From: Wolfgang M. <wol...@gm...> - 2007-02-07 22:01:40
|
> Yep I completely agree, I even now know how to work around this bug on > the server side simply stripping the $x in the where clause for the > 'older' clients that might connect. But as I say there seems to be some > bug maybe it's not worth fixing, though you never like to see anything > hang/stall. Recap: the bug seems to be when using a 'where' clause > which uses a variable such as '$x' with a beginning wildcard for &= > because 'pegase*' is ok but not '*pegase' or '*pegase*' on a big > collection. Fulltext indexing changed quite a lot since the last release, so we would need to recheck this on the current trunk. I just tried some queries on a large collection, but the difference in execution times between e.g. "*query" and "xquery" was as I expected. Querying with a wildcard at the start of the search string requires a scan over the index keys and will thus always be a little slower, but certainly not as much as to block the server. I also don't understand how the index lookup could be related to the structure of the FLWOR expression. I will keep this in mind and retest with other data. Wolfgang |