Hello, 2012/9/17 Gilles Sadowski <gil...@harfang.homelinux.org>: > On Fri, Sep 14, 2012 at 11:29:41AM -0700, Phil Steitz wrote: >> OK, I give up. Lets do option 2. Just warn users in the User Guide >> somewhere that our APIs are in general not null-safe and unless the >> javadoc explicitly allows nulls, they can expect NPE when passing nulls. > > Thanks, Phil; we are making progress, and I hope that you'll be convinced of > that at some point. > > Another decision still needs to be made. > > I think that everybody now agrees that wherever an argument will be directly > used (i.e. "dereferenced") inside the body of the method[1], it is safe to > not check for null (since the JVM will throw an NPE). > > But, whenever an argument is passed for _later_ use (e.g. stored by a > constructor or passed to another method), we also all expressed that it is > better to fail early if the object is supposed to be non-null. In those > cases, checks are not mandatory (since the JVM will throw NPE at some > point), but we must agree on how optional checks are to be implemented. > 1. The current state of affairs was to throw "NullArgumentException"; in 4.0 > this exception will become a subclass of NPE and we can continue to use > it in the same way as now (i.e. possibly providing additional localizable > error messages). > 2. The alternative is to directly throw a standard NPE, but without any > message, since it wouldn't be localizable with the support of CM's > "context". > > So, which of those alternatives do people want to settle on? > > > Regards, > Gilles > > [1] As opposed to being passed on to another method. > I have another question: are we going to little by little remove all null checks that we feel are not absolutely necessary? I would not mind going through this cleaning phase while working on MATH-854.
Best regards, Sébastien --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
