lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Lance Norskog <goks...@gmail.com>
Subject Re: Solr architecture diagram
Date Sun, 10 Apr 2011 21:04:33 GMT
Very cool! "The Life Cycle of the IndexSearcher" would also be a great
diagram. The whole dance that happens during a commit is hard to
explain. Also, it would help show why garbage collection can act up
around commits.

Lance

On Sun, Apr 10, 2011 at 2:05 AM, Jan Høydahl <jan.asf@cominvent.com> wrote:
>> Looks really good, but two bits that i think might confuse people are
>> the implications that a "Query Parser" then invokes a series of search
>> components; and that "analysis" (and the pieces of an analyzer chain)
>> are what to lookups in the underlying lucene index.
>>
>> the first might just be the ambiguity of "Query" .. using the term
>> "request parser" might make more sense, in comparison to the "update
>> parsing" from the other side of hte diagram.
>
> Thanks for commenting.
>
> Yea, the purpose is more to show a conceptual rather than actual relation
> between the different components, focusing on the flow. A 100% technical
> correct diagram would be too complex for beginners to comprehend,
> although it could certainly be useful for developers.
>
> I've removed the arrow between QueryParser and search components to clarify.
> The boxes first and foremost show that query parsing and response writers
> are within the realm of search request handler.
>
>> the analysis piece is a little harder to fix cleanly.  you really want the
>> end of the analysis chain to feed back up to the searh components, and
>> then show it (most of hte search components really) talking to the Lucene
>> index.
>
> Yea, I know. Showing how Faceting communicate with the main index and
> spellchecker with its spellchecker index could also be useful, but I think
> that would be for another more detailed diagram.
>
> I felt it was more important for beginners to realize visually that
> analysis happens both at index and search time, and that the analyzers
> align 1:1. At this stage in the digram I often explain the importance
> of matching up the analysis on both sides to get a match in the index.
>
> --
> Jan Høydahl, search solution architect
> Cominvent AS - www.cominvent.com
>
>



-- 
Lance Norskog
goksron@gmail.com

Mime
View raw message