Join two tables and remove duplicates - mysql

Question

I'm trying to join two tables. Where table2 has duplates.
The tables look something like
CREATE TABLE ta
(
id int,
cno varchar(30),
d1 varchar(30),
d2 int
);
CREATE TABLE tb
(
id int,
cno varchar(30),
cn1 varchar(30),
cn2 int
);
INSERT INTO ta
(id, cno, d1, d2)
VALUES
(1, '1234','a',2),
(2, '6456','j',3),
(3, '5456','h',4),
(4, '4454','g',5);
INSERT INTO tb
(id, cno, cn1, cn2)
VALUES
(1, '1234', 'a', 21),
(1, '1234', 'a', 22),
(2, '6456', 'b', 33),
(2, '6456', 'c', 34),
(2, '6456', 'c', 35),
(3, '5456', 'c', 36),
(4, '4454', 'c', 37);
I was able to get the result http://sqlfiddle.com/#!2/b282e3/1 in MySQL. However when I run it in Postgresql I get an error http://sqlfiddle.com/#!15/b282e/4
Output should be like http://sqlfiddle.com/#!2/b282e3/1
CNO CN1 CN2 D1 D2
1234 a 21 a 2
4454 c 37 g 5
5456 c 36 h 4
6456 b 33 j 3
Any alternatives for this in Psql?

Answer 1

Use aggregate functions for columns that are not used in GROUP BY:
select t2.cno,
min(t2.cn1) as a,
min(t2.cn2) as b,
min(t1.d1) as c,
min(t1.d2) as d
from ta as t1
inner join tb as t2
on t1.cno=t2.cno
group by t2.cno
http://sqlfiddle.com/#!15/b282e/23

Answer 2

This query in MySQL:
select t2.cno, t2.cn1, t2.cn2, t1.d1, t1.d2
from ta t1 inner join
tb t2
on t1.cno = t2.cno
group by t2.cno;
Is not valid SQL (according to the standard or other databases). The problem is that there are columns in the select that are neither in the group by nor are they arguments to aggregation functions (and they are not "functionally dependent" either). Your use of the group by extension in MySQL is officially discouraged. You can read the documentation about it here.
Ironically, Postgres has an extension called distinct on that does something similar. The syntax is:
select distinct on (t2.cno) t2.cno, t2.cn1, t2.cn2, t1.d1, t1.d2
from ta t1 inner join
tb t2
on t1.cno = t2.cno
order by t2.cno;
distinct on takes a list in parentheses and returns one row per value in the parentheses -- taking the first row and ignoring the rest. These columns need to match the columns in the order by, otherwise Postgres generates a compile-time error.
In most other databases, you would do something similar using row_number(). And you can use that as well in Postgres.

Answer 3

select t2.cno, min (t2.cn1), min(t2.cn2), t1.d1 , t1.d2
from ta as t1
inner join tb as t2 on t1.cno=t2.cno
group by t2.cno, t1.d1 , t1.d2

Answer 4

WITH Queries (Common Table Expressions)
with cte as
(
select cno,cn1,cn2 from tb where cn2 in (select min(cn2) from tb group by cno)
),
cte1 as
(
select d1,d2,cno from ta where cno in (select cno from tb where cn2 in (select
min(cn2) from tb group by cno))
)
select cte.cno,cn1,cn2,d1,d2 from cte inner join cte1 on cte1.cno = cte.cno order
by cte.cno