I did in the original post, and that's why I used C++ as example instead of Java. A method declares `throws X, Y, Z` and _runtime_ checks that no other exception escapes. No source changes needed if you add W to the list. And if some other exception escapes, it's wrapped in `UnexpectedException` that is reserved for and throwable only by the runtime.
I guess we could extend Project Lombok to do this. You'll still have your "throws" statement but then the tool would wrap calls to catch those that are not listed in "throws" statement and wrap them as you say.
It just occurred to me that I could probably do the same with attributes and DynamicProxy for C#. And also implement additional checks like "IF X is thrown, its properties must satisfiy some constraints.". (I program both in Java and C# these days.)
Such "UnexpectedException" as I suggested would serve two purposes: 1) well, knowing that something unexpected happened and allowing you to handle it with "last chance handler", 2) helping the developers maintain the contract. If you change a method so that it can throw some new exceptions (compared to the previous version), you've broken its contract/compatibility. This would then show up during testing.
> You'll still have your "throws" statement but then the tool would wrap calls t
Yes. Fortunately, Java allows you to mention subclasses of RuntimeException in "throws" declaration.
But... both purposes 1) and 2) are already served perfectly fine with the current exception models already: you catch Exception at the very top-level in the "last chance handler", and if during testing an unexpected exception is thrown, it bubbles up past the existing handlers right into this "last chance handler". What does re-wrapping help with, exactly?
Unless the method 1) throws an exception which it should not have according to its declarative contract (annotations in Java, attributes in C#), and 2) it gets (erroneously) handled by an intermediate handler.
> What does re-wrapping help with, exactly?
It makes it clear that the method broke its declarative contract, which is what exceptions are _for_. And given the restriction that only the runtime can throw such exceptions, you're sure that no other method can randomly throw such exceptions because... they like it so.
Well, "(erroneously) handled by an intermediate handler" is a tricky situation: would it be really handled incorrectly?
Another question is ergonomics. It's trivial (but tedious) to write code like this:
class Foo {
// ...
public void Frob(...) throws FooException {
try {
// ...
}
catch (Exception e) {
throw new FooException(e);
}
}
public void Blarg(...) throws FooException {
try {
// ...
}
catch (Exception e) {
throw new FooException(e);
}
}
}
which is actually a "best practice" already ("annotate inner exceptions with some high-level context and wrap them in high-level exceptions") — and adhering to it makes your proposition completely extraneous, because nothing can throw an UnknownException ever.
And in before "don't catch and wrap Exception!", consider that Foo maybe parameterized by some dependency that may be implemented as a network service, or a disk file, or a DB: three different implementations will throw completely different exceptions: FileNotFound vs NetworkConnectionClosed vs OdbcInvalidManufaturer. Either dependency interface allows implementations to throw any of those (so that the user of Foo, who knows which implementation it specified, can handle those), or it makes them to wrap them all into DepException, but then, again, this means that Foo's methods will catch-wrap-rethrow DepExceptions instead of just Exceptions.
If previously only DoB could throw SomethingEx and now DoA() can also throw it, the handler is almost certainly incorrect handling if this is some "generic" exception. C#/.NET base library is notorious for using InvalidOperationException for all kinds of unrelated crap.
> It's trivial (but tedious) to write code like this
Eh. Methods without "throws" would be "unchecked" at runtime as well. So my (imagined) best practice would be that non-private methods declare their exceptions and leave it to the caller whether and how to wrap them.
> And in before "don't catch and wrap Exception!"
Oh, I do that all the time for precisely the reasons you mentioned. Exceptions from the lowest level are most precise and least useful as there's no information about the context. (Unless you go down the unmaintainable rabbit hole of parsing the call stack.)
So, if I understand your proposal correctly, you want to split the current 2-pronged approach "catch everything, handle what you can, re-throw what you can't wrapped in FooException" into a 3-pronged one: "catch what you believe can be handled by you or your users, handle what you can, re-throw what you can't wrapped in FooException, and the system will re-throw everything else wrapped in UnexpectedException". The difference is that if now some dependency of Foo would change drastically, the Foo's users won't be able to accidentally swallow or even see the new exceptions, those will go all the way up as UnexpectedExceptions and will draw the due attention.
> if now some dependency of Foo would change drastically
The most frequent "drastic change" being writing new code and fixing bugs. It's extremely easy to widen the set of exceptions thrown by a method and zero tooling to help you with finding out what exceptions can be thrown from the code.
Say what you want about Java, but its division of throwables into errors and exceptions makes sense. Errors like stack overflow, VM faults, etc., should not be exceptions. Under this scheme, they would always propagate out of the method unwrapped. Again, .NETs predefined exception types are botched beyond repair.
