[Exist-open] query speed problem: eXist or inefficient XQuery statement

[Jeff W. <jef...@tb...>]

Hi,

We are building a web-based repository of Tibetan texts.  We are using  
eXist as our read only web index and have been very happy to date with  
the XQueryServlet and XQueryGenerator API's, in addition to the others  
supported.  Also very happy with the amount of attention paid here on  
the list serve.

  We are building our entire web interface using XQuery.  Currently our  
testing is on my MAC OS X 10.3.8,  G4  1.67 GHz Processor, 1 GB RAM .   
Our production server will be a Linux Redhat server, with at least 2 GB  
RAM with 3 GHz processor, 1 GB RAM.

A very common query for us runs quite slow and I am just wondering if  
anyone has thoughts about how to improve the performance.  I dont know  
if it is an inefficient XQuery statement, or lack of experience on my  
part with eXist's indexing method.  There are roughly 100K XML docs for  
TBRC-WORKS (in its own collection) and 10K docs for TBRC-PERSONS (in  
its own collection) and 2K docs for TBRC-TOPICS (in its own  
collection).

Here is a sample query to get all the works by an author.  For this I  
have to query the tbrc-works collection.  Author is an element within a  
single Works document.

     for $works in  
collection($tbrc-xutils:db-path-works)/w:work/w:creator[@person=$RID]/ 
../w:title[position()=1]
         let $doc-name :=  
concat($tbrc-xutils:db-path-works,"/M",$works/../@RID,".xml")
          order by $works ascending
     return <li><a href="tbrc-work-detail.xq?RID={ $works/../@RID }"> {  
$works }</a>{ tbrc-xutils:publication-info-summary($doc-name) }</li>

Within the FLOWR statement, I call a function  
tbrc-xutils:publication-info-summary($doc-name) which grabs a few  
elements from an associated XML document. I also return some basic HTML  
formating.

The problem is that this a very common query and it runs too slow.  I  
also notice that if I just return the XML without formating it seems to  
be faster.  That is

return $works

rather than

return <li><a href="tbrc-work-detail.xq?RID={ $works/../@RID }"> {  
$works }</a>{ tbrc-xutils:publication-info-summary($doc-name) }</li>

A similar query is used to generate all works with a specific subject  
classification (no function call within the return)

     for $works in  
collection($tbrc-xutils:db-path-works)/w:work/w:subject[@class=$RID]/ 
../w:title[position()=1]
          order by $works ascending
     return <br><a href="tbrc-work-detail.xq?RID={ $works/../@RID }"> {  
$works }</a></br>

This query is just as slow as the first.


So the question is, should I be structuring the collections  
differently?  Or specifying indices differently?  Or using a different  
XQuery statement.  Was wondering if anyone has any thoughts.

Best wishes,
Jeff Wallman
Tibetan Buddhist Resource Center
tbrc.org