I'm using the following query to create a view in Oracle 11g (11.2.0.3.0).
CREATE OR REPLACE FORCE VIEW V_DOCUMENTS_LIST
(
ID_DOC,
ATTACHMENTS_COUNT,
TOTAL_DIMENSION,
INSERT_DATE,
ID_STATE,
STATE,
ID_INSTITUTE,
INSTITUTE,
HASJOB
)
AS
SELECT D.ID_DOC,
COUNT (F.ID_FILE) AS ATTACHMENTS_COUNT,
CASE
WHEN SUM (F.DIMENSION) IS NULL THEN 0
ELSE SUM (F.DIMENSION)
END
AS TOTAL_DIMENSION,
D.INSERT_DATE,
D.ID_STATE,
S.STATE_DESC AS STATE,
D.ID_INSTITUTE,
E.NAME AS INSTITUTE,
CASE
WHEN EXISTS (SELECT D.ID_DOC FROM JOB) THEN 'true'
ELSE 'false'
END
AS HASJOB
FROM DOCUMENTS D
LEFT JOIN FILES F ON D.ID_DOC = F.ID_DOC
JOIN STATES S ON D.ID_STATE = S.ID_STATE
JOIN INSTITUTES E ON D.ID_INSTITUTE = E.ID_INSTITUTE
GROUP BY D.ID_DOC,
D.INSERT_DATE,
D.ID_STATE,
S.STATE_DESC,
D.ID_INSTITUTE,
E.NAME;
Then I query that view to get the values for a DataGridView
in an ASPX
page.
SELECT *
FROM V_DOCUMENTS_LIST
ORDER BY ID_STATE DESC, INSTITUTE, INSERT_DATE DESC;
Relevant tables and relations
DOCUMENTS; FILES; JOBS;
DOCUMENTS (1-1) <----> (0-N) FILES
JOBS (0-1) <----> (0-N) DOCUMENTS
Querying the view I get the complete list of documents with all their associated information (ID, description, dates, state, etc.) and also for each one:
JOB
associated to the DOCUMENT
or not.Everything worked fine untile the view contained a few thousand records. Now the records amount is increasing and the SELECT * FROM
on the view takes about 2:30 mins with 15.000-20.000 records.
I know that a really time consuming part of my view is the nested SELECT
:
CASE
WHEN EXISTS (SELECT D.ID_DOC FROM JOB) THEN 'true'
ELSE 'false'
END
AS HASJOB
How can I optimize my view?
To address the not exists issue, you can add a join:
LEFT JOIN (select distinct id_doc from JOB) J
ON d.id_doc = J.id_doc
The Has_job column would be:
CASE
WHEN j.id_doc is not null THEN 'true'
ELSE 'false'
END AS HASJOB
PS: Your current implementation has a problem, as SELECT D.ID_DOC FROM JOB
would allways contain rows if job table has rows. It is equivalent with select * from job
, because exists just test existence of rows. A logically correct implementation would be: SELECT 1 FROM JOB j where j.id_doc = D.ID_DOC
.