How to complare one Master table against several child tables - sql

Question

I have one master table (PO_BreakOutAll) with ~3000 rows made up of only two columns (PO_ID, PO_LN_NO) which together make up the primary key. I also have several other tables each with a subset of the data from the master table (or so it is supposed to be). All the tables are the same schema as the master table.
All tables have this exact schema:
PO_ID char(5) PK
PO_LN_NO int PK
I need to do two different types of comparisons for validation and to find duplicates.
First to make sure that every row in the master table exists in one, and only one, of the other child tables.
Second I need to make sure that no row is duplicated across any of the child tables. The same row can exist in two or more child tables and I need to find them.
I can do each table in a separate query but have not figured out how to write one query that compares all the child tables at once.
Here is what I have so far but it does not work:
SELECT a.PO_ID as all_PO,
a.PO_LN_NO,
c.PO_ID as Cummings_PO,
c.PO_LN_NO,
f.PO_ID as filter_PO,
f.PO_LN_NO,
fo.PO_ID as fixedObl_PO,
fo.PO_LN_NO
FROM
PO_BreakOutAll a
LEFT OUTER JOIN
PO_Cummins c ON (c.PO_ID = a.PO_ID AND c.PO_LN_NO = a.PO_LN_NO)
LEFT OUTER JOIN
PO_Filters f ON (f.PO_ID = a.PO_ID AND f.PO_LN_NO = a.PO_LN_NO)
LEFT OUTER JOIN
PO_FixedOblig fo ON (fo.PO_ID = a.PO_ID AND fo.PO_LN_NO = a.PO_LN_NO)

Answer 1

I think #gordon linoff has an overall solution. If you want to work with the CTE paradigm here is an example based on your Fiddle that answers the duplicate question:
WITH CTE (PO_ID,PO_LN_NO,TableName) AS
(SELECT
PO_ID,
PO_LN_NO,
'Cummings' as TableName
FROM PO_Cummins
UNION ALL
SELECT
PO_ID,
PO_LN_NO,
'Filters' as TableName
FROM PO_Filters
UNION ALL
SELECT
PO_ID,
PO_LN_NO,
'Office' as TableName
FROM PO_Office )
SELECT
PO_BreakOutAll.PO_ID,
PO_BreakOutAll.PO_LN_NO,
CHILD_DATA.TABLENAME AS DUP_TABLENAME
FROM
PO_BreakOutAll
INNER JOIN (
SELECT PO_ID, PO_LN_NO, COUNT(1) AS DUP_COUNTER
FROM CTE
GROUP BY PO_ID, PO_LN_NO
HAVING COUNT(1) > 1
) DUPS ON DUPS.PO_ID = PO_BreakOutAll.PO_ID AND DUPS.PO_LN_NO = PO_BreakOutAll.PO_LN_NO
INNER JOIN (
SELECT PO_ID, PO_LN_NO, TABLENAME
FROM CTE
) CHILD_DATA
ON CHILD_DATA.PO_ID = PO_BreakOutAll.PO_ID AND CHILD_DATA.PO_LN_NO = PO_BreakOutAll.PO_LN_NO
ORDER BY PO_ID, PO_LN_NO, DUP_TABLENAME

Answer 2

I wouldn't use join for this; I would use union all. Here is a way to get a count of how the records overlap among the tables:
select isAll, isCummins, isFilters, isOblig, count(*)
from (select PO_ID, PO_LN_NO, sum(isAll) as isAll, sum(isCummins) as isCummins,
sum(isFilters) as isFilters, sum(isOblig) as isOblig
from ((select PO_ID, PO_LN_NO, 1 as isAll, 0 as isCummins, 0 as isFilters, 1 as isOblig
from PO_BreakOutAll
) union all
(select PO_ID, PO_LN_NO, 0, 1, 0, 0
from PO_Cummins
) union all
(select PO_ID, PO_LN_NO, 0, 0, 1, 0
from PO_Filters
) union all
(select PO_ID, PO_LN_NO, 0, 0, 0, 1
from PO_FixedOblig
)
) t
group by PO_ID, PO_LN_NO
) t
group by isAll, isCummins, isFilters, isOblig;
If you want to find rows that fail your test, just use the subquery with where conditions:
select PO_ID, PO_LN_NO, sum(isAll) as isAll, sum(isCummins) as isCummins,
sum(isFilters) as isFilters, sum(isOblig) as isOblig
from ((select PO_ID, PO_LN_NO, 1 as isAll, 0 as isCummins, 0 as isFilters, 1 as isOblig
from PO_BreakOutAll
) union all
(select PO_ID, PO_LN_NO, 0, 1, 0, 0
from PO_Cummins
) union all
(select PO_ID, PO_LN_NO, 0, 0, 1, 0
from PO_Filters
) union all
(select PO_ID, PO_LN_NO, 0, 0, 0, 1
from PO_FixedOblig
)
) t
group by PO_ID, PO_LN_NO
having sum(isAll) <> 1 or
(sum(isAll) = 1 and (sum(isCummins) + sum(isFilters) + sum(isOblig) <> 1)
);