How do I handle exceptions in map() in an Observable in RxJava

Question

I want to do this:

Observable.just(bitmap)
            .map(new Func1<Bitmap, File>() {
                @Override
                public File call(Bitmap photoBitmap) {

                    //File creation throws IOException, 
                    //I just want it to hit the onError() inside subscribe()

                    File photoFile = new File(App.getAppContext().getCacheDir(), "userprofilepic_temp.jpg");
                    if(photoFile.isFile()) {//delete the file first if it exists otherwise the new file won't be created
                        photoFile.delete();
                    }
                    photoFile.createNewFile(); //saves the file in the cache dir

                    FileOutputStream fos = new FileOutputStream(photoFile);
                    photoBitmap.compress(Bitmap.CompressFormat.JPEG, 90, fos);//jpeg format
                    fos.close();

                    return photoFile;

                }
            })
            .subscribe(//continue implementation...);

Basically in the call() method, it can throw exceptions. How can I make the Observer handle it in onError(). Or is this not the right way to think about this?

Note that in RxJava 2 operators like `map` allow throwing of checked exceptions from the lambda. This was really a design flaw in RxJava 1 because the exact error thrown couldn't be propagated in 'onError` from the map lambda without being wrapped as a `RuntimeException`. — Dave Moten, Dec 01 '16 at 20:23

score 24 · Accepted Answer · edited Dec 01 '16 at 20:20

24

rx will always catch error, even if this is RuntimeException. So you can throw Some kind of Runtime exception in catch block. This is how actually it should work.

 Observable.just(bitmap)
                .map(b -> {
                    try {
                        // do some work which throws IOException
                        throw new IOException("something went wrong");
                    } catch (IOException e) {
                        throw new RXIOException(e);
                        // Or you can use 
                        throw Exceptions.propagate(e);
                        // This helper method will wrap your exception with runtime one
                    }
                }).subscribe(o -> {
                    // do something here
                }, exception -> exception.printStackTrace());

public static class RXIOException extends RuntimeException {
        public RXIOException(IOException throwable) {
            super(throwable);
        }
}

edited Dec 01 '16 at 20:20

Dave Moten

11,957
2
40
47

answered Nov 06 '15 at 21:44

wnc_21

1,751
13
20

1

when I try to do this, it asks me to surround the exception with a try-catch – Sree Nov 06 '15 at 23:27
2

You have to throw RuntimeException. JVM will not ask you to surround it with try catch. That's the idea – wnc_21 Nov 07 '15 at 20:09
I made `call()` throw an IOException but it is saying overridden method doesn't throw IOException – Sree Nov 09 '15 at 16:54
3

Come on. IOException is not runtime exception. Read my answer one more time. – wnc_21 Nov 09 '15 at 19:36
ah, sorry. makes total sense. can you look at @akarnokd 's answer above which uses `Observable.error`. Would you recommend your approach over that? – Sree Nov 09 '15 at 19:55
It depends on, what are you going to do. Observable.error() returns new Observable and will work only with flatMap If you want to have mObservable.map(something can trhow an error).filter(exception here).subscribe(...) in this case, it is not possible to use Observable.error() – wnc_21 Nov 09 '15 at 20:05
last question: so if I needed to handle an exception, I need to write a wrapper like that for each type of exception? – Sree Nov 09 '15 at 20:56
Only if you need it. You can throw RuntimeException or any class inherited from it. Wrapper adds additional sense. For example map() throws IOException and FileNotFondException and you need to handle them separately. In this case you need to have wrappers, to distinguish them in onError() – wnc_21 Nov 09 '15 at 21:25

score 8 · Answer 2 · edited Jun 05 '17 at 20:48

8

With 1.0.15, there is the fromCallable factory method which let's you run a Callable instance for each subscriber where you can throw checked exceptions as well:

Observable.fromCallable(() -> {      
    File photoFile = new File(App.getAppContext().getCacheDir(),
        "userprofilepic_temp.jpg");
    if (photoFile.isFile()) {
       //delete the file if it exists otherwise the new file won't be created
        photoFile.delete();
    }
    photoFile.createNewFile(); //saves the file in the cache dir

    FileOutputStream fos = new FileOutputStream(photoFile);
    photoBitmap.compress(Bitmap.CompressFormat.JPEG, 90, fos);//jpeg format
    fos.close();

    return photoFile;
})
.subscribe(...)

Edit:

source.flatMap(v -> {
    try {
        //...
        return Observable.just(result);
    } catch (Exception e) {
        return Observable.error(e);
    }
})
.subscribe(...);

edited Jun 05 '17 at 20:48

falvojr

3,060
4
31
54

answered Nov 05 '15 at 22:42

akarnokd

69,132
14
157
192

I guess this would work for this case, but what if I want to chain two maps together that rely on the output of the first map. Using `Func1()` I can get the return type and argument. – Sree Nov 05 '15 at 22:57
You have to use flatMap, try-catch and return either just() or error(). – akarnokd Nov 06 '15 at 08:01
Works like a champ!! – Raja Jawahar Apr 12 '17 at 15:58

score 3 · Answer 3 · answered Dec 01 '16 at 17:05

Just created helper class to extract this boilerplate to another place:

public class RxRethrow {
    public static <T, R> Func1<T, R> rethrow(Func1R<T, R> catchedFunc) {
        return t -> {
            try {
                return catchedFunc.call(t);
            } catch (Exception e) {
                throw Exceptions.propagate(e);
            }
        };
    }

    public interface Func1R<T, R> extends Function {
        R call(T t) throws Exception;
    }
}

You can call it like this:

.map(RxRethrow.rethrow(products -> mapper.writer(schema).writeValueAsString(products)))

score 3 · Answer 4 · answered Feb 03 '17 at 15:45

3

I don't know how the situation was when this question was first asked and answered, but RxJava currently contains a helper-method for this exact purpose: Exceptions.propagate(Throwable t)

RxJava Javadoc

Convenience method to throw a RuntimeException and Error directly or wrap any other exception type into a RuntimeException.

answered Feb 03 '17 at 15:45

Thorbear

2,303
28
23

@Sree Wow, I actually didn't spot that, I was blinded by the custom exception class. Now that I look closer I can see that there is actually another answer here using the same method. – Thorbear Feb 04 '17 at 10:18

How do I handle exceptions in map() in an Observable in RxJava

4 Answers4

Linked