Comment by vbezhenar
2 days ago
> This provides no automatic verification that indeed all likely error situation that can and should be handled were indeed handled.
And some people write code in Python which provides no automatic verification whatsoever.
Actually unchecked exceptions are very similar to dynamically typed languages. And that's fine. As Python and other languages proved by their mere existence: dynamic typing is not inherently bad. Sometimes it's good. I think that for error handling, it's good.
> Java class loaders can do anything including loading resources from the network. Which is admittedly not that common these days after the demise of applets.
Technically they can, but in my particular case I know very well, that my resources are residing inside of JAR file. And if that JAR file happened to reside on the failed HDD block, that's not going to be a recoverable error.
When we're talking about IO Exceptions, it's almost always failed operation which requires complete abort. It's either failed hardware, or, typically, disconnected client. Can't do much about it, other than clean up and proceed to the next client.
And the same could be said about SQL Exceptions. Like 99% of SQL exceptions are application bugs which are not going to be handled in any way other than wind up and return high level HTTP 500 error or something like that. There are cases when SQL exception should be handled, but those are rare. Yet JDBC developers decided that programmers must execute error handling rituals on every call site.
> The NullPointerException indicates a bug in the application. By the very reason it occurs, the current thread cannot continue execution normally.
That's not exactly true. In JVM, NullPointerException is absolutely well defined and you can continue execution after catching it. You might suspect, that logical application state is corrupted, but sometimes you know well that everything's fine (and in most cases you hope that everything's fine, if your Spring MVC handler threw NPE, Spring is not going to crash, it'll continue to serve requests). It's not C++ with its undefined stuff. JVM is pretty reliable when it comes to every error, including NPE, stack overflow or OOM. Latter is special, because even handling error might prove challenging, when memory allocations fail, but JVM will not hang up or crash.
> And some people write code in Python which provides no automatic verification whatsoever.
Python is a language with almost no static validation whatsoever. It would be very odd if it cared about checked exceptions. This dynamism makes big Python code bases infuriating to work with.
> When we're talking about IO Exceptions, it's almost always failed operation which requires complete abort. It's either failed hardware, or, typically, disconnected client. Can't do much about it, other than clean up and proceed to the next client.
If this is the case then the solution is to add it to the `throws` list.
> That's not exactly true. In JVM, NullPointerException is absolutely well defined and you can continue execution after catching it.
Why would I catch a NullPointerException instead of fixing my application? The JVM is indeed completely fine, but processing still cannot continue because that code simply does not exist.