[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-openjfx-dev
Subject:    Re: Proposal: Deprecate Builders
From:       Tom Eugelink <tbee () tbee ! org>
Date:       2013-04-29 19:22:54
Message-ID: 517EC88E.4040003 () tbee ! org
[Download RAW message or body]


To get a wither based setup going, you need to add generics in order to have the \
correct return type. And we do not want to use these classes-with-generics as "end \
user" classes (leafs), but they construct the inheritance tree (branches). So IMHO \
the answer is no. But that does not mean I do not want to add that test, just in case \
/ to probe it will work.

Tom


On 2013-04-29 20:58, Richard Bair wrote:
> I think we also need to test when a middle guy is not abstract but also concrete -- \
> that is, for cases such as Parent where it is both concrete and also subclassed? 
> Richard
> 
> On Apr 6, 2013, at 11:17 AM, Tom Eugelink <tbee@tbee.org> wrote:
> 
> > I see the attachment got removed from the email, so here is a download link:
> > 
> > http://www.softworks.nl/stuff/withertest.zip
> > 
> > Tom
> > 
> > 
> > On 2013-04-06 09:14, Tom Eugelink wrote:
> > > On 2013-03-26 08:07, Richard Bair wrote:
> > > > The way to try out the compatibility issues is to just have two projects -- \
> > > > one with a minimal hierarchy (A and B is sufficient where B extends A). The \
> > > > other project just has a main method which creates an instance of B and \
> > > > initializes it. Then make various changes to the A / B API and try running \
> > > > the main project unmodified (i.e.: not recompiled) and see what happens! \
> > > > Richard
> > > 
> > > Ok, attached is exactly such a project. Project JavaApplication1 contains \
> > > LeafNode -> AbstractNodeA<T> -> AbstractNodeB<T>. Each one has \
> > > get/set/withName, where only one should be commented in. The setter prints the \
> > > class name where it is in so it is clear with setter is being used. LeafNode \
> > > also has a special method that is called on the return value of the wither, to \
> > > test it is LeafNode that is returned. 
> > > The second project JavaApplication2 holds only a main which:
> > > - creates LeafNode,
> > > - calls withName,
> > > - call the LeafNode method on the returned value
> > > 
> > > Just for test JavaApplication1 has exactly the same main.
> > > 
> > > In the directory root there is a test.cmd with sets the classpath and calls the \
> > > mains in both applications. You can now easily comment in/out the get/set/with \
> > > and only rebuild JavaApplication1. JavaApplication2 keeps on running without \
> > > problems. 
> > > I also have tried to see if the wither can be moved out to an interface (using \
> > > default) but that is giving me troubles. I need to figure out why it doesn't \
> > > accept the default body. 
> > > So far so good.
> > > 
> > > Tom
> > 


[prev in list] [next in list] [prev in thread] [next in thread]