how to retrieve latest data from a table

Question

Currently I am working on a project where I need to extract latest data from a table for a report purpose. Below is the sample table structure:-

Every student has several courses and course_id for programming language is +ve and for non programming language -ve. I want to extract latest programming language and non programming language course_id for each and every student.

I use the below SQL query and able to extract the data.

CREATE TABLE COURSE
    ("STUDENT_ID" int, "COURSE_ID" int, "COURSE_NAME" varchar2(31), "COURSE_START_DATE" timestamp)
;

INSERT ALL 
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100001, -100, 'C Programming Language', '04-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100001, -200, 'Java Programming Language', '11-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100001, -300, 'C# Programming Language', '07-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100001, 100, 'Data Structure and algorithms', '05-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100001, 200, 'Computer Graphics', '13-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100001, 300, 'Networking', '02-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100002, -300, 'C# Programming Language', '12-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100002, -400, 'Python Programming Language', '07-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100002, -500, 'JavaScript Programming Language', '08-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100002, 100, 'Data Structure and algorithms', '17-Jan-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100002, 300, 'Computer Graphics', '26-Jan-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100002, 400, 'DataBase Management', '10-Jan-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100003, -500, 'JavaScript Programming Language', '07-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100003, -600, 'SQL', '13-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100003, -200, 'Java Programming Language', '17-Jan-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100003, 300, 'Networking', '04-Feb-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100003, 400, 'DataBase Management', '05-Jan-2019 12:00:00 AM')
    INTO COURSE ("STUDENT_ID", "COURSE_ID", "COURSE_NAME", "COURSE_START_DATE")
         VALUES (100003, 600, 'Cryptography', '18-Jan-2019 12:00:00 AM')
SELECT * FROM dual
;

SELECT STUDENT_ID
,COURSE_ID
,COURSE_NAME
,COURSE_START_DATE
  FROM (
SELECT 
ROW_NUMBER() OVER(PARTITION BY STUDENT_ID ORDER BY COURSE_START_DATE DESC) AS ROW_NUM
,STUDENT_ID
,COURSE_ID
,COURSE_NAME
,COURSE_START_DATE
  FROM
COURSE
WHERE COURSE_ID  0) TEMP1 WHERE TEMP1.ROW_NUM = 1;

But the problem is the real table is very large. There are almost 85k rows and this query is taking sometime. Is there any other better approach. I am using Oracle 11g R2. Please suggest

This is SQLfiddle link http://sqlfiddle.com/#!4/b3fe1/8

Answer 1

Another way to get those results is by using a NOT EXISTS

Because if you want the record with the lastest COURSE_START_DATE for a student.
Then for that record, there won't exists any with a date that's higher.
(unless there's 2 with the same max date, then it would return both)

SELECT 
 STUDENT_ID, 
 COURSE_ID, 
 COURSE_NAME, 
 COURSE_START_DATE
FROM COURSE t
WHERE COURSE_ID != 0
AND NOT EXISTS
(
  SELECT 1
  FROM COURSE d
  WHERE d.STUDENT_ID = t.STUDENT_ID
    AND d.COURSE_START_DATE > t.COURSE_START_DATE
    AND SIGN(d.COURSE_ID) = SIGN(t.COURSE_ID)
    AND d.COURSE_ID != 0
)
ORDER BY SIGN(COURSE_ID), STUDENT_ID

Such query would probably benefit from a non-unique index on STUDENT_ID.

A test on db<>fiddle here

By the way, in Oracle 12c you could order by ROW_NUMBER, then fetch only the first with ties.

SELECT 
 STUDENT_ID, 
 COURSE_ID, 
 COURSE_NAME, 
 COURSE_START_DATE
FROM COURSE t
WHERE COURSE_ID != 0
ORDER BY row_number() over(partition by student_id, SIGN(COURSE_ID) order by course_start_date desc)
FETCH FIRST ROW WITH TIES

Answer 2

You can use your query and add SIGN( course_id ) to the partition:

Query:

SELECT STUDENT_ID
     , COURSE_ID
     , COURSE_NAME
     , COURSE_START_DATE
FROM (
  SELECT ROW_NUMBER() OVER (
           PARTITION BY STUDENT_ID, SIGN( COURSE_ID )
           ORDER BY COURSE_START_DATE DESC
         ) AS ROW_NUM
       , STUDENT_ID
       , COURSE_ID
       , COURSE_NAME
       , COURSE_START_DATE
  FROM   COURSE
  WHERE COURSE_ID != 0
)
WHERE  ROW_NUM = 1;

Output:

STUDENT_ID | COURSE_ID | COURSE_NAME               | COURSE_START_DATE           
---------: | --------: | :------------------------ | :---------------------------
    100001 |      -200 | Java Programming Language | 11-FEB-19 12.00.00.000000 AM
    100001 |       200 | Computer Graphics         | 13-FEB-19 12.00.00.000000 AM
    100002 |      -300 | C# Programming Language   | 12-FEB-19 12.00.00.000000 AM
    100002 |       300 | Computer Graphics         | 26-JAN-19 12.00.00.000000 AM
    100003 |      -600 | SQL                       | 13-FEB-19 12.00.00.000000 AM
    100003 |       300 | Networking                | 04-FEB-19 12.00.00.000000 AM

db<>fiddle here

Answer 3

Use CTEs with row_number, then union them

with pro as 
(
select t1.*, row_number() over(partition by student_id order by course_start_date desc) rn
from course 
where course_id > 0 -- programming
)
, nonpro as 
(
select t1.*, row_number() over(partition by student_id order by course_start_date desc) rn
from course 
where course_id < 0 -- non-programming
)
select *
from pro
where rn = 1
union
select *
from nonpro
where rn = 1

Answer 4

You can try below - you need to add PARTITION BY STUDENT_ID,case when course_id<0 then 1 else 2 end order by COURSE_START_DATE DESC in over cluase

SELECT STUDENT_ID,COURSE_ID,COURSE_NAME,COURSE_START_DATE
FROM 
(
  SELECT 
  ROW_NUMBER() OVER(PARTITION BY STUDENT_ID,case when course_id<0 then 1 else 2 end order by COURSE_START_DATE DESC) AS ROW_NUM
, STUDENT_ID,COURSE_ID,COURSE_NAME,COURSE_START_DATE FROM COURSE
 )TEMP1 WHERE TEMP1.ROW_NUM = 1;