Path: csiph.com!eeepc.pasdenom.info!news.pasdenom.info!news.dougwise.org!news-transit.tcx.org.uk!tr22g12.aset.psu.edu!news.glorb.com!news2.glorb.com!npeer01.iad.highwinds-media.com!news.highwinds-media.com!feed-me.highwinds-media.com!post01.iad.highwinds-media.com!newsfe13.iad.POSTED!00000000!not-for-mail From: Owen Jacobson Newsgroups: comp.lang.java.programmer Message-ID: <201102112200563996-angrybaldguy@gmailcom> References: <99453fde-6bb5-4b7d-9817-03ec1ffc7616@o13g2000yql.googlegroups.com> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1; format=flowed Content-Transfer-Encoding: 8bit Subject: Re: Default Interfaces: possible Java extension? User-Agent: Unison/2.1.3 Lines: 43 X-Complaints-To: abuse@UsenetServer.com NNTP-Posting-Date: Sat, 12 Feb 2011 03:00:56 UTC Date: Fri, 11 Feb 2011 22:00:56 -0500 Xref: csiph.com comp.lang.java.programmer:25527 On 2011-02-11 21:44:14 -0500, Eric Sosman said: > On 2/11/2011 9:20 PM, Tom McGlynn wrote: ... >> Add a new optional element in the definition of an interface which >> defines a class that does the default implementation of the >> interface. ... >> Are there other ways to do this? A factory method requires knowing >> the class the factory resides in and I can't really think of other >> patterns that address this. E.g., one could add >> newSet(), newList() and newMap() methods to Collections but that's not >> especially elegant to my eye since there's no special language >> relation between the List interface and the Collections class. > > I think that's the crux: There's no special relationship between > an interface and its many implementations; the relationship is one-way. > Adding the other-way relation -- the ability of an interface to name a > class and say "This is My beloved Son, in Whom I am well pleased" -- > doesn't seem to me to add much utility. This proposal also "bakes in" a circular dependency between an interface (List) and its default implementation (ArrayList), such that there is no way to compile or load either class without the other. While such circularity is sometimes hard to avoid (enums with per-constant bodies, for example, are inherently circular), they're not recommended style and they can lead to hard-to-debug classloading problems if you're not careful. I can't think of a use case for this outside of the collection types, either, and there are already idioms for those. This feels like a sublimated complaint about the standard library's choice of naming conventions (List & ArrayList, rather than IList and List as with .Net or informal protocols and list() like Python). Interesting proposal, but I think it contains unfixable flaws. -o