[infinispan-dev] Re: property tags in configuration files

[Manik Surtani]

Interesting discussions.

I agree that JAXB *would* have been the ideal way about this, but that  
brings in a) additional deps with Java5, and b) a not-so-pretty object  
model for configuration beans, and c) does not help us generate docs.

So essentially we have:

1) a simple(-ish) object model for programmatic configuration.
2) an XML parser to parse a coherent XML file and generate the beans  
in (1).
3) documentation for all setters in (1) and all elements/attributes in  
(2).

As Vladimir pointed out, this effort is to remove the hand-written  
parser we have which handles (2), and to save us writing hand-written  
docs for (3).  By annotating the object model in (1), we can generate  
appropriate documentation for (3) as well as parse XML to populate the  
bean for (2).

Emmanuel, note that these annotations are internal only and are not  
public API.  End users would configure stuff by doing:

	Configuration c = new Configuration();
	c.setBlah( blah );

The annotations are purely for core devs who would need to add a new  
configuration element for new feature X.  By adding a new getter and  
setter to the Configuration bean, and appropriately annotating it,  
this new config element is automagically added to the XSD, will be  
parsed from XML, and is documented.

That said, I do like the "fluent config" approach where folks can do:

	c.setBlah( blah ).setFoo( foo );

HTH,
Cheers
Manik

On 22 Jun 2009, at 18:41, Vladimir Blagojevic wrote:

> On 6/22/09 10:52 AM, Emmanuel Bernard wrote:
>> OK here is my take.
>>
>> Populating bean is fairly PITA for a user and a fluent API approach  
>> makes configuration more readable.
> Fluent config API is the one that allows chaining calls, right?  
> These are very elegant, but they seem more suitable for configs that  
> follow more of a grammar like structure. Infinispan's config tree is  
> not so nice so to speak.
>>
>> http://anonsvn.jboss.org/repos/hibernate/search/trunk/src/test/java/org/hibernate/search/test/configuration/ProgrammaticMappingTest.java
>> (check NotUseddefineMapping)
>> and the root class
>> http://anonsvn.jboss.org/repos/hibernate/search/trunk/src/main/java/org/hibernate/search/cfg/SearchMapping.java
>>
>> if you want to stay on the bean side, I am wondering why you don't  
>> use JAXB or something like that to bind the XML model to the object  
>> model.
>
> We want to reuse these annotations to kill three birds with one  
> stone. Configuration beans are the source, or if you want - code is  
> the source. We annotate those beans and get for free human readable  
> configuration documentation, configuration schema, and, if possible,  
> use those annotations in conjunction with beans to populate the  
> beans automatically during configuration reading from xml.
>
> Does it make better sense now? Appreciate your input!
>
> Regards,
> Vladimir
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev

--
Manik Surtani
manik at jboss.org
Lead, Infinispan
Lead, JBoss Cache
http://www.infinispan.org
http://www.jbosscache.org