[sword-devel] Indexed searching

[DM Smith]

I think that we are a lot closer than we were, but I don't think we are 
there yet.

As I have noted in a previous note, there are specific hurdles that we 
have to overcome before we can offer indexes for download.

Here is where we stand with the Sword indexing implementation:
When a verse is indexed, it is treated as a single document. We store 
with that document the OSIS verse reference.
At this time we also index Strong numbers for that verse, if present. We 
do not index footnotes, notes or any other gloss that is present in the 
document. It is conceivable that we will want to do that at some time. 
Fortunately, adding more fields does not break previous code that does 
not look for them.

Also, we are using an English analyzer to analyze all Bibles regardless 
of language. This will need to be changed to handle other languages 
properly. Since the search needs to use the same analyzer as the the 
indexer, we will need to figure out how to indicate what analyzer was 
used to build the index. The conf may be the ideal location.

Chris Little wrote:

> Are we certain (or reasonably certain) that we have settled on a set 
> of (C)Lucene settings? Can we assume binary compatability with 
> versions of Sword going forward from here?
>
> Alternately, should we think about indicating some info about the 
> index in the .conf files? Something to indicate the version of Lucene, 
> the settings, etc. perhaps.
>
> --Chris
>
> DM Smith wrote:
>
>> Not only can you copy them to a different PC, the machine can be 
>> running a different OS (linux, mac) and it can be a different 
>> program, say JSword.
>>
>> Barry Drake wrote:
>>
>>> Hi there ......
>>>
>>> I've been playing with the indexed search .......  it's wonderful!  
>>> Now, can someone tell me; if I transfer the lucene directories to a 
>>> different box, do I have to do anything else to get the indexed 
>>> search to work on that one?  I ask, because I have a very slow 
>>> Pentium 3 that I sometimes work on, and the indexed search would be 
>>> useful on that - but it would take forever to build the indexes.
>>>
>>> God bless,
>>>
>>> Barry
>>>
>