| Query expansion and FIRST_ROWS problem [message #353690] |
Tue, 14 October 2008 11:22 |
mys007
Messages: 1
Registered: October 2008
|
Junior Member |
|
|
Hello!
I have the following construction:
SELECT p.* FROM ( SELECT q.*, rownum rn FROM (
SELECT /*+ FIRST_ROWS(10) */ SCORE(1) AS score, idf FROM table WHERE CONTAINS(column, **query**, 1) > 0 ORDER BY SCORE(1) DESC, ROWID
) q WHERE rownum <= 10) p WHERE rn > 0;
When I use query:
<query><textquery lang="ENGLISH" grammar="CONTEXT"><progression>
<seq>string within section</seq>
</progression></textquery><score datatype="FLOAT" algorithm="DEFAULT"></query>
I get the results (>10) in some normal time. But with the following query:
<query><textquery lang="ENGLISH" grammar="CONTEXT"><progression>
<seq>string within section</seq>
<seq>fuzzy(string) within section</seq>
<seq>%string% within section</seq>
.....
</progression></textquery><score datatype="FLOAT" algorithm="DEFAULT"></query>
the search is soooo slow. I don't understand it: I set the limit to 10 first rows, which is already satisfied by the first sequence in the query expansion. But Oracle (obviously) seems to execute all the other ones as well!
Another observation: the presence of /*+ FIRST_ROWS(10) */ doesn't really matter from the time viewpoint... Removing the "within" part makes the searching much faster.
So what I am doing wrong? Thank you SOO much for answers!
Martin
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
