’s a Science not an art J So you can generate the database physical design by pushing a button given the logical design , but the logical design is not a science its an art , thankfully otherwise we would be out of business JJ. 

 For 1 to 1 and many to many relationships create a new table for the relationship. Include the primary key of each participating entry. The union of all the primary keys which are referenced as foreign keys can uniquely identify any record in the table. So if the cardinality of both the participating entries is the same make a separate table. However if it is not take the primary key of the entity where cardinality is 1 and include it in the table where it is many as a foreign key. This is also called an identifying relationship sometimes. I would talk a bit about the enhanced ER in my next article; hope this would help you in learning the basics of database design.Some good database design tools are Dezign for databases by datanamic, Microsoft Visio, ERWin, Rational Rose and perhaps some other free design tool. If you have any queries, suggestions or feedback I would really appreciate it.This is a sample DB Design for a STUDENT GRADING SYSTEM which is a solved exercise from “Fundamentals of Database Systems” from ELmasri/ Navathe Question 16 and Question 18

Here is Question 18 where a new structure is to be designed that holds the previous college info , degrees and courses taken at another college:

Note: If I missed an important design concept you were looking for or is there anything you might wanna discuss ? Was this helpful to you ? I would appreciate your comments . Thanks

Remote Software Update Techniques ?

I have worked on various Self Service terminal applications so far . Kiosks for bill payment , Mini statement and balance inquiries and permit printing to self checkout solutions . One requirement that I always come across and I am not absolutely sure how and what would be the better way of solving it is remotely updating software . I want my exes and dlls to be updated and who does it ? Well it can be a windows service or any other application or perhaps a separate thread watching for the updates .

I have some experience with Microsft BITS and i wrote a wrapper for BITS in C# but all i could use the wrapper for was basically file transfer . The object didnt expose some of the core methodologies that Micrsoft uses to update windows or to install windows updates …..

So if you have worked with any such scenario where software can remotely update itself or u have a design in mind do feel free to discuss . I can provide the code whatever i have experimented , but i would really appreciate any ideas / logics or implementation that could help me make a better design for these kind of problems that i often run into … The Software/Configurations Remote update !! If u have the idea and not the time or patience to develope it feel free !! I will develope it for u !!

If you have any experience with making installation patches without hurting the installe application . I have tried Orca editor and the msp files , thats a pain a pain in the back side …. can u think of a smarter way ? Should we write a software which does that ? If you deploy software on a regular basis i m sure you probably might have some idea for a possible work around , I am amazaed why theres nothing from microsoft , doesnt sound that complicated :$

Cheers !!