RE: Greatest Value with column name - postgresql

Question

Greatest value of multiple columns with column name?
I was reading the question above (link above) and the "ACCEPTED" answer (which seems correct) and have several questions concerning this answer.
(Sorry I have to create a new post, don't have a high enough reputation to comment on the old post as it seems very old)
Questions
My first question is what is the significance of "#var_max_val:= "? I reran the query without it and everything ran fine.
My second question is can someone explain how this achieve it's desired result:
CASE #var_max_val WHEN col1 THEN 'col1'
WHEN col2 THEN 'col2'
...
END AS max_value_column_name
My third question is as follows:
It seems that in this "case" statement he manually has to write a line of code ("when x then y") for every column in the table. This is fine if you have 1-5 columns. But what if you had 10,000? How would you go about it?
PS: I might be violating some forum rules in this post, do let me know if I am.
Thank you for reading, and thank you for your time!

Answer 1

The linked question is about mysql so it does not apply to postgresql (e.g. the #var_max_val syntax is specific to mysql). To accomplish the same thing in postgresql you can use a LATERAL subquery. For example, suppose that you have the following table and sample data:
CREATE TABLE t(col1 int, col2 int, col3 int);
INSERT INTO t VALUES (1,2,3), (5,8,6);
Then you can identify the maximum column for each row with the following query:
SELECT *
FROM t, LATERAL (
VALUES ('col1',col1),('col2',col2),('col3',col3)
ORDER BY 2 DESC
LIMIT 1
) l(maxcolname, maxcolval);
which produces the following output:
col1 | col2 | col3 | maxcolname | maxcolval
------+------+------+------------+-----------
1 | 2 | 3 | col3 | 3
5 | 8 | 6 | col2 | 8
I think this solution is much more elegant than the one presented in the linked article for mysql.
As for having to manually write the code, unfortunately, I do not think you can avoid that.

Answer 2

In Postgres 9.5 you can use jsonb functions to get column names. In this case you do not have to write manually all the columns names. The solution needs a primary key (or a unique column) for proper
grouping:
create table a_table(id serial primary key, col1 int, col2 int, col3 int);
insert into a_table (col1, col2, col3) values (1,2,3), (5,8,6);
select distinct on(id) id, key, value
from a_table t, jsonb_each(to_jsonb(t))
where key <> 'id'
order by id, value desc;
id | key | value
----+------+-------
1 | col3 | 3
2 | col2 | 8
(2 rows)