While we are hearing tips on Clean code, the most common tip is maintaining SRP while writing class or methods, in a broader scope Module/Package/Service/API.
But Irony is, SRP is the most powerful but obscured principle in the Design toolbox, this throws a web of confusion and you are stoned by thinking should I use it in the right way or not? That's why I like to call SRP Medusa. Most of the time it succumbs us and we are ending up with anti-KISS code but If we use it in the right proportion, then we can create a cohesive and robust architecture.
It is time to quickly recap the definition of SRP.
“The single-responsibility principle (SRP) is a computer-programming principle that states that every module, class, or function in a computer program should have responsibility over a single part of that program's functionality, which it should encapsulate. All of that module, class, or function's services should be narrowly aligned with that responsibility.”
Let’s demystify the Confusions
What does it mean by SIngle responsibility?
How to define a single responsibility in the Class/Orchestrator method./Package level?
How SRP can be maintained in method level/ Class level/Service level/, isn't it contradictory? If a module/Service maintains SRP then what multiple Classes/methods inside it maintains?
1. What does it mean by SIngle responsibility?
According to Robert C Martin, Single responsibility means each component has one reason to change, if you have to modify the component for multiple reasons then it breaks SRP. but the question is, what does it mean by reason, how do I find the reason? How can we restrict components to serve just one reason?
let me take an example code snippet to discuss it more, I keep this example confined to the method level.
public void displayEvenNumberfromInput(){
List <Integer> numberList = CommandLine.getNumberInputAsList();
List<Integer> evenList = new ArrayList<Integer>();
for(Integer number : numberList){
if(number %2 ==0){
evenList.add(number);
}
}
System.out.println(“Even List :: ” + evenList );
}
Now read the method carefully and ask yourself, in which reason, you have to edit the method, those will be the reason and we need to separate them out by wrapping by a class/method according to the needs.
The above example invoked a method on an associated class called CommandLIne, but think If requirement changes and we do not want to read from Command line but from JSON, so you have to modify your code. So this is one reason to change.
If I want to change the algorithm, then also I need to change the method, so this is the second reason to change.
At last, rather than print in the console If I want to save the output in File so, this is another reason for the change.
So, I can see the above method has three reasons to change, a small method but doing so much.
2. How to define a single responsibility in the Class/Orchestrator method/Package level?
Take away from the above paragraph is, by asking questions finding the reason for the change. If you are new I would suggest writing a method then checking for the algorithm steps, extracting out the non-related steps, do not separate related step to achieve granular level SRP, it will create Code smell and increase coupling,
Now for architecting service or class study the acceptance criteria and encapsulate non-related acceptance criteria.
Another approach is trying to find out commit which classes are often committed together pay a closer look at them any find dependency you can extract out from them, or in a class change the field name and found what are the methods gives you the error, probably they are very related and closed, so avoid to segregate them in a separate class, again these are just strategy but to use SRP in the right proportion, not making code too granular, maintaining KISS and achieving cohesion need well-versed knowledge in the domain, and architecting.
sometime you will see some methods/Class are breaking SRP but if you pay closer look reasons are strangled together so no need to separate them out it will increase the overhead of maintainability cost,
Example:: Like if the minimum balance is over some amount then do the debit operation, here validation and debit operation are linked together so let it be in the same class.
So, domain knowledge and what context you are writing the class/methods is important, initially while architecting a class/module may be wrong in the long run but next time when you need to touch the component try to understand why you need to touch that component and separate out that concern. So Refactor is a must - to get a proper SRP, you have to do multiple refactoring, so take time, take a buffer in your story point. Refactor is a first-class citizen while you doing code
Let see How we can curb out the responsibility from the last example?
public void displayEvenNumberFromInputSource(){
List<Integer> numberList = processFactory. processInputSource();
List<Integer> evenList = algoFactory.findEven(numberList)
displayable.display(evenList);
}
Now I abstract the three reasons and separate them out into different classes, I am assuming there are multiple Inputs Source we need to support so I separated it into a different class, If that is not the case only we have to read from the command line, I will separate them in a separate method in Same class, Now the hardest question comes, you may say my class breaks SRP as it handles three concerns even the orchestrator method displayEvenNumberFromInputSource handling three separate concerns so sure breach of SRP.
With this context, I will dive deep into the most obscured part of SRP.
3. How SRP can be maintained in method level/ Class level/Service level/, isn't it contradictory? If a module/Service maintains SRP then what responsibility Classes/methods maintain?
I personally think, there is one part missing in SRP definition, the term hierarchy/Level, SRP does not make sense if you apply it to a specific point it is a range.
It is like a Pyramid, like our Food chain Pyramid. In a particular domain space if we visualize it using a bottom-up approach, it is starting from method level then responsibility merging in Orchestrator method, then Class, then package, then service/module level. It goes from more specialization to generalization in an upward direction.
So, in a top-level, it will do one responsibility which consists of multiple sub responsibilities, and each level of the SRP pyramid delegates the responsibility to the next level and it is continuously going on until it reaches the bottom level until each method handles a single responsibility.
But at the same level of the SRP pyramid, no component will perform more than one responsibility of that particular level. Each component level must not cross each other then it is breaching the SRP pyramid level and your higher-order component will know the lower order components, so against the Dependency Inversion principle. Only the upper-level talks to the next level through encapsulation.
Think about the HR department, what is its job in the organization -- managing Human resource, In an organization This is one of the SRP Pyramid Tower, Payroll, Finance are the different SRP Pyramid Towers. Each tower talks to the other tower through an Organization in-house API contract and each tower specifically handles one broad responsibility of that organization.
Now if we deep down to HR SRP Pyramid, next level we can have multiple sub responsibilities like managing different geography, different policies, so each time you deep down to SRP Pyramid you are dividing the responsibilities to sub responsibilities and give it to a more specific component to do that job.
That is the main essence of SRP, Divide, and conquer the work.
In SRP Pyramid, You can not compare top-level component with bottom level and then thinking top-level component is breaking SRP, you are then comparing Apple with Oranges.
So I would like to rephrase the definition of SRP.
“The single-responsibility principle (SRP) is a computer-programming principle that states that SRP starts with a SRP Pyramid , where a module, class or function in a computer program placed in a level and in that level they have responsibility over a single part of that program's functionality, which it should encapsulate from the next level . All of that module, class or function's services should be narrowly aligned with that responsibility.”
Or in a simple
In SRP Pyramid each component in a Level has one reason to change.
Posts
Post a Comment