[Laszlo-dev] tagnames

[André Bargull]

Simply invert the (debug-)invariant to distinguish those classes:
   lz[this.constructor.tagname] !== this.constructor
If this condition is true, the current object is an anonymous subclass. 
Or for more fail-safe code:
   lz[this.constructor.tagname] !== this.constructor
     && this is lz[this.constructor.tagname]


On 8/6/2008 1:01 PM, P T Withington wrote:
> You know my code better than I do!  Ok, let's remove the munging of the 
> tagName, although we need to make sure that in debug mode you can 
> distinguish the true button class from one of the anonymous button 
> subclasses.
> 
> On 2008-08-05, at 20:03EDT, André Bargull wrote:
> 
>>
>> On 8/5/2008 10:04 PM, P T Withington wrote:
>>> I see what you are getting at, but doing so would break this invariant:
>>>  lz[this.constructor.tagname] == this.constructor
>>
>> This is only true in debug-mode!
>>
>> From NodeModel.java:
>>> // If there are methods, make a class
>>> if (debug) {
>>>  String name = id;
>>>  if (name == null) {
>>>    name = CompilerUtils.attributeUniqueName(element, "class");
>>>  }       // Update tagname to our custom class
>>>  tagName = tagName + "_" + name;
>>> } else {
>>>  *// Don't create a tag for anonymous classes
>>>  tagName = null;*
>>> }
>>
>>
>> And a small testcase:
>>> <canvas debug="false" >
>>>  <button onclick="txt.setText(this.constructor.tagname)" />
>>>  <text id="txt" x="50" />
>>> </canvas>
>>
>>
>>> which is presently true for all classes.  We would have to say that 
>>> for 'anonymous' classes that does not hold.  That's probably ok, but 
>>> I expect that is why it is currently the way it is (at some point in 
>>> the evolution, the anonymous classes were instantiated with a tagname 
>>> which had to be unique; now that they are instantiated directly with 
>>> their classname, we could avoid creating the unique tagnames).
>>> Perhaps there should be some amendment to the debugger also to print 
>>> these anonymous classes in a prettier fashion too.
>>> On 2008-08-05, at 13:43EDT, André Bargull wrote:
>>>> An user asked me on the forum 
>>>> (http://forum.openlaszlo.org/showthread.php?t=12198):
>>>>
>>>>> what i want to ask is:
>>>>> does OL support the original "tagname" in customized instances 
>>>>> (constraint and method filled in tag )?
>>>>> or
>>>>> it is a defect?
>>>>
>>>> So, whenever an anonymous class was created (because of constraints, 
>>>> handlers etc), do we provide a way to retrieve the original tagname?
>>>> In my mind, a reasonable question, names like 
>>>> "/button_$$2Etmp_C697C77B63C2714AA45A51E9F3163AFA_hello$2Elzx_2_32/" 
>>>> look ugly and do confuse people (at least me when I see those names 
>>>> in the debugger :-) ).
>>>>
>>>>
>>>> ---
>>>> And I don't want to get this answer: ;-)
>>>>> /foo/.constructor.prototype.constructor.tagname
>>>>
>>
> 
>