For people not following:
http://www.nabble.com/SolrConfig.Initializable-tf4665036.html
In looking at how initialization throughout solr, I think the cleanest
and best approach to move forward is to break the existing init() method
for:
1. TokenFilterFactory.init( Map<String,String> args )
2. TokenizerFactory.init( Map<String,String> args )
3. SolrRequestHandler.init( NamedList args )
and replace them with:
1. TokenFilterFactory.init( SolrConfig, Map<String,String> args )
2. TokenizerFactory.init( SolrConfig, Map<String,String> args )
3. SolrRequestHandler.init( SolrCore, NamedList args )
In SOLR-215, Henri added some logic to deprecate the existing
TokenFilterFactory/TokenizerFactory -- it works, but it is messy and
means we can't just use o.a.s.util.AbstractPluginLoader<T> to load each
case.
I know that 1.3 should not break APIs, if we make this change
RequestHandlers/Tokenizers/TokenFilters build under 1.2 could not run
under 1.3
In my opinion, the simplicity of the change far outweighs the loops and
ugliness we would need to add to be able to support 1.2 libs under 1.3.
Thoughts?
thanks
ryan
