How to select for all cases within a group? - sql

Question

I’d like to get metrics on the number of cases in my table where my data fits different criteria. Currently, I have groups of 3+ that we’d like to inspect based on certain criteria. For the time being, I’d like to get counts of the following:
The breakdown of records in a group of 3+ where CHESS = White.
The breakdown of records in a group of 3+ where DATE like ‘%16’
The breakdown of records in a group of 3+ with all combinations of chess = white and DATE like ‘%16’
+------------------------------------------------+
| GroupID | CHESS | DATE |
| 1 | White | n16 |
| 1 | Black | n16 |
| 1 | Black | n03 |
| 2 | White | n16 |
| 2 | White | n10 |
| 2 | White | n11 |
| 3 | Black | n12 |
| 3 | White | n14 |
| 3 | Black | n16 |
---------------------------------------------------
The output would be something like:
Chess count of 1 = 2
Chess count of 2 = 0
Chess count of 3 = 1
Date count of 1 = 2
Date count of 2 = 1
Date count of 3 = 0
Cases with Chess count of 1 and Date count of 1 = 1
Cases with Chess count of 2 and Date count of 2 = 0
Cases with Chess count of 3 and Date count of 3= 0
Cases with Chess count of 1 and Date of count 2 = 1
Cases with Chess count of 1 and Date of count 3 = 0
Cases with Chess count of 2 and Date of count 1 = 0
Cases with Chess count of 2 and Date of count 3= 0
Cases with Chess count of 3 and Date of count 1 = 1
Cases with Chess count of 3 and Date of count 2 = 0
Can this be done in a way that takes into account groups of any sizes, or would it have to be specific to the group size (for example would the query only work on groups of 3)?

Answer 1

You could do it like here, using group by cube and others:
with groups as (select count(case when chess = 'White' then 1 end) cnt_white,
count(case when tdate = 'n16' then 1 end) cnt_n16
from t group by groupid),
numbers as (select level grp from dual connect by level <= 3)
select white, n16, cnt
from (
select n1.grp white, n2.grp n16, count(cnt_white) cnt, grouping_id(n1.grp, n2.grp) gid
from numbers n1
cross join numbers n2
left join groups on n1.grp = cnt_white and n2.grp = cnt_n16
group by cube(n1.grp, n2.grp)
having grouping_id(n1.grp, n2.grp) <> 3 )
order by case when gid = 1 then 1 when gid = 2 then 2 when gid = 0 then 3 end, white, n16
rextester demo
Modify level 3 in subquery numbers to change number of groups.