SQL group by count across two tables - sql

Question

I have got two tables called baseline and revisits
baseline
formid-------NoOfIssues
1--------------3
2--------------4
3--------------5
revisits
id------formid-------NoOfIssues-----------date--------------fid
1---------2--------------4-------------5/06/2016------------1
2---------3--------------3-------------15/06/2016-----------1
3---------1--------------4-------------20/07/2016-----------1
4---------1--------------3-------------25/07/2016-----------1
5---------2--------------5-------------28/07/2016-----------1
6---------1--------------5-------------01/06/2016-----------1
7---------3--------------8-------------21/02/2016-----------1
8---------3--------------2-------------21/02/2016-----------2
These tables are joined by 'formid'.
I need to compare number of issues in baseline vs revisits(only first) and get a count as reduced, increased or equal
Based upon the above table i am expecting the following, for example across all three baseline entries no equals were found comparing NoOfissues in first revisit against same formid, but 1 equal and 2 increased were found
Addition: if same date and same formid is found than take the lower fid, so in the last two rows of revisits table both formid and date are equal but need to consider the lower formid which is 1
status----------Count
reduced----------0
equal------------1
increased--------2

Answer 1

I'm not familiar with intersystems-cache, but you can see if the following is valid SQL with that DB:
SELECT
CASE
WHEN BL.NoOfIssues = FR.NoOfIssues THEN 'equal'
WHEN BL.NoOfIssues > FR.NoOfIssues THEN 'reduced'
WHEN BL.NoOfIssues < FR.NoOfIssues THEN 'increased'
END AS status,
COUNT(*) AS Count
FROM
Baseline BL
INNER JOIN Revisits FR ON FR.formid = BL.formid
LEFT OUTER JOIN Revisits R ON
R.formid = BL.formid AND
(
R.date < FR.date OR
(R.date = FR.date AND R.fid > FR.fid)
)
WHERE
R.formid IS NULL
GROUP BY
CASE
WHEN BL.NoOfIssues = FR.NoOfIssues THEN 'equal'
WHEN BL.NoOfIssues > FR.NoOfIssues THEN 'reduced'
WHEN BL.NoOfIssues < FR.NoOfIssues THEN 'increased'
END
Some quick notes on your database though - You should probably decide on a standard of plural or singular table names and stick with it. Also, try to avoid common reserved words for object names, like date. Finally, if a revisit is basically the same as a visit, just on a later date then you should consider keeping them all in the same table.

Answer 2

I would do this in one row rather than three:
select sum(case when numissues = rcnt then 1 else 0 end) as equal,
sum(case when numissues > rcnt then 1 else 0 end) as reduced,
sum(case when numissues < rcnt then 1 else 0 end) as incrased
from (select b.form_id, b.numissues, count(r.form_id) as rcnt
from baseline b left join
revisits r
on b.form_id = r.form_id
group by b.form_id, b.numissues
) br;

Answer 3

I would use a window function to get the number of ussues at the minimum date, then compare that to the baseline number of issues
select
case when baseline.NoOfIssues = rev.NoOfIssues then 'equal'
when baseline.NoOfIssues > rev.NoOfIssues then 'reduced'
when baseline.NoOfIssues < rev.NoOfIssues then 'increased'
end as status,
count(*) as count
from baseline
inner join(
select
formid,
case when date = min(date) over(partition by formid) then NoOfIssues else null end as first_rev_issues
from revisits
) rev
on baseline.formid = rev.formid
group by
case when baseline.NoOfIssues = rev.NoOfIssues then 'equal'
when baseline.NoOfIssues > rev.NoOfIssues then 'reduced'
when baseline.NoOfIssues < rev.NoOfIssues then 'increased'
end

Answer 4

Or like this:
WITH CTE AS
(
SELECT
BL.FORMID,
BL.NOOFISSUES AS BLI,
RV.NOOFISSUES AS RVI,
RV.DATE,
ROW_NUMBER() OVER(PARTITION BY BL.FORMID ORDER BY RV.DATE) AS RN
FROM Baseline BL
INNER JOIN Revisits RV ON RV.FORMID = BL.FORMID
)
SELECT COALESCE(SUM(CASE WHEN RVI > BLI THEN 1 END), 0) AS INCREASED,
COALESCE(SUM(CASE WHEN RVI < BLI THEN 1 END), 0) AS DECREASED,
COALESCE(SUM(CASE WHEN RVI = BLI THEN 1 END), 0) AS EQUAL
FROM CTE
WHERE RN=1;