There is a big difference in the two snippets you've presented, e.g. when the catch
block itself throws an exception, the finally
block would still be executed by its semantics.
That is the following snippet prints "Finally!"
, but not "What about me???"
:
try {
throw null; // throws NullPointerException!
} catch (Exception e) {
int oops = 1/0; // throws ArithmeticException!
} finally {
System.out.println("Finally!"); // still gets executed!
}
System.out.println("What about me???"); // doesn't get executed!
Generally speaking, the finally
of a try
block practically always gets executed. There's no such guarantee for any code following the try
block.
But what if my catch
block is just a simple print
statement?
There's still no guarantee that it won't throw
something. Something could still go wrong in e.g. the construction for the exception detailed message.
Even if you make a best effort guarantee that the catch
code is "safe" and the code following the try
statement will always be executed, the question then becomes "Why?". Why avoid finally
but then try so hard to replicate its semantics?
finally
semantics is guaranteed, requiring no burden of proof from either the writer or the reader of the code. Precisely because of this, it's idiomatic to use finally
block to put mandatory "clean-up" code. Using finally
guarantees correctness and enhance both writability and readability.