如何根据最小日期行中的条件返回最大日期结果？

Question

I have a list of data like so:

ID  step               date
1   SECOND_ATTEMPT     03/19/2018
1   QC_READY           03/23/2018
1   QC_REJECTS         03/26/2018
2   SCHEDULED          02/01/2018
2   FINISHED           02/04/2018
3   SECOND_ATTEMPT     04/02/2018
3   QC_READY           04/03/2018
4   SECOND_ATTEMPT     01/15/2018
4   FINISHED           01/25/2018

My query, stripped down, looks like this.

select 

j.id,
wfh.step,
wfh.date

from Job j
join work_flow_history wfh on j.id = wfh.job_id

the wfh IDs are unique for each step in each job. So if viewing the step ID's it would look like this:

wfh_ID   step
001      SECOND_ATTEMPT
002      QC_READY
etc...

I would like to return all of the information in the row with the most recent date in it per ID when the step column contains SECOND_ATTEMPT and doesn't contain FINISHED . The correct result set would look like so:

ID   step              date
1    QC_REJECTS        03/26/2018
3    QC_READY          04/03/2018

Answer 1

This is a little weird as a requirement. Here is one method:

select t.*
from t
where not exists (select 1 from t t2 where t2.id = t.id and t2.step = 'FINISHED') and
      exists (select 1 from t t2 where t2.id = t.id and t2.step = 'SECOND_ATTEMPT') and
      t.date = (select max(t2.date) from t t2 where t2.id = t.id);

Another way with just one subquery:

select t.*
from t join
     (select t.id, max(t.date) as maxd
      from t
      group by t.id
      having sum(t.step = 'SECOND_ATTEMPT') > 0 and
             sum(t.step = 'FINISHED') = 0
     ) tt
     on tt.id = t.id and tt.maxd = t.date;

Answer 2

SELECT *
FROM
(
SELECT
j.id,
wfh.step,
wfh.date
FROM Job j
JOIN work_flow_history wfh on j.id = wfh.job_id
WHERE wfh.step NOT LIKE '%FINISHED%'
AND wfh.step LIKE '%SECOND_ATTEMPT%'
ORDER BY DATE_FORMAT(wfh.date, "%Y-%m-%d") DESC
) t
GROUP BY t.id

GROUP BY takes first row when grouping, subquery orders it beforehand. Since there won't be any SUM() or else impacted, no need for more complexity

Having your date in better format in database would avoid using date_format() though.

Answer 3

Thanks to the answers posted in previous comments I got some ideas and figured out how to make it work.

select
j.id
,sum(case when wfh.step = 'SECOND_ATTEMPT' then 1 else 0 end) sum_
,(select max(wfh1.date) 
  from work_flow_history wfh1
  join job j1 on j1.id = wfh1.job_id
  where j1.id = j.id
  order by wfh1.date desc
  limit 1) date_
,(select wfh1.step
  from work_flow_history wfh1
  join job j1 on j1.id = wfh1.job_id
  where j1.id = j.ie
  and wfh1.step <> 'FINISHED'
  order by wfh1.date desc
  limit 1) step_

from job j
join wofk_flow_history wfh on j.id = wfh.job_id
group by j.id

This query successfully looks at all wfh steps per job id and returns the max wfh step if SECOND_ATTEMPT is in any of the rows associated with that job id.

I know it's really messy. I'm not sure how to change the subquery columns into proper joins to make it look nicer, and I also need to add logic in order to only include instances where sum_ = 1. But I am very glad to have moved past this roadblock. Thanks everyone for your help!