[infinispan-dev] Module configuration proposal

[Navin Surtani]

On 3 Nov 2009, at 18:24, Vladimir Blagojevic wrote:

> On 09-11-03 12:59 PM, Galder Zamarreno wrote:
>>
>> On 11/03/2009 04:52 PM, Vladimir Blagojevic wrote:
>>
>>> Hey guys,
>>>
>>> I completed module configuration proposal Manik and I agreed upon.  
>>> It is
>>> very important to get this one right so I want to run it by you  
>>> before
>>> proceeding further!
>>>
>>> Have a look at [1] and follow proposed module configuration  
>>> example below:
>>>
>>> <infinispan xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
>>> xmlns="urn:infinispan:config:4.0">
>>> <global>
>>> <transport clusterName="demoCluster"/>
>>> </global>
>>>
>>> <default>
>>> <clustering mode="replication">
>>> </clustering>
>>> <modules>
>>> <module name="query"
>>> configClassName 
>>> ="org.infinispan.query.config.QueryConfigurationBean">
>>> <indexing enabled="true" indexLocalOnly="true"/>
>>> </module>
>>> </modules>
>>> </default>
>>> </infinispan>
>>>
>>> A few things to notice:
>>>
>>> a) Any configuration (including default and namedCache) can have a
>>> module definition
>>> b) Instance of configClassName is a rich object structure filled  
>>> from
>>> module's child XML content
>>>
>> I think configClassName should simply be 'class' as per the rest of
>> situations where we're configuring FQCNs.
>>
>
> Ok, makes sense!
>>
>>> c) Module's XML content can be arbitrary
>>>
>> So,<indexing enabled="true" indexLocalOnly="true"/>  is the XML  
>> from the
>> query module that the query module can read itself?
>>
>>
> Yes, exactly. And it is read into object graph rooted in class  
> specified
> by configClassName.
> XML unmarshalling into object graph is done by JAXB, just as we do it
> elsewhere!

I disagree. I could be wrong, but I think it's easiest if the  
QueryConfigurationBean does the work of setting up stuff kind of like  
how the QueryHelper [1] currently does. This way, if the query module  
is to be used, once these configurations are registered the  
appropriate interceptor can be added to the chain.

Just a quick question on general Cache setup. Currently the  
QueryInterceptor is added to chain after the locking interceptor. Just  
wondering that if things are done this way, i.e. we create and add the  
QueryInterceptor to the chain when a new instance of the  
QueryConfigurationBean is created, is that going to cause problems  
about where on the chain the QueryInterceptor sits? [Just thinking out  
loud]


>
>
>>> d) Custom module configuration bean is accessed by module name:
>>>
>>>
>>>         Configuration def = c.parseDefaultConfiguration();
>>>         ModuleConfigurationBean extensionConfig =
>>> def.getModuleConfigurationBean("query");
>>>         QueryConfigurationBean bean = (QueryConfigurationBean)
>>> extensionConfig.getConfigurationBean();
>>>         assert bean.isEnabled();
>>>         assert bean.isIndexLocalOnly();
>>>
>>>


</SNIP>

[1] http://fisheye.jboss.org/browse/Infinispan/trunk/query/src/main/java/org/infinispan/query/backend/QueryHelper.java?r=trunk

Navin Surtani

Intern Infinispan
Intern JBoss Cache Searchable

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.jboss.org/pipermail/infinispan-dev/attachments/20091104/b376b371/attachment-0002.html