Select a row of first non-null values - database

Question

I have the following table. How do I select the first non-null value of reviewers and voting column if they have the same product_id? The first here mean the first row sorting by created_at
+------------+-----------+--------+---------------------+
| product_id | reviewers | voting | created_at |
+------------+-----------+--------+---------------------+
| B0021ZFV9M | null | null | 2015-03-20 00:34:09 |
| B0021ZFV9M | 4 | 3 | 2015-03-24 00:34:09 |
| B0021ZFV9M | null | null | 2015-04-13 00:55:51 |
| B0021ZFV9M | 30 | 4 | 2015-04-15 00:44:38 |
| B00JKO4CHO | null | null | 2015-09-17 00:41:40 |
| B00JKO4CHO | null | null | 2015-09-19 00:41:47 |
| B00JKO4CHO | 50 | 1 | 2015-09-21 00:41:31 |
+------------+-----------+--------+---------------------+
Expected
+------------+-----------+--------+---------------------+
| product_id | reviewers | voting | created_at |
+------------+-----------+--------+---------------------+
| B0021ZFV9M | 4 | 3 | 2015-03-20 00:34:09 |
| B0021ZFV9M | 4 | 3 | 2015-03-24 00:34:09 |
| B0021ZFV9M | 30 | 4 | 2015-04-13 00:55:51 |
| B0021ZFV9M | 30 | 4 | 2015-04-15 00:44:38 |
| B00JKO4CHO | 50 | 1 | 2015-09-17 00:41:40 |
| B00JKO4CHO | 50 | 1 | 2015-09-19 00:41:47 |
| B00JKO4CHO | 50 | 1 | 2015-09-21 00:41:31 |
+------------+-----------+--------+---------------------+

Answer 1

Try this:
select
product_id,
case
when reviewers is null then (
select reviewers from test
where product_id = a.product_id
and created_at > a.created_at
and reviewers is not null
limit 1)
else reviewers
end as reviewers,
case
when voting is null then (
select voting from test
where product_id = a.product_id
and created_at > a.created_at
and voting is not null
limit 1)
else voting
end as voting,
created_at
from test a;
Example: http://sqlfiddle.com/#!9/546dff/3
create table test (
product_id varchar(20),
reviewers int,
voting int,
created_at datetime
);
insert into test values
('B0021ZFV9M',null , null ,'2015-03-20 00:34:09')
,('B0021ZFV9M',4 , 3 ,'2015-03-24 00:34:09')
,('B0021ZFV9M',null , null ,'2015-04-13 00:55:51')
,('B0021ZFV9M',30 , 4 ,'2015-04-15 00:44:38')
,('B00JKO4CHO',null , null ,'2015-09-17 00:41:40')
,('B00JKO4CHO',null , null ,'2015-09-19 00:41:47')
,('B00JKO4CHO',50 , 1 ,'2015-09-21 00:41:31');
Result:
| product_id | reviewers | voting | created_at |
|------------|-----------|--------|-----------------------------|
| B0021ZFV9M | 4 | 3 | March, 20 2015 00:34:09 |
| B0021ZFV9M | 4 | 3 | March, 24 2015 00:34:09 |
| B0021ZFV9M | 30 | 4 | April, 13 2015 00:55:51 |
| B0021ZFV9M | 30 | 4 | April, 15 2015 00:44:38 |
| B00JKO4CHO | 50 | 1 | September, 17 2015 00:41:40 |
| B00JKO4CHO | 50 | 1 | September, 19 2015 00:41:47 |
| B00JKO4CHO | 50 | 1 | September, 21 2015 00:41:31 |
EDIT:
To update old data, you could do this:
-- create a duplicate empty table
create table test1 like test;
-- insert good data into this duplicate table
insert into test1
select
product_id,
case
when reviewers is null then (
select reviewers from test
where product_id = a.product_id
and created_at > a.created_at
and reviewers is not null
limit 1)
else reviewers
end as reviewers,
case
when voting is null then (
select voting from test
where product_id = a.product_id
and created_at > a.created_at
and voting is not null
limit 1)
else voting
end as voting,
created_at
from test a;
-- remove data from original table
truncate table test;
-- re-insert good data into original table
insert into test select * from test1;
-- drop the duplicate table
drop table test1;
Make a backup of test (original) table before you try this.

Answer 2

select distinct on (a.product_id, a.created_at)
a.product_id,
coalesce(a.reviewers, b.reviewers) reviewers,
coalesce(a.voting, b.voting) voting,
a.created_at
from a_table a
left join a_table b
on a.product_id = b.product_id
and b.reviewers notnull
and b.created_at > a.created_at
order by 1, 4;
SqlFiddle.
Note: it is assumed that if reviewers is not null then voting is not null too.