Generics page

Overview

In this section, you will:

• Understand the fundamentals and objectives of generic functions
• Create generic classes
• Combine recursion with generic classes
• Create recursive generics

Objective 1: Basic Generics

Generics are a way of writing abstract code that allows the determination of types to be handled when the code is used. Generics let us reuse code for different types and improve maintainability. Let's see how with a small example.

Consider a function that wraps a value in an object:

function wrapAsValue(value) {
    return {value: value};
}

Ideally, you’d want to use this function to wrap all sorts of values:

let fourObj = wrapAsValue(4);  //-> {value: 4}
let hiObj = wrapAsValue("hi"); //-> {value: "hi"}

And you might want to pass those objects to other functions:

function getDollars(obj: {value: number}){
    return "$"+obj.value.toFixed(2)
}

function getMessage(obj: {value: string}) {
    return obj.value + " world";
}

getDollars(fourObj); //-> "$4.00"
getMessage(hiObj);   //-> "hi world"

But watch out! The following will not error until runtime because strings do not have a toFixed() method.

getDollars(hiObj);  

You don’t see a compile time error because hiObj object looks like {value: any} to TypeScript.

Getting a compile time error can be solved in a variety of inelegant ways:

• Way 1 - Define the type of the variables:

  let fourObj: {value: number} = wrapAsValue(4);
  let hiObj:   {value: string} = wrapAsValue("hi");
  

• Way 2 - Write multiple functions:

  function wrapStringAsValue(value: string) {
    return {value: value};
  }
  function wrapNumberAsValue(value: number) {
    return {value: value};
  }
  

• Way 3 - Overload wrapAsValue signatures:

  function wrapAsValue(value: string): {value: string};
  function wrapAsValue(value: number): {value: number};
  function wrapAsValue(value: any) {
      return {value: value};
  }
  

With generics, this problem can be solved more simply:

function wrapAsValue<MyType>(value: MyType): {value: MyType} {
    return {value: value};
}

let fourObj = wrapAsValue<number>(4);
let hiObj = wrapAsValue("hi");


function getDollars(obj: {value: number}){
    return "$"+obj.value.toFixed(2)
}

function getMessage(obj: {value: string}) {
    return obj.value + " world";
}

getDollars(fourObj);
getMessage(hiObj);
getDollars(hiObj);

The <MyType> part of the wrapAsValue definition is the Generics part. This <MyType> allows us to capture the type the user provides so that we can use that information later. In this case, we are using it to specify that the return type is an object with a MyType value property ({value: MyType}). This allows us to traffic that type of information in one side of the function and out the other.

We can call generic functions in two ways:

We can explicitly pass the type:

 wrapAsValue<number>(4)

Notice that <number> acts as a special set of arguments. Instead of arguments passed like func(arg1, arg2, arg3), generic type arguments are passed like func<Type1, Type2, Type3>.

The type can be inferred:

wrapAsValue("hi")

Notice that we didn’t explicitly pass the type n the angle brackets (<>). Instead, the compiler just looked at the value "hi" and set MyType to string.

Setup 1

✏️ Create src/generics/return-last.ts and update it to be:

export function returnLast(arr) {
  return arr[arr.length - 1];
}

Verify 1

✏️ Create src/generics/return-last.test.ts and update it to be:

import { returnLast } from "./return-last";
import { describe, it } from "node:test";
import { strict as assert } from "assert";

describe("Generics", () => {
  it("Returns last element with a string", () => {
    const lastString = returnLast<string>(["A", "B", "C"]);

    assert.equal(typeof lastString, "string", "It returns a string");
  });
  it("Returns last element with a number", () => {
    const lastNumber = returnLast<number>([1, 2, 3]);

    assert.equal(typeof lastNumber, "number", "It returns a string");
  });
});

Exercise 1

Update the return-last.ts file to inform the function that it will be accepting an array of a certain type and return a single element of the same - type.

Have issues with your local setup? You can use either StackBlitz or CodeSandbox to do this exercise in an online code editor.

Solution 1

export function returnLast<T>(arr: T[]): T {
  return arr[arr.length - 1];
}

Objective 2: Generic Classes

Generic classes are quite common.

const cardNumber = new Subject<string>();

cardNumber.next("1234")

In the example above, RxJS subjects are a generic class that can publish values of a particular type.

Let’s look at making a basic class to collect a list of things.

class Collection {
  private list: any[] = [];
  push(thing) {
    this.list.push(thing);
  }
}

let myList = Collection();
myList.push(25);
myList.push('25');

The good - we can push any type to this list.
The bad - we can push any type to this list.

myList now holds an assortment of types and will be a likely source of runtime errors.

Let’s build a generic Collection class instead.

class GenericCollection<T> {
  private list: T[] = [];
  pushItem( thing:T ) {
    this.list.push(thing);
  }
}

Now when we initialize this class we can specify a type to use.

class GenericCollection<T> {
  private list: T[] = [];
  pushItem(thing:T) {
    this.list.push(thing);
  }
}

const myListOfStrings = new GenericCollection<string>();
myListOfStrings.pushItem('booop');
myListOfStrings.pushItem(5);
//error Argument type of '5' is not assignable to parameter of type 'string'


const myListOfNumbers = new GenericCollection<number>();
myListOfNumbers.pushItem(5);
myListOfNumbers.pushItem('boop');
//error Argument type of '"boop"' is not assignable to parameter of type 'number'

interface Dinosaur {
  name: string;
  breed: string;
  teeth: number;
}

const myListOfDinosaurs = new GenericCollection<Dinosaur>();
const otherDino = {
  name: 'Blue',
  breed: 'Velociraptor',
  teeth: 100
}

myListOfDinosaurs.pushItem(otherDino);

myListOfDinosaurs.pushItem({name: 'Charlie'});
//error Argument type '{ name: string; }' is not assignable to parameter of type 'Dinosaur'.

In the example above, we are utilizing generics to inform GenericCollection what type it is receiving. string, number, and Dinosaur.

Objective 3: Recursive Generic Classes

A great example of the power of generics is creating a linked list with type safety. We will create a simple linked list that supports:

• Adding values to the front of the list with linkedList.unshift(value).
• Removing and returning the front values with linkedList.shift().
• Reading the front of the list with linkedList.head.
• Reading the end of the list with linkedList.tail.

We can use it with strings like:

const linkedList = new LinkedList<string>();

linkedList.unshift("a");
linkedList.unshift("b");

console.info( linkedList.shift() ) //logs "b"

console.info( linkedList.shift() ) //logs "a"

Or with numbers like:

const linkedList = new LinkedList<number>();

linkedList.unshift(100);
linkedList.unshift(200);

console.info( linkedList.head ) //logs 200
console.info( linkedList.tail ) //logs 100

The implementation looks like this:

// Define node that has a value and points to the
// next item in the list.
class LinkedListNode<T> {
    value: T;
    next?: LinkedListNode<T>;

    constructor(val: T) {
        this.value = val;
        this.next = null;
    }
}

class LinkedList<T> {
    private _head: LinkedListNode<T>;
    private _tail: LinkedListNode<T>;

    // Adds to the start of the list.
    unshift(value: T) {
        var node = new LinkedListNode(value);

        // The existing head is now next.
        if(this._head) {
            node.next = this._head;
        }

        this._head = node;

        // If there wasn’t a tail, this is the first node
        if(!this._tail) {
            this._tail = node;
        }
    }
    // removes first
    shift(){
        let value: T;

        // If there was a head,
        // set head to whatever is after it.
        if(this._head) {
            value = this._head.value;
            this._head = this._head.next;
        }

        // If there is no more head, the
        // list is empty.
        if(!this._head) {
            this._tail = null;
        }
        return value;
    }

    get head() { return this._head.value }
    get tail() {return this._tail.value }
}

Thanks to generics, we’re able to use the same LinkedList class in multiple different scenarios with any type.

Setup 2

✏️ Create src/generics/tree-node.ts and update it to be:

interface Comparison<T> {
  (v1: T, v2: T): number;
}

class TreeNode {
  value: any;
  compare: Comparison<any>;
  left?: TreeNode;
  right?: TreeNode;

  constructor(val, compare: Comparison<any>) {
    this.value = val;
    this.compare = compare;
  }

  add(val) {
    if (this.compare(this.value, val) >= 1) {
      if (this.left == null) {
        this.left = new TreeNode(val, this.compare);
      } else {
        this.left.add(val);
      }
    } else {
      if (this.right == null) {
        this.right = new TreeNode(val, this.compare);
      } else {
        this.right.add(val);
      }
    }
  }
}

export default TreeNode;

Verify 2

✏️ Create src/generics/tree-node.test.ts and update it to be:

import TreeNode from "./tree-node";
import { describe, it } from "node:test";
import { strict as assert } from "assert";

describe("Generics", () => {
  it("TreeNode can add numbers", () => {
    function numberComparison(v1: number, v2: number) {
      return v1 - v2;
    }

    let root = new TreeNode<number>(100, numberComparison);

    root.add(50);

    assert.equal(root.left?.value, 50, "50 to the left of 100");

    root.add(150);
    root.add(125);

    assert.equal(root.right?.value, 150, "150 to the right of 100");
    assert.equal(root.right?.left?.value, 125, "125 to the left of 150");
  });

  it("TreeNode can specify string", () => {
    function stringComparison(v1: string, v2: string): number {
      if (v1 > v2) {
        return 1;
      } else {
        return -1;
      }
    }

    let root = new TreeNode<string>("Jennifer", stringComparison);

    root.add("Chasen");

    assert.equal(root.left?.value, "Chasen", "Chasen to the left of Jennifer");

    root.add("Tom");
    root.add("Matthew");

    assert.equal(root.right?.value, "Tom", "Tom to the right of Jennifer");
    assert.equal(
      root.right?.left?.value,
      "Matthew",
      "Matthew to the left of Tom"
    );
  });
});

Run the following to verify your solution:

npm run test

Exercise 2

Update the tree-node.ts file to create a recursive TreeNode class that can house a value and be used to create a tree structure of left and right nodes.

For example, we will be able to create a TreeNode with a root value and comparison function as follows:

import TreeNode from "./tree-node";

function stringComparison(v1: string, v2: string): number {
  if(v1 > v2) {
      return 1;
  } else {
      return -1;
  }
};

const root = new TreeNode<string>("Jennifer", stringComparison);

Then we can add values to root like:

root.add("Chasen");

This will add Chasen to a left TreeNode of root because the stringComparison will return 1 (Jennifer > Chasen):

root.left.value //-> "Chasen"

As we add other values, they will be added to either the right or left nodes recursively:

root.add("Tom");
root.add("Matthew");

root.right.value      //-> "Tom"
root.right.left.value //-> "Matthew"

Have issues with your local setup? You can use either StackBlitz or CodeSandbox to do this exercise in an online code editor.

Solution 2

interface Comparison<T> {
  (v1: T, v2: T): number;
}

class TreeNode<T> {
  value: T;
  compare: Comparison<T>;
  left?: TreeNode<T>;
  right?: TreeNode<T>;

  constructor(val: T, compare: Comparison<T>) {
    this.value = val;
    this.compare = compare;
  }

  add(val: T) {
    if (this.compare(this.value, val) >= 1) {
      if (this.left == null) {
        this.left = new TreeNode(val, this.compare);
      } else {
        this.left.add(val);
      }
    } else {
      if (this.right == null) {
        this.right = new TreeNode(val, this.compare);
      } else {
        this.right.add(val);
      }
    }
  }
}

export default TreeNode;

Next steps

Next, let’s take a look at utility types for type transformations.