Image of decorating the sidewalk with Chalk Art
Staying in tune with Polymorphic Composition where the majority of our code parts are Javascript or Typescript functions, we will begin to see a pattern whereby base functions can easily have more functionality simply using another function. This other function uses what is known as decorator pattern. The base function never changes.
By "never changing working code" we simply add a function which calls an existing function but returns a different result.
Consider this Angular route function:
/** Using the route, parse the id param
* @returns id e.g. route/id
*/
export function funcGetRouteId(
ar: ActivatedRoute
) {
let currentRoute = ar.snapshot;
let id = currentRoute.paramMap.get("id");
let person id = Number.parseInt(id);
return personId;
}
In the code above the id of the route is returned using angular's route snapshot. e.g. /person/id where id is the number.
The Decorator
/**Parses the route id
* @returns True if id is greater than zero
* (meaning existing person)
* False if person id is not greater than zero
*/
export function funcIsExistingPerson(ar:ActivatedRoute) {
// we reuse the existing code but
// provide a different result
let id = funcGetRouteId(ar);
if (id > 0) return true;
return false;
}
The caller's code looks like this:
if (funcIsExistingPerson(this.ar)){
// don't show icon until change;
this.SaveIconAction
.saveIconVisibility = "none";
}
None of the internals of how it was done are exposed, however, the naming of the function makes it clear what it's doing.
The caller merely needs to hover over the function to pick up the API comments.
A New Way of Thinking
When we start dialing-into what decorators do for us, we spend more time thinking about the base-function. Each base-function can have a minimum of 2n-1 decorators based on the number of parameters it requires.
When we start seeing code possibilities based on the Permutations of the parameters used, we are thinking outside of all specific solutions, and more into reusable parts.
Warning: It's pretty easy to get carried away with decorators, once we dial into thinking of Permutations. It's better; however, not to just write code because "someday" we may need it. Unless of course, we are writing utility programs.<\sub>
JWP2020 Functional Decorators
Top comments (6)
Hmm, reading this post, I failed to understand 2 things.
First, if
funcIsExistingPerson
function actually verified person existence (checked against DB or API or otherwise, that givenid
actually existed), how would hiding the logic from where theid
comes be useful in any way? We provide the active route anyway, there is no point in callingwhen we can write the following and obtain better transparency and code reusability of the
funcIsExistingPerson
function:Why? Well how about
funcIsExistingPerson( my.friend.user.id )
... You get the idea.Second, since the
funcIsExistingPerson
does not check for person existence, how is better than simple? πIn case the point is encapsulation of the
id > 0
comparison, a much more reusable approach would simply be to create a generic functionisIdValid
and thenwould:
id
(encapsulateid > 0
check)funcIsExistingPerson
)For your first question, the example shown was to demonstrate the most simple example of a decorator possible to illustrate the concept. It's usefullness is questionable except where it's reused throughout the project. Which is the jest of this series.
The function turns out to be a business rule based on the person/Id route design, using numbers are not self evident whereas the function name is.
The other advantage is one time one place for that rule. Composing the function as you suggest is more visible but requires more knowledge of the API. Plus we have to change the implementation. One of the points in this series is to never change working code.
Knowledge of entire functional modules, becomes a large problem in using large libraries (Angular, React, Vue et. al.) because we don't really become that overtly knowledgeable until we've spent plenty of time with them.
Can we use this on other routes? No, this function is only for the person route which is why we put it in the person Typescript module.
I agree that the concept you describe is usable, but the example used is not.
The example mixes routing with business logic. I would never ever do that.
Furthermore, "never change working code" is just an excuse for not writing proper tests. A well tested aplication can be refactored without introducing bugs. The refactor I propose is "inversion of control", which is one of the most basic possible refactorings.
Yet, I agree that different working environments require different development practices. If "never changing working code" is one's dogma, it might make sense to encapsulate stuff into one-off functions.
This article shows the huge advantage of this series. About halfway down we see how intellisense auto discovers all functions of a module. If each function is atomic and each module they are in, is properly named, there is no concerns regarding mixing of business logic with routing.
Our Person module/object is the aggregator of Person functions. This is Separation of Concerns in action. However as you stated before, we could easily composed functions explicitly, by using two modules: One for routing to get the ID, the other from Person to apply the rules. Or as suggested we use a decorator function with just one call. Because functions are fully atomic there is no mixing of Business logic, a function either has business logic or it doesn't. We may want to create a Business Logic module to keep them in the same place, but as far as function discovery they all become Viewable to intellisense with no distinction on where they came from other than the import statement.
Never change working code is "no excuse", it's the implementation of the Open/Closed principle. The principle is older than Javascript itself. Javascript breathes new life into the principle because functions are first class citizens. Open for extension is perfect when using the function. Closed for modification means all our previous tests should always work. We do however, need to create new tests for all new code.
One off functions in this example adhere to proper functional decorator techniques. Most newer programmers don't know about decorators. The results are always monolithic and repeated code patterns. This is always a total nightmare.
I believe I understand your point of view now π
I might have mistaken this example for a general one, it makes more sense in the context of Angular.
This is not an Angular only concept it works in any Typescript or Javascript code. It's not limited by Angular in any way.