Reputation: 359
I want to select the 2nd response column value of each new_threads group, with a zero as the value if it is a group of 1 row.
new_treads|response
------------------
1 | 0
1 | 1
2 | 0
2 | 0
2 | 1
... | ...
9 | 0
9 | 1
9 | 0
10 | 0
The output being:
new_treads|response
------------------
1 | 1
2 | 0
... | ...
9 | 1
10 | 0
So far, I understand how to get the first with MIN, but I need the 2nd
SELECT
thread,
min(response)
FROM messages
GROUP BY thread;
I would like to use GROUP BY because I'm using GROUP BY for other SELECTs as well
Thanks!
Upvotes: 0
Views: 2502
Reputation: 2156
I would like to elaborate on the answer above. While it worked well for me, it took some time to piece together the context and generalize it so I could apply it to my code. I hope that this answer will better generalize what is laid out above...
SELECT *
FROM (SELECT distinct keyField --- keyField is the field the query is grouping by
FROM TABLE
-- Add keyField Constraint --
-- Add non-keyField Constraint --
INNER JOIN (SELECT *,
@n:=(CASE -- Iterating through...
WHEN keyField = @prev_keyField -- When keyField value == previous keyField value
THEN @n:=@n+1 -- Define n as the row in the group
ELSE 1 -- When keyField value != previous keyField value, then n is the 1st row in the group
END) as n,
@prev_keyField:= keyField -- Define previous keyField value for the next iteration
FROM (SELECT @n:=0,@prev_keyField:=0) r,TABLE as p
-- Add non-keyField Constraint--
ORDER BY keyField,sortField DESC -- Order by keyField and the field you are sorting by
-- ei. keyField could be `thread`,
-- and sort field could be `timestamp` if you are sorting by time
) s ON s.keyField = p.keyField
WHERE s.n = 2 -- Define which row in the group you want in the query
Upvotes: 0
Reputation: 21047
Since the rows are not "numbered", you need to create a number for each group and then select it. I'd do that with user variables:
select thread, response
from (
select @n := (case
when m.thread = @prev_thread then @n
else 0
end) + 1 as n -- If the current thread is the same as the
-- previous row, then increase the counter,
-- else, reset it
, @prev_thread := m.thread as thread -- Update the value of
-- @prev_thread
, m.response
from
(select @n := 0, @prev_thread := 0) as init
-- The 'init' subquery initializes the
-- temp variables:
-- @n is a counter
-- @prev_thread is an identifier for the
-- previous thread id
, messages as m
order by m.thread -- You need to add a second column to order
-- each response (such as "response_id", or
-- something like that), otherwise the returned
-- row may be a random one
) as a
where n = 2; -- Select the rows from the subquery where the counter equals 2
The above works quite fine to find the 2nd row of each group, but only if there's one. So now: how to get a NULL
value if there isn't a second row?
The easiest way to do this would be to use a left join:
select t.thread, b.response
from (select distinct thread from messages) as t
left join (
-- Put the above query here
) as b on t.thread = b.thread;
Upvotes: 2
Reputation: 519
SELECT
thread,
min(response)
FROM messages
GROUP BY thread
HAVING response > min(response)
try this just want to know if it works
Upvotes: 0