On Sun, Nov 22, 2009 at 4:36 PM, Stephen Colebourne <scolebou...@btopenworld.com> wrote: > Henri Yandell wrote: >> >> On Thu, Nov 19, 2009 at 9:24 AM, Paul Benedict <pbened...@apache.org> >> wrote: >>> >>> On Thu, Nov 19, 2009 at 1:39 AM, Henri Yandell <flame...@gmail.com> >>> wrote: >>> >>>> Your argument being that the JDK and others do NPE when null? i.e. NPE >>>> = IAE(null)? >>> >>> Yes. Some things to note: >>> * The JDK does this as their pattern >>> * JDK 7 definitely encodes this patter in their new Object.notNull() >>> method. >>> * Google Collections -- being a JDK wannabe addition -- does this >>> * Josuha Bloch's "Effective Java" (two editions) has codified this >>> expectation into the minds of developers. >> >> +1 to NPE. > > While I know this is the modern advice, there will be other places where any > NPE is a bug, and thus Validate should throw IAE.
Other places in the JDK, or in some other library, or that there are places that choose the non-'modern advice'? > There are no pretty solutions > - two classes ValidateNPE and ValidateIAE? I don't think Validate is that important a class to warrant two impls. > - some kind of static flag? Seems brittle. Though a -D property could be doable, that implies singleton-for-jvm better than just a static flag where people will want to play funny games with classloaders. Still feels bad though as wouldn't it mean the API throwing Exception? > - picking one, and offending the others? +1 :) I've got better things to do in life than burn time supporting alternative opinions. Unless the JDK is wrong, JDK wins imo. Hen --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
