Problem with SQL

Question

I am a java developer and having problem writing a simple sql query. Any help is highly appreciated.
FILE_TBL
CREATE TABLE FILE_TBL
FILE_ID VARCHAR2(8 BYTE),
PUBLICATION_DATE       DATE,
LOCATION_CODE VARCHAR2(2 BYTE),
Sample Data
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('1', '2011-04-11', 'AF001');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('2', '2011-04-11', 'SF057');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('2', '2011-04-14', 'AF081');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('3', '2011-04-11', 'SF022');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('3', '2011-04-14', 'AF012');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('4', '2011-04-11', 'SF057');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('5', '2011-04-11', 'AF036');
This is the structure of my table, from where data needs to be retrieved.
Now I have a requirement to retrieve the file_ids for all the matching files whose publication date is either or after 04/11/2011. Everything works fine with the below query.
SELECT FILE_ID
FROM FILE_TBL
WHERE PUBLICATION_DATE >= TO_DATE ('2011-04-11', 'yyyy-mm-dd')
AND ( (LOCATION_CODE LIKE 'SF%') OR (LOCATION_CODE LIKE 'AF%'))
Now there is a specific requirement, where in if there are multiple records for a particular file id with dates either or after 04/11/2011, then those records should be filtered.
For eg: in the above data set, for file id 2 and 3, I have two records with publication dates '04/11/2011' and '04/14/2011'. In this case since there are more entries than one record, this shouldn't be displayed and needs to be filtered from the result set.
Any help on correcting the query would be highly appreciated.
Thanks.

Answer 1

Hi,
SirGeneral wrote:
Thanks for the response!
Apologize for not testing the queries before posting them and this will not repeat. It was an error caused while simplifying the case. Thanks. They're working great now.
What are the results you want with this new requirement?
I also have one more requirement, and am not sure if I can do that without looping the data. What does "looping the data" mean?
Basically file data comes in as a batch with publication_dates for various file ids. In each such batch, the publication_date would be the same for all the different file ids. So, when a new batch comes in, all the files from old batch needs to be ignored.
Say with the corrected data set published:
First Batch:
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('1', TO_DATE('4-11-2011', 'MM-dd-YYYY'), 'AF001');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('2', TO_DATE('4-11-2011', 'MM-dd-YYYY'), 'SF057');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('3', TO_DATE('4-11-2011', 'MM-dd-YYYY'), 'SF022');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('4', TO_DATE('4-11-2011', 'MM-dd-YYYY'), 'SF057');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('5', TO_DATE('4-11-2011', 'MM-dd-YYYY'), 'AF036');
Second Batch:
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('2', TO_DATE('4-14-2011', 'MM-dd-YYYY'), 'AF081');
INSERT INTO FILE_TBL (FILE_ID, PUBLICATION_DATE, LOCATION_CODE) VALUES ('3', TO_DATE('4-14-2011', 'MM-dd-YYYY'), 'AF012');
Corrected create stmt:
CREATE TABLE FILE_TBL
FILE_ID VARCHAR2 (8 BYTE),
PUBLICATION_DATE DATE,
LOCATION_CODE VARCHAR2 (5 BYTE)
Now as per the new requirement I have to retrieve the file_ids for all the matching files whose publication date is either or after todays date (04/11/2011 ~ changes depending on when its being executed) only if a new batch hasn't arrived. If a new batch has arrived, retrieve only the files from new batch.
One solution that I can think about is to retrieve the data in descending order of PUBLICATION_DATE > TODAY and loop through by collecting file ids. If the PUBLICATION_DATE changes, then stop looping and return the collected file ids. You can use ROW_NUMBER, or similar analytic functions, to do that in SQL; no user-defined function or PL/SQL is required.
Is there a way to write this in a single SQL query? Appreciate any help on this.I'm not sure I understand.
It's clear you don't want to hard-code April 11, 2011 in the query, but I'm not sure what you want in its place. The query below uses today's date, whatever that happens to be.
Are you saying that only the most recent publication_date counts for each file_id? That is, file_ids 1, 4 and 5 only have rows dated April 11, so the April 11 rows count for those file_ids, but file_ids 2 and 3 have a later publication_date as well, so only the later date counts for them, so the rows dated April 11 are ignored for file_ids 2 and 3?
If so:
WITH     got_rnk         AS
     SELECT       file_id
     ,       COUNT (*)          AS cnt
     ,       ROW_NUMBER () OVER ( PARTITION BY  file_id
                                  ORDER BY          publication_date     DESC
                         )  AS rnk
        FROM        file_tbl
     WHERE       publication_date          >= TRUNC (SYSDATE)
     AND       SUBSTR (location_code, 1, 2)     IN ('AF', 'SF')
     GROUP BY  file_id
     ,            publication_date
SELECT     file_id
FROM     got_rnk
WHERE     rnk     = 1
AND     cnt     = 1
;With the sample data you posted, all 5 file_ids are selected, assuming this is run on April 11, 2011. (No file_id has more than 1 row on any one date.)
Or are you saying that all rows dated April 11 are to be ignored if there is any row with a later date?
In that case:
SELECT       file_id
FROM        file_tbl
WHERE       publication_date     = (
                         SELECT  MAX (publication_date)
                         FROM     file_tbl
                         WHERE     publication_date >= TRUNC (SYSDATE)
AND       SUBSTR (location_code, 1, 2)     IN ('AF', 'SF')
GROUP BY  file_id
HAVING       COUNT (*)     = 1
;The output from this query will include only file_ids 2 and 3. Here, April 14 is the latest date in the table, so all rows from April 11, from any file_id, are ignored.
It's unclear if you want the condition
SUBSTR (location_code, 1, 2)     IN ('AF', 'SF')to be in the main query (as shown above), the sub-query, or both.
Edited by: Frank Kulash on Apr 11, 2011 8:42 PM