> >> So go ahead with the change, removing the throws from the declaration but > >> keeping the javadoc as suggested previously. > > > > Again, what is it that you try to convey by specifying a single exception in > > the Javadoc? Any unchecked exception can be thrown from a class that > > implement the interface. > > If the user code doesn't care that the evaluation fails, it should catch > > everything and continue. Alternatively, if it cannot continue, it should let > > the exception propagate. In either case, there isn't any useful information > > from a Javadoc "@throws" tag: We already know that unchecked exceptions can > > arise. > > Seems to me we should document any Exceptions that our code throws, > along with the reasons for them. > That way, the user can create specific catch clauses.
No disagreement about this principle. But the "@throws" Javadoc comment should be placed where we _know_ that the code throws the exception. In "UnivariateRealFunction" it is impossible to know which exception is thrown and which not. > Also, the other RuntimeExceptions are generally unpredictable, whereas > in this case, these Exceptions are only > generated under specific circumstances. Which exceptions? Regards, Gilles --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
