Also, Bob raised the question of generators and their sensitivity to
user.agent values (or other properties, perhaps), and whether we would need
GeneratorContext.propertyValueMatches(String actual, String
possiblyInherited) or the like.

I was focused on rule selection, so I hadn't thought about it, but I'm not
sure it would be needed: the generator is, after all, still generating for
safari, and not doing anything special for iphone or android; the
inheritance is just asserting that the generated code is also applicable for
iphone.  When we do an <any><when-property-is.../></any> today, as I
understand it, the generator does get a chance to do different work for each
user agent, but may also return the same classname to do the same job once
for all... my value-inheritance would effectively force the second case.  If
you want something different, you want a specific rule, not the implicit
inheritance.  No?




On Tue, May 5, 2009 at 10:16 AM, Freeland Abbott <[email protected]> wrote:

> Before I started working up a patch for this, I thought it was worth
> floating the basic idea for comment:
>
> I recently wanted an app with special interfaces for mobile browsers
> (specificially, iphone and android).  No problem; we redefined user.agent to
> allow those values, and to provide a property provider for them.  However,
> it turns out that both are WebKit, like Safari, and in a lot of cases, all
> the Safari rebindings were correct for the mobile ones, too.  So I wanted a
> way to express that iphone was "just like Safari, unless explicitly noted
> otherwise."
>
> What would people think about a GWT XML directive like:
>
> <property-value-inherits property="user.agent" value="iphone"
> inherits="safari"/>
>
> ...with all of the field names up for debate, as well as the basic
> concept.  The same mechanism could be used to specialize user agents by
> version, i.e. ie-base could be inherited by ie6 and ie8.
>
> Other, similar concepts I've heard discussed include a
> <when-property-begins-with .../> or <when-property-matches .../>
> formulation, but I'm not sure either of them let me do the notion of
> "inheritance" that I wanted.  (They do cover the version number problem,
> however.)
>
>

--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---

Reply via email to