How should we use Command/Query Segreration (part 2)

In this post I talk a little about command/query segregation. I will try to go deep on this topic using some example from real word.

public bool Update(Foo foo) { … }

Basically, this method updates the foo item in the database based on some unique id that Foo object contains. First of all, why we need to return a value? Basically update is a command and not a query. Because of this we shouldn’t return any value.
You would say: “Yep, okay… but I want to notify the caller if the update ended with success or not”. Not a problem. In this case we should throw an error. On the update flow, something went wrong and we need to notify the caller. In this cases don’t be afraid to use exceptions. Exceptions are good when are used in the proper place.
In our update method, what I would do. I would return void and if something happen in the update flow I would throw an error. If we think a little, we would realize that usually we need to tell the caller what exactly went wrong. Because of this throwing an exception can offer all the information that the caller need to resolve the update problem.
On the other perspective, we should think that a method should be self-explanatory. If the update command returns a Boolean value or a state object, than a user will need to read the documentation to see what the return value represents.
Another example that I would like to talk is the Add method.

public Foo Add(Foo foo) { … }

How many times did you saw this kind of methods? I saw this kind of add method a lot of times. The same question appears like in the first example? This is a command or a query? From how it is declared is a query. From the name of the method it’s a command.
Why we try to return the object that we added. Because we are setting an “id” to the object maybe, that only the Add method can do. In this case, why the Add method doesn’t set the “id” value of the object itself that is given as parameter. In this way we would end with a method that doesn’t return any value. This is what we expect from a command.
But be aware, command/query segregation principle is good, but can increase your code complexity, can increase the numbers of lines of code. On the other side, the code will be more easily to understand and used.
Don’t try to go to the other extreme and use only command/query segregation everywhere. There can be cases when violating this principle is acceptable. It is not a good thing but we can leave with it.
A classic example for this is the purchase item command. Normally when you implement this command you will need to return a receipt. I saw some implementations of this command that look like this:
public Receipt Purchase(List itemList,Customer customer) { … }
This method would return the receipt for this command. It we think a little is not so bad, because on our flow we need the same think in all our application: to make the purchase and generate the receipt after this. So basically we have 2 commands and a query in the same command if we look from the command/query segregation perspective:

• Making the purchase
• Generate the receipt
• Return the receipt

We could have 3 different methods that could be called separately. But because we don’t want have the same code written all over the application we can define a method that make this three calls.
In conclusion, command/query separation principle is a very good thing. These rules are not very easy to follow and in time may cost you. For example this principle helps us to write the code in an asynchronous manner more easily and to change in time. For example for billing we could have pay by phone, pay by credit card and so on. If we have this segregation already made it will be easier for us to change the current flow.
We need to find a balance around all this principle that we have.