From: Birnbaum, David J <djbpitt@pi...> - 2009-05-22 03:49:52
To follow up a bit on my own post (subject heading "indexing blind spot?"), I've noticed that:
for $j in (1,2,3) return data($i/following-sibling::word[$j])
is reasonably fast, while
for $j in (1,2,3) return data($i/following::word[$j])
is excruciatingly slow. Unfortunately, I need the latter because I have structures like:
where $i might point to the second <word> element in the first <line>, and the next <word> element (the one with the value "third") is on the "following" axis, but not on the "following-sibling" axis. Is eXist supposed to be much faster with the "following-sibling" axis than with the "following" axis, or is there some optimization or index use that happens in the first case but, by accident, not in the second? Should I be coding this some other way?
Tested with eXist 1.2.5 on Vista.
Get latest updates about Open Source Projects, Conferences and News.