Question

I am doing little hobby project in C#, a language I do not know well, and have stumbled upon the following:

Suppose you have an asynchronous operation implemented by using BackgroundWorker. Now if there is an exception, event RunWorkerCompleted will be raised and RunWorkerCompletedEventArgs.Error will be non-null.

Is the following the canonical way then to handle different exception types? (Here all the exception kinds are siblings WRT inheritance)

if (e.Error != null)
{
    FirstKindOfException e1 = e as OneKindOfException;
    SecondKindOfException e2 = e as SecondKindOfException;
    ...
    LastKindOfException en = e as LastKindOfException;
    if (e1 != null)
    {
        ...
    }
    else if (e2 != null)
    {
        ...
    }
    ...
    else
    {
        ...
    }
}

It works, but… it doesn’t feel right.

Answer 1

You could use is to keep each test tightly scoped:

if (e.Error is FirstKindOfException )
{
    ...
}
else if (e.Error is SecondKindOfException)
{
    ...
}

(then re-cast if you want special values from the exception)

To be honest, though, it is pretty rare that I need to handle lots of different types of exceptions. In most cases, it is fine to simply recover (compensate) to a known state and report the error appropriately. Generally I prefer to test for likely errors before I start the action, so an exception truly is something exceptional.