Hi Robert,
thanks for your precise observations. I agree that the second query should have been optimized as well, and is now rewritten in the latest snapshot [1] (the optimizer was bold enough to reject all path expressions that contained any positional predicates).
Hope this helps, Christian
[1] http://files.basex.org/releases/latest/
On Thu, Jan 30, 2014 at 3:25 PM, Robert schlichtner r_schlichtner@hotmail.com wrote:
Hello,
I found that a number of similar queries have quite different performance:
xquery doc('mcev_temp_5/20140130_11_55_40.xml')//column[@name='KLV_m1_rdr']/v takes only 0.68ms, which is great.
But xquery doc('mcev_temp_5/20140130_11_55_40.xml')//column[@name='KLV_m1_rdr']/v[1] takes 1479.81ms.
If I change this to the equivalent (for my data) xquery (doc('mcev_temp_5/20140130_11_55_40.xml')//column[@name='KLV_m1_rdr'])/v[1] it runs fast again (0.53ms)
Checking the query execution plan I found that only query 1 and 3 utilize the attribute index.
I find this rather surprising, and I would like to ask if there is some general reason/rule why the second can not use the index, so I can avoid such 'bad' queries in the future.
Thanks!
(I am using BaseX 7.7.2 on Windows XP.)
Regards,
Robert
BaseX-Talk mailing list BaseX-Talk@mailman.uni-konstanz.de https://mailman.uni-konstanz.de/mailman/listinfo/basex-talk