Refactoring:improving the design of existing code (book review)

Standard

Hi, Dear readers! Welcome to my blog. On this post, I will review a famous book of Martin Fowler, which focus on a refactoring techniques. But after all, why refactoring matters?

Definition

According to Fowler, a refactoring consists of modifying code in order to improve his readability and capacity to change, without changing his behavior. When refactoring, our objectives is to make the code easier to be read by humans and also improving his structure and design, making changes motivated by business rules easier to implement. Other benefits are that a cleaner code makes it easier to spot bugs, alongside fastening the development of new code on top of a well organized production code.

When refactor?

Fowler defends that refactoring should be done on 3 situations:

  • When you add a new functionality;
  • When you find a bug;
  • When you do a code review;

On this situations, you are forced to make changes on the code structure, making ideal situations for refactoring.

Pitfalls

When refactoring, there is some common pitfalls that could hinder the refactoring. The most common ones are the databases and the interfaces from the code.

Database schemas could be hard to change, specially if the database is old with millions of rows. This produces a splash effect on the code that manipulates the database, making more difficult to make changes on the code. Also the interfaces (APIs, libraries or even a single class inside a component) could be a challenge to refactoring, since a change on a interface could cascade to a change on lots of client’s code.

In order to solve this problem, the better approach for databases is to isolate the database’s logic on his own layer, allowing the “dirtier” code to be evolved in a more controlled manner. As for the interfaces issue, the better approach is to allow the old and new interfaces to coexist, while a migration work is conducted.

When not refactor?

According to Fowler, there is one situation when you shouldn’t refactor: when the code is so bad, that it is better to be written from scratch. This is a difficult rule to be measured as to when the code is bad enough to be rewritten. Some good hints could be if the code is infested by bugs or if it is identified that it has so much refactoring points, that fixing it up could end up rewritten most of the code.

Refactoring and performance

Sometimes, when refactoring, we could incur on refactorings that cause some performance degradation. Of course that it is up to the business to measure up how much this degradation is unbearable to meet the requirements, but as a general rule, we can assume that a more organized code is a easier code to fine tune. So, if we refactor first and improve his readability and design, it will be easier to make a performance tune later.

Unit testings

Another key point defended by Fowler is the need to develop unit tests for the code. With unit tests, we can develop refactorings in small steps (“baby steps”), receiving rapid feedback from the tests, so if anything breaks we can easily and fast make fixes during the refactoring process.

Refactoring catalog

Here there are some brief descriptions of some of the refactoring patterns that I found more interesting. Complete descriptions with examples can be found on Fowler’s book, that you can find on the links at the end of this post.

Extract Method

This refactoring consists of taking some code that can be grouped together and extract the code as his own method, this way improving readability.

Introduce Explaining Variable

This refactoring consists of taking a big and complex conditional and simplifying by turning his operators onto variables, this way making the conditional more self explanatory.

Replace Method with Method Object

This refactoring consists of a situation when you have a method that is better to have some code extracted to his own method, but it refers to a lot of variables that hinders the operation. On this case, this refactoring applies, consisting of taking all the variables and the method and moving to a new object, making a easier environment to make the extract method refactoring.

Move Method

This refactoring consists of moving a method from one class to another. This makes sense when the old class has less uses for his own method then the class he is moved to.

Extract Class

This refactoring consists of when you have a class that it is doing work that could be better organized if divided in two. On this case, we move the common behavior and data (methods and fields) that could form a new class and move it, making a delegation from the old one.

Remove Middle Man

This refactoring consists of when a class has lots of delegating methods to another class, which introduces unnecessary code. On this case, we create a accessor to the instance of the object itself, making it so the callers can call the methods from the class themselves, so after creating the accessor we remove all delegating methods.

Consolidate Conditional Expression

This refactoring consists of when you have several conditionals that returns the same value. On this case, we refactor the conditionals by creating a single one, commonly by creating a method, making the code more clear and simple.

Remove Control Flag

This refactoring consists of when you have a conditional flag that controls the behavior inside a loop. By using control commands like break and continue, we can remove the control flag, simplifying the code.

Replace Conditional with Polymorphism

This refactoring consists of when you have a method on a class that has conditional behavior depending on the type of the object. On this case, we extract each leg of the conditional and create a subclass around the different behaviors, until the method turns out to be empty, in which case we turn the method to abstract on the now superclass of the hierarchy. We may have to change the constructor of the class on a factory method.

Introduce Nul Object

This refactoring consists of when we have various null checks for data on the callers of a object. On this case, we create a object to represent null, that returns all the default data that should be used when the data was null. This way, we don’t have to make checks for null on the callers anymore, since the behavior on the null object will cover the check’s circumstances.

Preserve Whole Object

This refactoring consists of when you have a method call that is preceded by calling several data accessors to get lots of data from other object to pass as parameters for the call. On this case, we change the method to pass the object itself, removing the calls from the data accessors by moving them to inside the method itself. This refactoring not only simplifies the code on the caller’s side, but also simplifies changes if the method needs more data from the object passed by parameter on the future.

Replace Constructor with Factory Method

This refactoring consists of when we want to include more behavior on a constructor then it normally has it. This is specially true on class hierarchies, where the object construction must reflect the corresponding subclass depending on the type of the object. On this case, we change the constructors to a more restrictive access (private or protected)  and create a static factory method at the top of the hierarchy, allowing the dynamic creation of the objects.

Replace Error Code with Exception

This refactoring comes in handy when we have code that returns error codes when something breaks. Error codes are common on languages such as Unix and C, but on Java, we have a much more powerful tool: exceptions. With exceptions, we can easily separate the code that fix the errors from the normal code. So in this case, our best approach is to change the error code’s return to exception throws, which make a much more readable and organized structure.

Replace Exception with Test

This refactoring occurs when we have a code on a try-catch block, that has on the catch block some code that could be moved to be performed before the error occurs. This is typically found when we have predicted errors that can occur on some cases, but we use the catch block as part of our program’s logic. By changing the logic on the catch block to a test (if) before the code that breaks, we can remove the try -catch altogether, making a better readable and consistent code, that doesn’t rely on errors to work.

Extract Subclass

This refactoring consists of when we have some methods and fields that are used only on some instances of the class. On this case, we move the methods and fields to a new subclass, where they could be better organized and maintained.

Extract Superclass

This refactoring is opposed to the previous one, since we create a superclass instead of a subclass. If we have identified common behavior from two different classes, we create a superclass with the common behavior from the two and make both of them subclasses from the created superclass.

Form Template Method

This refactoring consists of when we have two classes that has methods with equal or very similar logic, that needs to be called on a certain order. On this case, we equalize the interfaces of the methods of both classes to be equal and creates a superclass where we move the common methods from both classes and create a orchestration method for the order of the calls. This improves the code on reusability and hierarchy organization.

Conclusion

And so we conclude our introduction to Martin Fowler’s Refactoring book. With good didactic and good examples, the book is a must read that I highly recommend!

Thank you for following me on this post, until next time.

Buy the book now!