The two key concepts you would want to know to handle errors in Observables are: catch and retry. As the name suggests catch allows us to catch the errors and retry will enable us to retry the action in case of an error.

Catch

Catch does not prevent the error from happening. It merely allows us to catch the error and do something with that error. Generally, we can wrap that error in an Observable so that the Observable chain can continue. We could also display that error to the end-user in the catch block while we continue the Observable chain.

Let us go through some examples to understand this better. I will take the example of concatenating two intervals from my previous blog post. This time I will add a third Observable that produces an error.

import Rx from 'rxjs';

const interval1 = Rx.Observable.interval(1000).map(i => `first: ${i}`).take(5);
const errorObservable = Rx.Observable.throw(new Error("An error occurred, cannot proceed"));
const interval2 = Rx.Observable.interval(500).map(i => `second: ${i}`).take(5);

const combinedInterval = Rx.Observable.concat(interval1, errorObservable, interval2);

combinedInterval.subscribe(
    data => console.log(`${data}`)
);

The output will be:

first: 0
first: 1
first: 2
first: 3
first: 4
Error: An error occurred, cannot proceed

The output is as expected. We got an error after the first Observable, so the second one never ran. The problem is, what if we still wanted to continue the concat chain despite the error? In this situation is where catch comes in. We will modify the example to use catch and display a message.

import Rx from 'rxjs';

const interval1 = Rx.Observable.interval(1000).map(i => `first: ${i}`).take(5);
const errorObservable = Rx.Observable.throw(new Error("An error occurred, cannot proceed")).catch(e => Rx.Observable.of(e.message));
const interval2 = Rx.Observable.interval(500).map(i => `second: ${i}`).take(5);

const combinedInterval = Rx.Observable.concat(interval1, errorObservable, interval2);

combinedInterval.subscribe(
    data => console.log(`${data}`)
);

Since now we have a catch statement for any error in the Observable, it returns a regular Observable with an error message. The output will be:

first: 0
first: 1
first: 2
first: 3
first: 4
An error occurred, cannot proceed
second: 0
second: 1
second: 2
second: 3
second: 4

We see the execution of all Observables despite the error in one of them.

P.S. The concat example might not be the best example to demonstrate catch. Don't lose hope! I use catch and retry both to explain retry in the next section.

Retry

retry method retries the Observable that generated the error. retry is helpful in case you are making an API call and you would like to keep retrying until you get success. Key things to remember about retry.

1. By default, it will retry infinite times.
2. It does take a numerical argument if you would like to limit the number of retries.
3. Do not use retry if you are converting a promise into an Observable (explained below).
4. retry truly unsubscribes from an error generating Observable and subscribes again.
5. A retry needs to execute the generator function of the Observable again. So, retry is only useful in case of cold Observables.
6. Hot Observable retry will not invoke the generator again, so it is useless.

When we resubscribe to a fromPromise, it still caches the resolve/reject status of the promise. It does not invoke the complete action again. This is why retry does not work with Observables made from promises.

To showcase an example of retry, I will create a function called dummyApi. The function will mimic an API call to the backend and return an error Observable. We will try and catch the error as well as retry the call.

Without catch or retry

import Rx from 'rxjs';

const dummyApi = () => {
    return new Rx.Observable(observer => {
        setTimeout(() => {
            observer.error(new Error("API call failed. Sorry!")); // API call responds with an error
        }, 1000); // API call takes 1 second to respond
    });
}

dummyApi()
    .do(() => console.log("Executing next Observable, chain continues"))
    .subscribe(
        data => console.log(data),
        error => console.log(error.message) // We handle error here by displaying the message
    )

The output will be:

API call failed. Sorry!

We logged the error message, but the problem is do operator in the chain never got executed. Now we use the catch operator (and this is a better example as I promise 😉).

Without retry (with catch)

import Rx from 'rxjs';

const dummyApi = () => {
  return new Rx.Observable(observer => {
    setTimeout(() => {
      observer.error(new Error("API call failed. Sorry!"))
    }, 1000);
  });
}

dummyApi()
  .catch(err => Rx.Observable.of(err.message)) // Wrap the error in a regular Observable so chain continues
  .do(() => console.log("Executing next Observable, chain continues")) // `do` operator logs the message
  .subscribe(
    data => console.log(data) // The error wrapped in a regular observable could not be logged
  )

The output will be:

Executing next Observable, chain continues
API call failed. Sorry!

Much better but we are still not retrying!

With retry and catch!

import Rx from 'rxjs';

const dummyApi = () => {
    return new Rx.Observable(observer => {
        console.log("Calling API"); // Added the log to display retry
        setTimeout(() => {
            observer.error(new Error("API call failed. Sorry!"))
        }, 1);
    });
}

dummyApi()
    .retry(3) // Retry 3 times
    .catch(err => Rx.Observable.of(err.message))
    .do(() => console.log("Executing next Observable, chain continues"))
    .subscribe(
        data => console.log(data)
    )

I added a console log statement to dummyApi so we can see retry attempts. The output will be:

Calling API
Calling API
Calling API
Calling API
Executing next Observable, chain continues
API call failed. Sorry!

The API gets called, it fails, and then it is retried three more times. That is why we see four logs with "Calling API" (original call plus three retries).

The above code handles retries, logs the error message if any, and continues the chain of Observable operators. Voila!

Happy coding 👋🏼