I think that all of us heard about command/query separation and command and query responsibility segregation. This is a design principal that says that a method should be a command or a query, but not both. For example a command is a method that modifies an object, but don’t return any value. The same think is for a query. A query is a method that returns a value/values but don’t modify the object itself. This would be the simple definition of this design principal.
What happens if we don’t respect this principal? You would say nothing. I can have a method that change the value of an item and return the object that was changed for example. We can have a lot of side effects that can be undetectable at beginning. How many times you had to look on an implementation of a method or to look in the documentation to know exactly that command is executed and what value is returned.
Basically, when we execute a command we should not return any kind of value. If we want to notify the caller that something happen, we should throw an exception.
Using this pattern we can define a software architecture where we have the queries and the commands separately. But in a small project, maybe we don’t need this. Implementing such architecture in a simple project can increase the complexity without having the need of such architecture. We don’t need an airplane electric circuit for an iron.
This principal is very beautiful and we can use it complex project, but when we have simple project I wouldn’t go on this solution, because we consume a lot of time defining all the separation and in a lot of cases we don’t need different classes for commands and queries (for simple projects). Simple is better and for projects with low complexity we should have this in mind.
You would say what happens if the project in time will be more complex. In that case we can change the architecture of it. Not all the project will become mammoths (only a very small part of them). If we have a good design and a healthy code than the redesign will be very simple.
The command/query separation can be used not only when we design our application, but also when we write the implementation. When we write a method that execute a query and return some values we should never change an item in that query or execute a command. If we draw a well-defined line between these two tips of methods we will have the command/query separation principal used. The code will be more easily understand and used.
There are a lot of implementations of the update command that return the object itself that was updated. What do you think about this kind of methods? Should this type of methods return the object that was updated?
What happens if we don’t respect this principal? You would say nothing. I can have a method that change the value of an item and return the object that was changed for example. We can have a lot of side effects that can be undetectable at beginning. How many times you had to look on an implementation of a method or to look in the documentation to know exactly that command is executed and what value is returned.
Basically, when we execute a command we should not return any kind of value. If we want to notify the caller that something happen, we should throw an exception.
Using this pattern we can define a software architecture where we have the queries and the commands separately. But in a small project, maybe we don’t need this. Implementing such architecture in a simple project can increase the complexity without having the need of such architecture. We don’t need an airplane electric circuit for an iron.
This principal is very beautiful and we can use it complex project, but when we have simple project I wouldn’t go on this solution, because we consume a lot of time defining all the separation and in a lot of cases we don’t need different classes for commands and queries (for simple projects). Simple is better and for projects with low complexity we should have this in mind.
You would say what happens if the project in time will be more complex. In that case we can change the architecture of it. Not all the project will become mammoths (only a very small part of them). If we have a good design and a healthy code than the redesign will be very simple.
The command/query separation can be used not only when we design our application, but also when we write the implementation. When we write a method that execute a query and return some values we should never change an item in that query or execute a command. If we draw a well-defined line between these two tips of methods we will have the command/query separation principal used. The code will be more easily understand and used.
There are a lot of implementations of the update command that return the object itself that was updated. What do you think about this kind of methods? Should this type of methods return the object that was updated?
public Foo Update(Foo unsavedFoo) { … }
public Foo Update(int fooId, string fooName, int fooAge) { … }
OK.
ReplyDeleteLet's take a simple example:
You have a LIst page, that must support paging / filtering / sorting. Also the page should know the toal number of pages to display to the user.
Please implement in CQS .
That's easy - in the above example there are only queries, not commands that change the state of the system, I would say...
DeleteWe have only queries. We don't have any kind of commands on the items.
DeleteIf we add to the request the CRUD operation on the list of item than maybe... we should think about CQS
I think/hope I was tired when I was posting the question....
DeleteMaybe when we create an object we should return the object that was created.
ReplyDeleteNot really. However how do you go about creating an entity that hasn't yet an ID that is generated by the datastore? Like :
ReplyDeleteint CreateUser(string firstName, string lastName) { ... }
The response of all this questions will come tomorow with a post.
Delete