I need to provide an argument!

Question

I'm sorry I'm having a lot of questions, but I appreciate all the great help I've been getting to date.
Here's one I've just been thrown. If I do a CONTAINS search for 'dog', matches will include 'dog collar', 'dog star', 'hot dog' and so on, as well as 'cat and dog', no matter how I 'phrase' the SQL:
SQL> select * from texttst where contains (col1,'dog')>0;
COL1
hot dog
dog collar
dog and cat
SQL> select * from texttst where contains (col1,'{dog}')>0;
COL1
hot dog
dog collar
dog and cat
SQL> select * from texttst where contains (col1,'{"dog"}')>0;
COL1
hot dog
dog collar
dog and catMy boss has just said that's wrong. He thinks a search for 'dog' should just return 'dog' items, not items in which dog happens to be a word. Their old search used simple 'like '%,dog,%' tests in the where clause, and the inclusion of the comma delimiters meant their old search really did do 'word search' instead of phrase and partial-phrase matching.
What can I say to persuade him that the Oracle Text mechanism is the more "appropriate" or correct way of doing things? Is there anything I can do to my code to make it more 'word sensitive' and less able to spot words within phrases? Or would that be a dumb way to go?
I did think of this, but it doesn't seem to work:
SQL> select * from texttst where contains (col1,'",dog,"')>0;
COL1
,hot dog,sausage,
,hot,dog,sausage,The request to find comma-dog has nevertheless matched space-dog. But even if you could persuade Oracle to pay attention to the commas, would it be a good idea to do so?
It's a more philosophical question, I suppose, than a technical one. Just interested in people's views.

Answer 1

You can use a standard Oracle Text search to obtain the larger result set then add additional filtering criteria using like or instr to obtain the smaller set from the larger set. This should enable it to use the domain index to obtain the larger set quickly and still get the desired results. Please see the demonstration below.
SCOTT@orcl_11g> CREATE TABLE texttst
  2    (col1  VARCHAR2 (60))
  3  /
Table created.
SCOTT@orcl_11g> INSERT ALL
  2  INTO texttst VALUES ('hot dog')
  3  INTO texttst VALUES ('dog collar')
  4  INTO texttst VALUES ('dog and cat')
  5  INTO texttst VALUES (',hot dog,sausage,')
  6  INTO texttst VALUES (',hot,dog,sausage,')
  7  SELECT * FROM DUAL
  8  /
5 rows created.
SCOTT@orcl_11g> INSERT INTO texttst SELECT object_name FROM all_objects
  2  /
68663 rows created.
SCOTT@orcl_11g> CREATE INDEX testtxt_col1_idx ON texttst (col1)
  2  INDEXTYPE IS CTXSYS.CONTEXT
  3  /
Index created.
SCOTT@orcl_11g> EXEC DBMS_STATS.GATHER_TABLE_STATS (USER, 'TEXTTST')
PL/SQL procedure successfully completed.
SCOTT@orcl_11g> SET AUTOTRACE ON EXPLAIN
SCOTT@orcl_11g> SELECT *
  2  FROM   texttst
  3  WHERE  CONTAINS (col1, 'dog') > 0
  4  AND    INSTR (col1, ',dog,') > 0
  5  /
COL1
,hot,dog,sausage,
Execution Plan
Plan hash value: 149214848
| Id  | Operation                   | Name             | Rows  | Bytes | Cost (%CPU)| Time     |
|   0 | SELECT STATEMENT            |                  |     1 |    24 |     1   (0)| 00:00:01 |
|*  1 |  TABLE ACCESS BY INDEX ROWID| TEXTTST          |     1 |    24 |     1   (0)| 00:00:01 |
|*  2 |   DOMAIN INDEX              | TESTTXT_COL1_IDX |       |       |     0   (0)| 00:00:01 |
Predicate Information (identified by operation id):
   1 - filter(INSTR("COL1",',dog,')>0)
   2 - access("CTXSYS"."CONTAINS"("COL1",'dog')>0)
SCOTT@orcl_11g> SET AUTOTRACE OFF
SCOTT@orcl_11g>