RxJS custom operators

Undoubtedly, it’s a wide array of operators that makes the RxJS library extremely powerful utility in a developer’s tool belt. I’ve recently written several custom operators in order to reuse some repetitive combinations of operators in a nifty way. In this blog post, I will introduce you to the concept of custom RxJS operators and present exemplary implementations.

Identity operator

Link to this section

An RxJS operator is simply a function which takes a source observable as an input and returns a resulting stream. Therefore, the task of creating a custom RxJS operator comes down to writing a regular JavaScript (TypeScript) function. Let’s start with a basic identity operator which simply mirrors a source observable:

<>Copy
import { interval, Observable } from "rxjs";
import { take } from "rxjs/operators";

const source$ = interval(1000).pipe(take(3));

function identity<T>(source$: Observable<T>): Observable<T> {
  return source$;
}

const results$ = source$.pipe(identity);

results$.subscribe(console.log);
  
// console output: 0, 1, 2

In the next step, let’s add some basic logic to a custom operator.

Logging operator

Link to this section

The following custom operator performs a side effect (logs values to the console) for each value of a source stream:

<>Copy
import { interval, Observable } from "rxjs";
import { take, tap } from "rxjs/operators";

const source$ = interval(1000).pipe(take(3));

function log<T>(source$: Observable<T>): Observable<T> {
  return source$.pipe(tap(v => console.log(`log: ${v}`)));
}

const results$ = source$.pipe(log);

results$.subscribe(console.log);
  
// console output: log: 0, log: 1, log: 2

The resulting stream is composed on the basis of the source$ and altered by applying built-in operators via the pipe method.

Operator’s factory

Link to this section

In certain scenarios, it’s convenient to provide a context for a custom operator. You can accomplish it by defining a function which returns an operator. The factory’s arguments belong to the operator’s lexical scope:

<>Copy
import { interval, Observable } from "rxjs";
import { take, tap } from "rxjs/operators";

const source$ = interval(1000).pipe(take(3));

function logWithTag<T>(tag: string): (source$: Observable<T>) => Observable<T> {
  return source$ =>
    source$.pipe(tap(v => console.log(`logWithTag(${tag}): ${v}`)));
}

const results$ = source$.pipe(logWithTag("RxJS"));

results$.subscribe(console.log);
  
// console output: logWithTag(RxJS): 0, logWithTag(RxJS): 1, logWithTag(RxJS): 2

You can simplify the return type annotation by making use of the MonoTypeOperatorFunction provided by the RxJS. In addition, the static pipe function enables to define an operator in a less verbose way:

<>Copy
import { interval, MonoTypeOperatorFunction, pipe } from "rxjs";
import { take, tap } from "rxjs/operators";

const source$ = interval(1000).pipe(take(3));

function logWithTag<T>(tag: string): MonoTypeOperatorFunction<T> {
  return pipe(tap(v => console.log(`logWithTag(${tag}): ${v}`)));
}

const results$ = source$.pipe(logWithTag("RxJS"));

results$.subscribe(console.log);
  
// console output: logWithTag(RxJS): 0, logWithTag(RxJS): 1, logWithTag(RxJS): 2

If you seek for more RxJS-related tips, check this out.

Observer-unique lexical scope

Link to this section

An operator’s factory function is called once only at the moment of a stream definition. As a result, there’s a shared lexical scope for all observers:

<>Copy
import { interval, MonoTypeOperatorFunction, pipe } from "rxjs";
import { take, tap } from "rxjs/operators";

const source$ = interval(1000).pipe(take(3));

function tapOnce<T>(job: Function): MonoTypeOperatorFunction<T> {
  let isFirst = true;

  return pipe(
    tap(v => {
      if (!isFirst) {
        return;
      }

      job(v);
      isFirst = false;
    })
  );
}

const results$ = source$.pipe(tapOnce(() => console.log("First value emitted")));

results$.subscribe(console.log);
results$.subscribe(console.log);
  
// console output: First value emitted, 0, 0, 1, 1, 2, 2

If you want to get a unique lexical scope for each observer, you can make use of the defer function:

<>Copy
import { defer, interval, MonoTypeOperatorFunction } from "rxjs";
import { take, tap } from "rxjs/operators";

const source$ = interval(1000).pipe(take(3));

function tapOnceUnique<T>(job: Function): MonoTypeOperatorFunction<T> {
  return source$ =>
    defer(() => {
      let isFirst = true;

      return source$.pipe(
        tap(v => {
          if (!isFirst) {
            return;
          }

          job(v);
          isFirst = false;
        })
      );
    });
}

const results$ = source$.pipe(tapOnceUnique(() => console.log("First value emitted")));

results$.subscribe(console.log);
results$.subscribe(console.log);
  
// console output: First value emitted, 0, First value emitted, 0, 1, 1, 2, 2

If you want to get to know how to solve the tapOnce task in a different way, take a look at one of my blog posts.

Real-life examples

Link to this section

firstTruthy operator:

<>Copy
import { MonoTypeOperatorFunction, of, pipe } from "rxjs";
import { first } from "rxjs/operators";

const source1$ = of(0, "", "foo", 69);

function firstTruthy<T>(): MonoTypeOperatorFunction<T> {
  return pipe(first(v => Boolean(v)));
}

const result1$ = source1$.pipe(firstTruthy());

result1$.subscribe(console.log);

// console output: foo

evenMultiplied operator:

<>Copy
import { interval, MonoTypeOperatorFunction, pipe } from "rxjs";
import { filter, map, take } from "rxjs/operators";

const source2$ = interval(10).pipe(take(3));

function evenMultiplied(multiplier: number): MonoTypeOperatorFunction<number> {
  return pipe(
    filter(v => v % 2 === 0),
    map(v => v * multiplier)
  );
}

const result2$ = source2$.pipe(evenMultiplied(3));

result2$.subscribe(console.log);
  
// console output: 0, 6

liveSearch operator:

<>Copy
import { ObservableInput, of, OperatorFunction, pipe  } from "rxjs";
import { debounceTime, delay, distinctUntilChanged, switchMap } from "rxjs/operators";

const source3$ = of("politics", "sport");

type DataProducer<T> = (q: string) => ObservableInput<T>;

function liveSearch<R>(
  time: number,
  dataProducer: DataProducer<R>
): OperatorFunction<string, R> {
  return pipe(
    debounceTime(time),
    distinctUntilChanged(),
    switchMap(dataProducer)
  );
}

const newsProducer = (q: string) =>
  of(`Data fetched for ${q}`).pipe(delay(2000));

const result3$ = source3$.pipe(liveSearch(500, newsProducer));

result3$.subscribe(console.log);
  
// console output: Data fetched for sport

Conclusions

Link to this section

Common RxJS operators’ combinations can be extracted into custom operators and reused in the future when implementing similar features. Generics enables proper type inference for emitted values further down in the pipe sequence.

Live example:

I hope you liked the post and learned something new.