Determining Where Hits Were Found (XML Section Groups)

Question

Hello all,
Hoping for a bit of help. It's my first time posting here, but I've benefitted from reading a lot of your posts. I have created a single Oracle Text index for a resume database of sorts. It utilizes a user datastore. The procedure I use gathers a a single user's information in my database and organizes it via XML tags that are assigned in the procedure - outputs a CLOB. I did this in order to try the AUTO SECTION GROUP option when indexing, so that I could use the WITHIN operator.
So, for example, the procedure is given the rowid for a single employee, and the output may be:
<ALL>
<PERSONAL>
83019
John Smith
10/05/1973
Baltimore, MD
</PERSONAL>
<SKILLS>
Oracle Database Administrator
.NET Developer
</SKILLS>
</ALL>
My problem is I wish to know the SPECIFIC section groups where hits were registered if I search all of the XML
SELECT EMPLOYEE_ID
FROM EMPLOYEE_TBL
WHERE CONTAINS (DUMMY_COLUMN, 'developer WITHIN ALL', 1) > 0
ORDER BY SCORE (1) DESC
My output is:
EMPLOYEE_ID
83019
Is there any way I can also list the sections where hits were registered for this user? In this example, I'd want the "SKILLS" section to show up in the result set.

Answer 1

skMed,
Yes, that is a problem. Alternatives are the first method using UNION ALL or creating a MATERIALIZED VIEW and indexing that, similar to what I demonstrated to the other poster on this thread, or CTX_DOC.HIGHLIGHT as Roger said. I have provided a demonstration of the problem below, followed by an alternative solution that uses CTX_DOC.HIGHLIGHT as Roger suggested. I used a function but Roger may have a more elegant method.
SCOTT@orcl_11g> CREATE TABLE employee_tbl
  2    (employee_id  NUMBER,
  3       fname          VARCHAR2 (7),
  4       lname          VARCHAR2 (7),
  5       dob          DATE,
  6       city          VARCHAR2 (9),
  7       state          VARCHAR2 (2),
  8       dummy_column VARCHAR2 (1),
  9       CONSTRAINT emp_id_pk PRIMARY KEY (employee_id))
10  /
Table created.
SCOTT@orcl_11g> CREATE TABLE skills
  2    (employee_id  NUMBER,
  3       skill          VARCHAR2 (30),
  4       CONSTRAINT skills_id_fk FOREIGN KEY (employee_id) REFERENCES employee_tbl (employee_id))
  5  /
Table created.
SCOTT@orcl_11g> CREATE TABLE jobs
  2    (employee_id  NUMBER,
  3       job          VARCHAR2 (30),
  4       CONSTRAINT job_id_fk FOREIGN KEY (employee_id) REFERENCES employee_tbl (employee_id))
  5  /
Table created.
SCOTT@orcl_11g> BEGIN
  2    INSERT INTO employee_tbl VALUES
  3        (1, 'Adam', NULL, NULL, NULL, NULL, NULL);
  4    INSERT INTO employee_tbl VALUES
  5        (2, 'Bob', NULL, NULL, NULL, NULL, NULL);
  6    INSERT INTO employee_tbl VALUES
  7        (3, 'Charles', NULL, NULL, NULL, NULL, NULL);
  8    INSERT INTO employee_tbl VALUES
  9        (4, 'David', NULL, NULL, NULL, NULL, NULL);
10    INSERT INTO skills VALUES (1, 'OTHERS');
11    INSERT INTO skills VALUES (2, '.NET');
12    INSERT INTO skills VALUES (3, 'OTHERS');
13    INSERT INTO skills VALUES (4, '.NET');
14    INSERT INTO jobs   VALUES (1, '.NET');
15    INSERT INTO jobs   VALUES (2, 'OTHERS');
16    INSERT INTO jobs   VALUES (3, '.NET');
17    INSERT INTO jobs   VALUES (4, '.NET');
18  END;
19  /
PL/SQL procedure successfully completed.
SCOTT@orcl_11g> CREATE OR REPLACE PROCEDURE resume
  2    (p_rowid IN     ROWID,
  3       p_clob     IN OUT CLOB)
  4  AS
  5  BEGIN
  6    FOR e IN
  7        (SELECT employee_id,
  8             employee_id || ' ' || fname || ' ' || lname || ' ' || dob || ' '
  9             || city || ',     ' || state AS personal
10         FROM      employee_tbl
11         WHERE  ROWID = p_rowid)
12    LOOP
13        DBMS_LOB.WRITEAPPEND (p_clob, 10, '<PERSONAL>');
14        DBMS_LOB.WRITEAPPEND (p_clob, LENGTH (e.personal), e.personal);
15        DBMS_LOB.WRITEAPPEND (p_clob, 11, '</PERSONAL>');
16        DBMS_LOB.WRITEAPPEND (p_clob, 8, '<SKILLS>');
17        FOR s IN (SELECT skill FROM skills WHERE employee_id = e.employee_id) LOOP
18          DBMS_LOB.WRITEAPPEND (p_clob, LENGTH (s.skill) + 1, s.skill || ' ');
19        END LOOP;
20        DBMS_LOB.WRITEAPPEND (p_clob, 9, '</SKILLS>');
21        DBMS_LOB.WRITEAPPEND (p_clob, 12, '<JOBHISTORY>');
22        FOR j IN (SELECT job FROM jobs WHERE employee_id = e.employee_id) LOOP
23          DBMS_LOB.WRITEAPPEND (p_clob, LENGTH (j.job) + 1, j.job || ' ');
24        END LOOP;
25        DBMS_LOB.WRITEAPPEND (p_clob, 13, '</JOBHISTORY>');
26    END LOOP;
27  END resume;
28  /
Procedure created.
SCOTT@orcl_11g> SHOW ERRORS
No errors.
SCOTT@orcl_11g> BEGIN
  2    CTX_DDL.CREATE_PREFERENCE ('your_datastore', 'USER_DATASTORE');
  3    CTX_DDL.SET_ATTRIBUTE ('your_datastore', 'PROCEDURE', 'resume');
  4    CTX_DDL.CREATE_SECTION_GROUP ('your_sec_group', 'XML_SECTION_GROUP');
  5    CTX_DDL.ADD_FIELD_SECTION ('your_sec_group', 'PERSONAL', '<PERSONAL>', TRUE);
  6    CTX_DDL.ADD_FIELD_SECTION ('your_sec_group', 'SKILLS', '<SKILLS>', TRUE);
  7    CTX_DDL.ADD_FIELD_SECTION ('your_sec_group', 'JOBHISTORY', '<JOBHISTORY>', TRUE);
  8  END;
  9  /
PL/SQL procedure successfully completed.
SCOTT@orcl_11g> CREATE INDEX resume_idx ON employee_tbl (dummy_column)
  2  INDEXTYPE IS CTXSYS.CONTEXT
  3  PARAMETERS
  4    ('DATASTORE     your_datastore
  5        SECTION GROUP     your_sec_group
  6        STOPLIST     CTXSYS.EMPTY_STOPLIST')
  7  /
Index created.
SCOTT@orcl_11g> CREATE OR REPLACE FUNCTION list_element
  2    (p_string     VARCHAR2,
  3       p_element    INTEGER,
  4       p_separator  VARCHAR2 DEFAULT ':')
  5    RETURN          VARCHAR2
  6  AS
  7    v_string      VARCHAR2 (32767);
  8  BEGIN
  9    v_string := p_separator || p_string || p_separator;
10    v_string := SUBSTR (v_string, INSTR (v_string, p_separator, 1, p_element) + LENGTH (p_separator));
11    RETURN SUBSTR (v_string, 1, INSTR (v_string, p_separator) - 1);
12  END list_element;
13  /
Function created.
SCOTT@orcl_11g> SELECT * FROM employee_tbl ORDER BY employee_id
  2  /
EMPLOYEE_ID FNAME   LNAME   DOB       CITY      ST D
          1 Adam
          2 Bob
          3 Charles
          4 David
SCOTT@orcl_11g> SELECT * FROM skills ORDER BY employee_id
  2  /
EMPLOYEE_ID SKILL
          1 OTHERS
          2 .NET
          3 OTHERS
          4 .NET
SCOTT@orcl_11g> SELECT * FROM jobs ORDER BY employee_id
  2  /
EMPLOYEE_ID JOB
          1 .NET
          2 OTHERS
          3 .NET
          4 .NET
SCOTT@orcl_11g> COLUMN ixv_value    FORMAT A26
SCOTT@orcl_11g> COLUMN from_section FORMAT A15
SCOTT@orcl_11g> SELECT ixv_value,
  2           list_element (ixv_value, 1) AS from_section,
  3           TO_NUMBER (list_element (ixv_value, 3)) AS token_type
  4  FROM   ctx_user_index_values
  5  WHERE  ixv_index_name = 'RESUME_IDX'
  6  AND    ixv_class = 'SECTION_GROUP'
  7  /
IXV_VALUE                  FROM_SECTION     TOKEN_TYPE
SKILLS:SKILLS:17:Y         SKILLS                   17
PERSONAL:PERSONAL:16:Y     PERSONAL                 16
JOBHISTORY:JOBHISTORY:18:Y JOBHISTORY               18
SCOTT@orcl_11g> COLUMN token_info FORMAT A30 WORD_WRAPPED
SCOTT@orcl_11g> SELECT token_text, token_type, token_first, token_last, token_count
  2  FROM   dr$resume_idx$i
  3  WHERE  token_text = 'NET'
  4  /
TOKEN_TEXT       TOKEN_TYPE TOKEN_FIRST TOKEN_LAST TOKEN_COUNT
NET                       0           1          4           4
NET                      17           2          4           2
NET                      18           1          4           3
SCOTT@orcl_11g> -- problem:
SCOTT@orcl_11g> SELECT t.employee_id, t.fname,
  2           SCORE (1),
  3           list_element (v.ixv_value, 1) AS from_section
  4  FROM   employee_tbl t, dr$resume_idx$i i, dr$resume_idx$k k, ctx_user_index_values v
  5  WHERE  CONTAINS (t.dummy_column, '.NET%', 1) > 0
  6  AND    i.token_text = 'NET'
  7  AND    i.token_type = TO_NUMBER (list_element (v.ixv_value, 3))
  8  AND    k.docid BETWEEN i.token_first AND i.token_last
  9  AND    k.textkey = t.ROWID
10  ORDER  BY t.employee_id
11  /
EMPLOYEE_ID FNAME     SCORE(1) FROM_SECTION
          1 Adam             3 JOBHISTORY
          2 Bob              3 JOBHISTORY
          2 Bob              3 SKILLS
          3 Charles          3 SKILLS
          3 Charles          3 JOBHISTORY
          4 David            6 JOBHISTORY
          4 David            6 SKILLS
7 rows selected.
SCOTT@orcl_11g> -- alternative solution:
SCOTT@orcl_11g> CREATE OR REPLACE FUNCTION from_section
  2    (p_emp_id IN VARCHAR2,
  3       p_text      IN VARCHAR2,
  4       p_rowid  IN ROWID)
  5    RETURN VARCHAR2
  6  AS
  7       hightab   ctx_doc.highlight_tab;
  8       v_clob       CLOB;
  9       v_clob2   CLOB;
10       v_section VARCHAR2 (32767);
11       v_result  VARCHAR2 (32767);
12  BEGIN
13    CTX_DOC.HIGHLIGHT ('resume_idx', p_emp_id, p_text, hightab, FALSE);
14    DBMS_LOB.CREATETEMPORARY (v_clob, TRUE);
15    DBMS_LOB.CREATETEMPORARY (v_clob2, TRUE);
16    resume (p_rowid, v_clob);
17    FOR i IN 1 .. hightab.COUNT LOOP
18        v_clob2 := SUBSTR (v_clob, 1, hightab(i).offset);
19        v_clob2 := SUBSTR (v_clob2, 1, INSTR (v_clob2, '>', -1) - 1);
20        v_section := SUBSTR (v_clob2, INSTR (v_clob2, '<', -1) + 1);
21        v_result := v_result || ',' || v_section;
22    END LOOP;
23    DBMS_LOB.FREETEMPORARY (v_clob);
24    DBMS_LOB.FREETEMPORARY (v_clob2);
25    RETURN LTRIM (v_result, ',');
26  END from_section;
27  /
Function created.
SCOTT@orcl_11g> SHOW ERRORS
No errors.
SCOTT@orcl_11g> COLUMN from_section FORMAT A60 WORD_WRAPPED
SCOTT@orcl_11g> SELECT t.employee_id, t.fname, SCORE (1),
  2           from_section (employee_id, '.NET%', ROWID) AS from_section
  3  FROM   employee_tbl t
  4  WHERE  CONTAINS (t.dummy_column, '.NET%', 1) > 0
  5  ORDER  BY t.employee_id
  6  /
EMPLOYEE_ID FNAME     SCORE(1) FROM_SECTION
          1 Adam             3 JOBHISTORY
          2 Bob              3 SKILLS
          3 Charles          3 JOBHISTORY
          4 David            6 SKILLS,JOBHISTORY
SCOTT@orcl_11g>