In PostgreSQL i want to fetch every users at once and order them by date.
This is my query:
SELECT id, useridx, isread, message, date
FROM messages
WHERE isread = 1
GROUP BY useridx
ORDER BY date DESC
This is a sample data:
------------------------------------------------------
+ id | useridx | isread | messsage | date +
------------------------------------------------------
1 | 1 | 0 | Hello | 2012-01-01
2 | 2 | 1 | Hi | 2012-01-02
3 | 3 | 1 | Test | 2012-01-03
4 | 3 | 0 | My Msg | 2012-01-04
5 | 4 | 1 | sadasd | 2012-01-05
6 | 4 | 1 | sdfsdfd | 2012-01-06
7 | 4 | 0 | sdfsdfsd | 2012-01-07
8 | 5 | 0 | 5345634 | 2012-01-08
9 | 6 | 0 | sdfdfsd | 2012-01-09
10 | 7 | 0 | sdfsdfsf | 2012-01-10
------------------------------------------------------
Now, what i want to do is fetch this table by grouping them via useridx and order by date.
Expected Result:
------------------------------------------------------
+ id | useridx | isread | messsage | date +
------------------------------------------------------
6 | 4 | 1 | sdfsdfd | 2012-01-06
3 | 3 | 1 | Test | 2012-01-03
2 | 2 | 1 | Hi | 2012-01-02
------------------------------------------------------
Actual Result
ERROR: column "messages.date" must appear in the GROUP BY clause or be used in an aggregate function
I do not want to group date either. I just want to group with useridx and sort them by date DESC.
Any help/idea is appreciated!
Note: I also tried Distinct. Not fit my needs or i did wrongly.
I am very confused and stuck between DISTINCT ON
and rank()
methods.
Conclusion: For who get the same problem here can read this as an answer. Both @kgrittn's and @mu is too short's answers are correct. I will continue to use both answers and schemas on my project and in time i can understand which one is the best -i guess-. So, pick one of them and continue to your work. You will be just fine.
Last Update: Sometimes, Distinct On excludes some ids from result. Lets say i have a id column and i have 6 rows which is same. So, distinct on exlude it from the result BUT rank() just result it. So, use rank()!
PostgreSQL, unlike MySQL, does not show random data for columns which are not aggregated in an aggregated query.
The solution is in the error message
ERROR: column "messages.date" must appear in the GROUP BY clause or be used in an aggregate function
Which means you must GROUP BY the "messages.date" column or use an aggregate function like MIN() or MAX() when selection this column
Example:
SELECT MIN(id), useridx, isread, message, MAX(date)
FROM messages WHERE isread = 1
GROUP BY useridx, isread, message
ORDER BY MAX(date) DESC