XQuery sorting issue - sorting

Question

Right now I am in a big trouble. I am new in XQuery, but right now I'm in a project using Xquery using PHP and XSLT..
In our project we have large no. data (its a property listing site) and I'm storing that data to Barkeley DB (XML DB).
The problem is when I am searching for a property its taking too much time for getting the result. The ORDER BY is creating the problem(Query 1).. with out sorting its working fine(Query 2). But for my project sorting is needed and its very impotent. So kindly please check my query(Query1) and please give me a solution as soon as possible. Following are the query:
Query1:
let $properties := (
for $property in collection('bdb/properties.dbxml')/properties/property
[ ( sale_price >=60000 and sale_price <=500000 ) and ( building_square_footage >=300 and building_square_footage <=3000 ) and ( bedrooms >=2 and bedrooms <=6 ) ]
order by
contains($property/mls_agent_id/text(), '505199') descending,
matches($property/mls_office_id/text(), '^CBRR') ascending,
$property/sale_price/number() descending
return $property
)
let $properties := subsequence($properties,10,10) return <properties>{$properties}</properties>
Query 2:
let $properties := (
for $property in subsequence (
collection('bdb/properties.dbxml')/properties/property
[ ( sale_price >=60000 and sale_price <=500000 ) and ( building_square_footage >=300 and building_square_footage <=3000 ) and ( bedrooms >=2 and bedrooms <=6 ) ]
, 1, 10)
)
descending return $property
) return <properties>{$properties}</properties>

Answer 1

I don't know if that helps.. But you may be more successful by trying an alternative XML Databases that offers more advanced query optimizations. Just a guess, Hannes

Answer 2

I think sorting is the wrong tool to use here. You're really getting 3 lists - the agent's listings, the office's listings other than the agent's, and other listings. It might work better to just make these 3 queries, and the optimizer can more efficiently select the nodes for each subquery.
let $properties := (
for $property in collection('bdb/properties.dbxml')/properties/property
[ ( sale_price >=60000 and sale_price <=500000 ) and ( building_square_footage >=300 and building_square_footage <=3000 ) and ( bedrooms >=2 and bedrooms <=6 ) and mls_agent_id = '505199']
order by
$property/sale_price/number() descending
return $property,
for $property in collection('bdb/properties.dbxml')/properties/property
[ ( sale_price >=60000 and sale_price <=500000 ) and ( building_square_footage >=300 and building_square_footage <=3000 ) and ( bedrooms >=2 and bedrooms <=6 ) and (starts_with(office_id, 'CBRR') and not(mls_agent_id = '505199'))]
order by
$property/sale_price/number() descending
return $property,
for $property in collection('bdb/properties.dbxml')/properties/property
[ ( sale_price >=60000 and sale_price <=500000 ) and ( building_square_footage >=300 and building_square_footage <=3000 ) and ( bedrooms >=2 and bedrooms <=6 ) and not (starts_with(office_id, 'CBRR')]
order by
$property/sale_price/number() descending
return $property
)
let $properties := subsequence($properties,10,10) return <properties>{$properties}</properties>
Other things that may be helpful are
try using starts-with instead of matches - a simple string match is likely to be a faster than a regular expression, and easier to optimize
if you break it into multiple selects as I did above, you can avoid the second or third selects if you already have enough items selected.
don't select the entire <property> node only to discard most of what you've selected with subsequence. Particularly when sorting, this likely means a lot of memory bandwidth. It's better to select just some unique identifier and then use those identifiers to get the rest of the property info later.
For example, rather than
let $properties:= for $property in collection('properties.dbxml')/properties/property[.....]
return $property
let $properties := subsequence($properties,10,10)
return <properties>{$properties}</properties>
do instead
let $property_ids:= for $property in collection('properties.dbxml')/properties/property[.....]
return $property/unique_id
return <properties>{
for $id in subsequence($property_ids,10,10) return
collection('properties.dbxml')/properties/property[unique_id = $id]
}</properties>
This means that the in-memory sequence will be a bunch of small ids rather than large nodes. Of course, this means that you need to have these unique ids to begin with, but I suspect a MLS database has such things.