For performance reasons, I would not recommend using the //* notation. Its expanded version would like the following: /descendant::node()[attribute::product_id='VPU-1001'] The descendant axis of the root node includes every other node in the tree, which means you would search every single node for your specified product_id attribute. Its not very specific and of limited use for large documents.
I'm assuming you mean this for the third query: /[EMAIL PROTECTED]'VPU-1001']] This query is not very useful in your document. It will return the entire products element, including all of its children. This is of course your entire document, so I don't think this is what you want to do. I hope this helps. John -----Original Message----- From: Dave Hurrell [mailto:[EMAIL PROTECTED] Sent: Wednesday, May 15, 2002 3:02 AM To: xindice-users Subject: Xpath query performance Given a large file of which the following is a small sample: <?xml version="1.0"?> <products xmlns:src="http://xml.apache.org/xindice/Query" src:col="/db/vp" src:key="prods"> <product product_id="VPU-1001"> <title>Mamadou Ly Mandinka Drum Master</title> <description>Mandinka drumming</description> </product> <product product_id="VPU-1002"> <title>Tabala Wolof</title> <description>Sufi drumming</description> </product> </products> I believe I can obtain the same document with any one of the following xpath queries: /products/[EMAIL PROTECTED]'VPU-1001'] //*/[EMAIL PROTECTED]'VPU-1001'] /[EMAIL PROTECTED]'VPU-1001'] Do they all the above xpath queries have the same performance? What effect does indexing have? I suspect it would make no difference if the products document was stored as a single document. I think the xquery engine would read the document sequentially searching for the product_id. What happens if I store each product separately into a products database. Indexing allows direct access to the desired product. Does an xindice xpath query use any indexes?
