[Htmlparser-developer] Re: Adding methods to Tag

[Derrick O. <Der...@ro...>]

Chris,

I'm opening this up to a wider audience, because it may have been solved 
before, or might be of interest to others with the same problem.

The basic problem is how to add functionality like supportsColor() to 
base classes, like Tag, without recompiling the whole class heirarchy.

One way would be to join the htmlparser project as a developer and just 
add it, if it's germane to others besides yourself. If it's not, then a 
bolt-on is needed.

One way to handle this problem is a 'Factory' mechanism.  A 
'deep-in-the-bowels' class would ask the 'factory' for a tag, i.e. 
factory.makeTag ("Form"). So you would wedge your own factory in there. 
 Choosing the factory is usually done with a Class.forName() where the 
string specifying the class comes from a configuration setting.  With 
some design effort, we should be able to come up with a definition for a 
factory class and a suitable set of interfaces which the whole project 
would be refactored to use, i.e. the IFormTag interface extends the 
ICompositeTag interface and adds form related methods; the ICompositeTag 
interface extends the IBaseTag interface and adds child accessors; and 
nothing references FormTag directly except the factory.

So then there is the problem of your factory supplying your special tag 
that implements IFormTag *and* IColorSupport when makeTag ("Form") is 
called. Most of what you need is already written in FormTag, you just 
need to add a couple of methods.  I think this is where dynamic proxies 
come in: http://java.sun.com/j2se/1.3/docs/guide/reflection/proxy.html. 
 The InvocationHandler would determine if the target method comes from 
IColorSupport, and if so perform the needful directly.  Otherwise it 
would delegate to the wrapped tag object.  This means the whole 
htmlparser project shuttles wrapped tag objects around and doesn't know 
it, till they bubble up to your code where you cast them to an 
IColorSupport and invoke the supportsColor() method:

Parser.setTagFactory ("ChrisBirdFactory");
Parser parser = new Parser (url);
parser.registerScanners ();
for (NodeIterator e = parser.elements (); e.hasMoreNodes (); )
{
    Node node = e.nextNode ();
    if (node instanceof Tag) // I presume all tags, but not nodes, 
support IColorSupport
        ((IColorSupport)node).supportsColor ();
}

Derrick

Christopher Bird wrote:

> Thanks for the reply, that is my dilemma.
>
> I am an old Smalltalk programmer from years gone by and have always 
> used what is sometimes called responsibility driven development. So 
> (at least in my head), the responsibilty for knowing that a tag 
> supports color or BG color is the tag's responsibilty, and not the 
> responsibility of some agent acting on the tag.
>
> The trouble with that style of development, especially for "packaged" 
> software is that you(I) find yourself(myself) in a bind like this one.
>
> Indeed I had to recompile the whole package! But since I have the 
> source in my project (to help me learn the intricacies of certain 
> behaviors - especially the creation of handlers for very complex tags) 
> that was no big deal for me. However now I am in violation of protocol 
> for OpenSource, I am sure.
>
> This really gets to the crux of OOness and Open Source development. 
> When there are requirements for classes high in the inheritance 
> hierarchy and they do "rightfully" belong there how does one get them 
> there - short term to overcome a specific issue, and long term as part 
> of the overall release cycle of the product.
>
> I am probably not the first to wonder this!
>
> BTW, I love the implementation. It took some mind-bending to get used 
> to it at first - again separating the responsibilities out so I can 
> factor my solutions properly was initialy a challenge, but I have 
> become very productive.
>
> Thank you so much for an excellent piece of technology.
>
> Regards
>
> Chris
>
>
>> From: Derrick Oswald <Der...@ro...>
>> To: Christopher Bird <se...@us...>
>> Subject: Re: Adding methods to Tag
>> Date: Fri, 29 Aug 2003 07:41:08 -0400
>> MIME-Version: 1.0
>> Received: from sc8-sf-mx1.sourceforge.net ([66.35.250.206]) by 
>> mc4-f42.law16.hotmail.com with Microsoft SMTPSVC(5.0.2195.5600); Fri, 
>> 29 Aug 2003 04:41:47 -0700
>> Received: from fep02-mail.bloor.is.net.cable.rogers.com 
>> ([66.185.86.72])by sc8-sf-mx1.sourceforge.net with esmtp (Exim 
>> 4.22)id 19shdQ-0004My-0Sfor se...@us...; Fri, 29 
>> Aug 2003 04:41:44 -0700
>> Received: from rogers.com ([24.102.205.244])          by 
>> fep02-mail.bloor.is.net.cable.rogers.com          (InterMail 
>> vM.5.01.05.12 201-253-122-126-112-20020820) with ESMTP          id 
>> <200...@ro...> 
>>          for <se...@us...>;          Fri, 29 Aug 
>> 2003 07:41:12 -0400
>> X-Message-Info: JGTYoYF78jGnyWgKUPy676KmG5L9JDoH
>> Message-ID: <3F4...@ro...>
>> User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2.1) 
>> Gecko/20030225
>> X-Accept-Language: en-us, en
>> References: <E19...@sc...>
>> In-Reply-To: <E19...@sc...>
>> X-Authentication-Info: Submitted using SMTP AUTH PLAIN at 
>> fep02-mail.bloor.is.net.cable.rogers.com from [24.102.205.244] using 
>> ID <der...@ro...> at Fri, 29 Aug 2003 07:41:12 -0400
>> X-Spam-Score: -2.1 (--)
>> X-Spam-Report: -2.1/5.0The original message has been attached along 
>> with this report, soyou can recognize or block similar mail in 
>> future.See http://spamassassin.org/tag/ for more details.Content 
>> preview:  Chris, It's unclear how your ChrisTag method would 
>> workwithout recompiling the whole package. The Tag class 
>> extendsAbstractNode, so presumably ChrisTag would extend AbstractNode 
>> and addthe methods you want, then Tag would extend ChrisTag. You 
>> would stillneed to 'fix' each new release by doctoring Tag. [...] 
>> Content analysis details:   (-2.10 points, 5 
>> required)USER_AGENT_MOZILLA_UA (0.0 points)  User-Agent header 
>> indicates a non-spam MUA (Mozilla)IN_REP_TO          (-0.5 points) 
>> Has a In-Reply-To headerX_ACCEPT_LANG      (-0.1 points) Has a 
>> X-Accept-Language  headerREFERENCES         (-0.5 points) Has a 
>> valid-looking References headerEMAIL_ATTRIBUTION  (-0.5 points) BODY: 
>> Contains what looks like an email attributionREPLY_WITH_QUOTES  (-0.5 
>> points) Reply with quoted text
>> Return-Path: Der...@ro...
>> X-OriginalArrivalTime: 29 Aug 2003 11:41:50.0512 (UTC) 
>> FILETIME=[89217300:01C36E22]
>>
>> Chris,
>>
>> It's unclear how your ChrisTag method would work without recompiling 
>> the whole package. The Tag class extends AbstractNode, so presumably 
>> ChrisTag would extend AbstractNode and add the methods you want, then 
>> Tag would extend ChrisTag. You would still need to 'fix' each new 
>> release by doctoring Tag.
>>
>> The best way is probably to have a class external to everything with 
>> the static methods needed (see Tag.breaksFlow() for example code):
>>    class ColorKnowledge {
>>        public static boolean supportsColor (Node node)
>>        {   return 
>> (listofNodesSupportingForegroundColor.contains(node.getText().toUpperCase()));} 
>>
>>        ...
>>
>> If it's generic enough, submit it and we'll add it to Node.
>>
>> Derrick
>>
>> Christopher Bird wrote:
>>
>>> I am new to using OpenSource code. I have found it very
>>>
>>> helpful, and am using the HTMLParser for a number of
>>>
>>> purposes.
>>>
>>>
>>>
>>> I am wanting to add some code to Tag - especially the
>>>
>>> following two methods:
>>>
>>>
>>>
>>> public boolean supportsColor()
>>>
>>> /* returns true iff the color attribute is valid for this tag
>>>
>>>
>>>
>>> public boolean supportsBGColor ()
>>>
>>>
>>>
>>> /* Returns true iff the bgColor attribute is valid for this tag
>>>
>>>
>>>
>>> I would be happy if you guys were to add that, but failing
>>>
>>> that what is the process if I have to do it myself? There may
>>>
>>> be a bunch of other things that I will want to add to Tag -
>>>
>>> for handling some of my own custom behaviors.
>>>
>>>
>>>
>>> I can see a couple of ways of doing this (none pretty). One is
>>>
>>> to create a new ChrisTag superclass and change Tag's
>>>
>>> implements clause to implements ChrisTag. I can then define
>>>
>>> my methods there. Of course the community doesn't get the
>>>
>>> benefit (? dubious in some cases, I fear) of my additions.
>>>
>>>
>>>
>>> The other obvious way is simply to add the methods to Tag
>>>
>>> itself. I am not wild about doing that either because as I
>>>
>>> download new editions of HTMLParser, my changes get lost -
>>>
>>> especially since I am a solo practitioner at the moment and
>>>
>>> am not using a source code management system.
>>>
>>>
>>>
>>> Any assistance would be gratefully appreciated - both to the
>>>
>>> short term (immediate) problem and to the general question.
>>>
>>>
>>>
>>> Thanks in advance
>>>
>>>
>>>
>>> Chris Bird
>>>
>>>