Generics

Generics

The key motivation for generics is to provide meaningful type constraints between members. The members can be:

• Class instance members

• Class methods

• function arguments

• function return value

Motivation and samples

Consider the simple Queue (first in, first out) data structure implementation. A simple one in TypeScript / JavaScript looks like:

class Queue {
  private data = [];
  push = (item) => this.data.push(item);
  pop = () => this.data.shift();
}

One issue with this implementation is that it allows people to add anything to the queue and when they pop it - it can be anything. This is shown below, where someone can push a string onto the queue while the usage actually assumes that only numbers were pushed in:

class Queue {
  private data = [];
  push = (item) => this.data.push(item);
  pop = () => this.data.shift();
}

const queue = new Queue();
queue.push(0);
queue.push("1"); // Oops a mistake

// a developer walks into a bar
console.log(queue.pop().toPrecision(1));
console.log(queue.pop().toPrecision(1)); // RUNTIME ERROR

One solution (and in fact the only one in languages that don't support generics) is to go ahead and create special classes just for these constraints. E.g. a quick and dirty number queue:

class QueueNumber {
  private data = [];
  push = (item: number) => this.data.push(item);
  pop = (): number => this.data.shift();
}

const queue = new QueueNumber();
queue.push(0);
queue.push("1"); // ERROR : cannot push a string. Only numbers allowed

// ^ if that error is fixed the rest would be fine too

Of course this can quickly become painful e.g. if you want a string queue you have to go through all that effort again. What you really want is a way to say that whatever the type is of the stuff getting pushed it should be the same for whatever gets popped. This is done easily with a generic parameter (in this case, at the class level):

/** A class definition with a generic parameter */
class Queue<T> {
  private data = [];
  push = (item: T) => this.data.push(item);
  pop = (): T => this.data.shift();
}

/** Again sample usage */
const queue = new Queue<number>();
queue.push(0);
queue.push("1"); // ERROR : cannot push a string. Only numbers allowed

// ^ if that error is fixed the rest would be fine too

Another example that we have already seen is that of a reverse function, here the constraint is between what gets passed into the function and what the function returns:

function reverse<T>(items: T[]): T[] {
    var toreturn = [];
    for (let i = items.length - 1; i >= 0; i--) {
        toreturn.push(items[i]);
    }
    return toreturn;
}

var sample = [1, 2, 3];
var reversed = reverse(sample);
console.log(reversed); // 3,2,1

// Safety!
reversed[0] = '1';     // Error!
reversed = ['1', '2']; // Error!

reversed[0] = 1;       // Okay
reversed = [1, 2];     // Okay

In this section you have seen examples of generics being defined at class level and at function level. One minor addition worth mentioning is that you can have generics created just for a member function. As a toy example consider the following where we move the reverse function into a Utility class:

class Utility {
  reverse<T>(items: T[]): T[] {
      var toreturn = [];
      for (let i = items.length - 1; i >= 0; i--) {
          toreturn.push(items[i]);
      }
      return toreturn;
  }
}

TIP: You can call the generic parameter whatever you want. It is conventional to use T, U, V when you have simple generics. If you have more than one generic argument try to use meaningful names e.g. TKey and TValue (conventional to prefix with T as generics are also called templates in other languages e.g. C++).

Useless Generic

I've seen people use generics just for the heck of it. The question to ask is what constraint are you trying to describe. If you can't answer it easily you might have a useless generic. E.g. the following function

declare function foo<T>(arg: T): void;

Here the generic T is completely useless as it is only used in a single argument position. It might as well be:

declare function foo(arg: any): void;

Design Pattern: Convenience generic

Consider the function:

declare function parse<T>(name: string): T;

In this case you can see that the type T is only used in one place. So there is no constraint between members. This is equivalent to a type assertion in terms of type safety:

declare function parse(name: string): any;

const something = parse('something') as TypeOfSomething;

Generics used only once are no better than an assertion in terms of type safety. That said they do provide convenience to your API.

A more obvious example is a function that loads a json response. It returns a promise of whatever type you pass in:

const getJSON = <T>(config: {
    url: string,
    headers?: { [key: string]: string },
  }): Promise<T> => {
    const fetchConfig = ({
      method: 'GET',
      'Accept': 'application/json',
      'Content-Type': 'application/json',
      ...(config.headers || {})
    });
    return fetch(config.url, fetchConfig)
      .then<T>(response => response.json());
  }

Note that you still have to explicitly annotate what you want, but the getJSON<T> signature (config) => Promise<T> saves you a few key strokes (you don't need to annotate the return type of loadUsers as it can be inferred):

type LoadUsersResponse = {
  users: {
    name: string;
    email: string;
  }[];  // array of user objects
}
function loadUsers() {
  return getJSON<LoadUsersResponse>({ url: 'https://example.com/users' });
}

Also Promise<T> as a return value is definitely better than alternatives like Promise<any>.