commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christopher Lenz <cml...@gmx.de>
Subject Re: digester 2.0 [WAS Re: [digester] [PROPOSAL] More pattern matching flexibility]
Date Tue, 03 Sep 2002 13:56:43 GMT
robert burrell donkin wrote:
[...]
> Rules is an interface but since it's poorly named anyway, we might 
> actually think about creating a new 'Matcher' class. we could then 
> retain backwards compatibility by using an adapter. we could use 
> cristopher's nice strategy to retain compatibility for the rule class.

Your Matcher idea in combination with an adapter for backwards 
compatibility is pretty damn good ;o)

And it got me experimenting...

Let's assume the responsibility of the Matcher interface would be only 
to match string patterns against the current XML document context. For 
the start, it could have an interface like this

   public interface Matcher {

       // add a pattern
       public void add(String pattern);

       // remove all patterns
       public void clear();

       // return all patterns that match the current XML document
       // context, in the order they've been added
       public List matches(DigesterContext context);

       // return all patterns in the order they've been added
       public List patterns();

   }

Note that the Matcher doesn't store the Rules themselves, that will be 
the responsibility of Digester. Digester will just request the matched 
patterns from the Matcher and then lookup the corresponding rules in a 
map (for example).

Now, there might be a simpler and yet more flexible design... what if 
someone wanted to create a Matcher that understands full-blown XPath 
expressions. In that case DigesterContext would be the limiting factor, 
as the Matcher would not have access to - for example - the attributes 
of the parent element. In other cases, the DigesterContext might have 
more info than the Matcher requires. The algorithm in the current 
RulesBase is enough for many use-cases and only needs a simple string to 
match against. Here the DigesterContext adds unneeded overhead.

So, my idea was to make the Matcher a SAX ContentHandler and not pass a 
DigesterContext (or something similar) at all:

   public interface Matcher extends ContentHandler {
       ...

       // return all patterns that match the current XML document
       // context
       public List matches();

       ...
   }

Digester in turn would call all the Matcher's ContentHandler methods 
from it's own ContentHandler implementation methods. The Matcher could 
get all the information (and *only* the information) about the parsed 
XML that it actually needs to perform it's matching tasks.

I've done some experimenting with a RulesAdapter (i.e. a class that 
implements Matcher but wraps around a Rules implementation), and it 
looks like we could get away with almost no breaking API changes.

Again, I actually like this approach much better than my own 
DigesterContext idea... it seems cleaner, leaner and more flexible. What 
do you think?

-- 
Christopher Lenz
/=/ cmlenz at gmx.de


--
To unsubscribe, e-mail:   <mailto:commons-dev-unsubscribe@jakarta.apache.org>
For additional commands, e-mail: <mailto:commons-dev-help@jakarta.apache.org>


Mime
View raw message