Vladimir, Il giorno mer 13 feb 2019 alle ore 21:49 Vladimir Sitnikov <sitnikov.vladi...@gmail.com> ha scritto: > > Hi, > > I've recently discovered that new RuntimeException(Throwable) results in a > duplicated error messages. > In fact Throwable#<init>(Throwable) just calls cause.getMessage() and uses > it as its own message. > > This makes exceptions harder to follow since the new RuntimeException(e) is > very often used just to overcome "checked exception" rule in Java. > > I've recently committed a deduplicate PR to Avatica: > https://github.com/apache/calcite-avatica/pull/84 > As you can see, it removed "RuntimeException: ..." nonsense and it made > exceptions easier to understand. For instance, see > https://github.com/apache/calcite-avatica/pull/84/files#diff-d499001f31481c5f7151a81380c01a60L332 > > I think we should avoid new RuntimeException(Throwable), and I wonder if we > should add that as a forbiddenapis rule. > > You can find a PR for *test* code: > https://github.com/apache/calcite/pull/1042 > > I'm not sure if I should just go ahead and dodge new > RuntimeException(Throwable) from the main codebase as well. > Do you have any negative experience with sneaky-throws pattern? > I'm inclined to do so provided no-one objects within a week.
I think sneaky-throws works well in your case, I mean to replace new RuntimeException(Throwable). The only problem is that users will find checked exception thrown in places where they aren't supposed to be thrown, anyway using RuntimeException is weird as well. So +1 from me just my 2 cents Enrico > > Vladimir
