[stringtemplate-interest] StringTemplate + Doclet

[Terence Parr]

Ah. right! Class.isAssignableFrom is the ticket.

@Sam, caching. you mean recording each o.getClass() -> isAssignableFrom mapping?

Seems I should upgrade regiserRenderer too...grr...

Ter
On Dec 22, 2010, at 10:47 AM, Rafael Chaves wrote:

> For any interface based API (which is a common practice), getClass()
> is a show stopper. A client of the API does not know the actual
> classes involved (they are not API).
> 
> I was not suggesting actually using the instanceof operator (I am not
> sure you really meant that, as it is not really possible), but
> Class.isInstanceOf(object). A similar thing would be
> Class.isAssignableFrom(object.getClass()).
> 
> Cheers,
> 
> Rafael
> 
> On Wed, Dec 22, 2010 at 10:06 AM, Terence Parr <parrt at cs.usfca.edu> wrote:
>> Seems like I could do instanceof by just walking the Class.superClass/implements chain, right?  I'm not considering generating code; makes it hard to port ;)
>> Ter
>> On Dec 22, 2010, at 10:01 AM, Brian Clapper wrote:
>> 
>>> On 12/22/10 12:51 PM, Terence Parr wrote:
>>>> 
>>>> On Dec 22, 2010, at 9:38 AM, Brian Clapper wrote:
>>>>>> contract), I ended up using ASM to generate interfaces on the fly, which I then
>>>>> implement via java.lang.reflect.Proxy.
>>>> 
>>>> BTW, for renderers i did == on obj.getClass() not instanceof (which only works
>>>> on class names not Class ptrs).  Is it a serious problem if i don't allow
>>>> instanceof?  Maybe if the model object is hidden by interfaces like jdbc.
>>>> hmm... i'd rather avoid the java.lang.reflect.Proxy stuff.
>>> 
>>> I don't use instanceof, so it won't affect me.
>>> 
>>> I deliberately chose java.lang.reflect.Proxy, for a couple reasons. First,
>>> generating the byte code for an interface, via ASM, is straightforward and
>>> simple. Generating a wrapper class, on the other hand, is more complicated and
>>> annoying. Second, using Proxy to implement an interface on the fly is also
>>> straightforward.
>>> 
>>> This technique allows for more maintainable code, in my opinion. If I'm going
>>> to resort to generating byte code on the fly, I'd like that part of the
>>> solution to be a dead-simple as possible. Use of Proxy made that possible. If I
>>> weren't generating the byte code for the wrapper interface on the fly, I
>>> wouldn't be using Proxy at all.
>>> 
>>> If I could get rid of byte code generation entirely, it'd be a real win.
>>> --
>>> -Brian
>>> 
>>> Brian Clapper, http://www.clapper.org/bmc/
>>> _______________________________________________
>>> stringtemplate-interest mailing list
>>> stringtemplate-interest at antlr.org
>>> http://www.antlr.org/mailman/listinfo/stringtemplate-interest
>> 
>> _______________________________________________
>> stringtemplate-interest mailing list
>> stringtemplate-interest at antlr.org
>> http://www.antlr.org/mailman/listinfo/stringtemplate-interest
>> 
> _______________________________________________
> stringtemplate-interest mailing list
> stringtemplate-interest at antlr.org
> http://www.antlr.org/mailman/listinfo/stringtemplate-interest