LIKE and CONTAINS

Question

I create index:
SQL> CREATE INDEX UM2_ATTRS_STR_I_ctx ON UM2_ENTRIES_ATTRS
(STRING_VALUE) INDEXTYPE IS CTXSYS.CONTEXT;
2
Index created.
I've got 2 query:
SQL> select count(e.id) from um2_entries e , um2_entries_attrs a where e.id = a.entry_id
and a.name = 'user.name'
and a.string_value like '%__%';
2 3
COUNT(E.ID)
1734
94 msec!!!!!!!!
SQL> select count(e.id) from um2_entries e , um2_entries_attrs a where e.id = a.entry_id
and a.name = 'user.name'
and contains(a.string_value, '%__%', 1) > 0;
2 3
COUNT(E.ID)
1734
7 sec!!!!!!!!!!!!!!
Why does one of them (with LIKE) goes much faster than the other?

Answer 1

FYI - you might want to check the documentation on when wildcards are needed for CONTAINS vs. LIKE. From your example I can't tell whether you are looking for a part of a term (where a wildcard would be appropriate for CONTAINS), or if you are going to look for a term within a string (no wildcard needed for CONTAINS, but one is required for LIKE).
Example - Assuming a string value of HELLO WORLD
select count(e.id) from um2_entries e , um2_entries_attrs a where e.id = a.entry_id
and a.name = 'user.name'
and a.string_value like 'Hello';
** No record found - case-sensitive and partial string
select count(e.id) from um2_entries e , um2_entries_attrs a where e.id = a.entry_id
and a.name = 'user.name'
and a.string_value like 'HELLO';
** No record found - Partial string
select count(e.id) from um2_entries e , um2_entries_attrs a where e.id = a.entry_id
and a.name = 'user.name'
and a.string_value like 'HELLO%';
** Record found
select count(e.id) from um2_entries e , um2_entries_attrs a where e.id = a.entry_id
and a.name = 'user.name'
and CONTAINS(a.string_value, 'Hello') > 0;
** Record found - Oracle Text is case-insensitive
** by default, so mixed-case here doesn't matter. Also,
** the search is against the DR$ tables where
** HELLO is a token. No need for a wildcard.
I just wanted to make the differences between these two searches at least partially clear. Your earlier comparison seemed to indicate that you expected no difference regardless of search criteria and that just isn't the case. Some of the reasons for using wildcard or string functions are not valid when Oracle Text is used. That's one reason it is very usefull. Add stemming, fuzzy search, or a thesaurus and you go FAR beyond LIKE.
Please do reference the CONTAINS sections in the online docs if you have more questions on these differences. They go far beyond where I just did.
-Ron