SQL Server query to find multiple rows that exists in another table - sql-server

Question

I'm having trouble figuring out how to write a query that checks if two tables have any overlapping rows, given a certain criteria. The main issue I have is that the number of rows that need to match can vary, sometimes it may be one row, sometimes it's several.
As an example, let's say I have two tables. I want to find which Parents are in both tables. For a parent to be in both tables, the tables need to have the same number of children, and the children should have the same names and ages. Each parent is identified by either a name or a number.
CREATE TABLE ParentsToSearch
(
ParentID INT NOT NULL,
ChildName NVARCHAR(800) NOT NULL,
ChildAge INT
);
CREATE TABLE ExistingParentsAndChildren
(
ParentName NVARCHAR(800) NOT NULL,
ChildName NVARCHAR(800) NOT NULL,
ChildAge INT
);
If I had the following sample data, I would want the query to return that ParentID 7 exists in the table ExistingParentAndChildren (as ParentID 7's children and their ages is a perfect match with John's children)
ParentsToSearch
ParentID
ChildName
ChildAge
7
Katie
17
7
Jacob
8
12
Robert
10
ExistingParentAndChildren
ParentName
ChildName
ChildAge
John
Katie
17
John
Jacob
8
Sue
Robert
5
Sue
Carter
14
Sue
Ralph
10
Alex
Rocky
12
I assume I need to use something like ALL or PIVOT? But I'm kind of lost, as I'm new to DB queries.

Answer 1

A simple join on the ChildName and ChildAge columns gets you close.
But then there is the possibility that the children of parent Y are a subset of the children with parent X (i.e. parent X has children x1 (5) and x2 (3) and parent Y also has child x1 (5), then the children from parent Y are a subset of the children from parent X).
If you would extend the available parent data with the count of his children in each table and matched on that number as well, then you would have a full match.
Adding the required counts can be done with a cross apply (a subquery that is executed for each row) or common table expressions (returns a table you can join with).
Sample data
CREATE TABLE ParentsToSearch
(
ParentID INT NOT NULL,
ChildName NVARCHAR(800) NOT NULL,
ChildAge INT
);
insert into ParentsToSearch (ParentId, ChildName, ChildAge) values
( 7, 'Katie', 17),
( 7, 'Jacob', 8),
(12, 'Robert', 10);
CREATE TABLE ExistingParentsAndChildren
(
ParentName NVARCHAR(800) NOT NULL,
ChildName NVARCHAR(800) NOT NULL,
ChildAge INT
);
insert into ExistingParentsAndChildren (ParentName, ChildName, ChildAge) values
('John', 'Katie', 17),
('John', 'Jacob', 8),
('Sue', 'Robert', 5),
('Sue', 'Carter', 14),
('Sue', 'Ralph', 10),
('Alex', 'Rocky', 12);
Solution
With cross apply.
select p.ParentId,
ep.ParentName
from ParentsToSearch p
cross apply ( select count(1) as ChildCount
from ParentsToSearch p2
where p2.ParentId = p.ParentId ) pc
join ExistingParentsAndChildren ep
on ep.ChildName = p.ChildName
and ep.ChildAge = p.ChildAge
cross apply ( select count(1) as ChildCount
from ExistingParentsAndChildren ep2
where ep2.ParentName = ep.ParentName ) epc
where pc.ChildCount = epc.ChildCount -- match the child counts
group by p.ParentId,
ep.ParentName;
With common table expressions.
with cte_pc as
(
select p2.ParentId,
count(1) as ChildCount
from ParentsToSearch p2
group by p2.ParentId
),
cte_epc as
(
select ep2.ParentName,
count(1) as ChildCount
from ExistingParentsAndChildren ep2
group by ep2.ParentName
)
select p.ParentId,
ep.ParentName
from ParentsToSearch p
join cte_pc p3
on p3.ParentId = p.ParentId
join ExistingParentsAndChildren ep
on ep.ChildName = p.ChildName
and ep.ChildAge = p.ChildAge
join cte_epc ep3
on ep3.ParentName = ep.ParentName
where p3.ChildCount = ep3.ChildCount -- match the child counts
group by p.ParentId,
ep.ParentName;
Result
ParentId ParentName
-------- ----------
7 John
Fiddle to see things in action (also includes some alternative queries to list the children etc.).

Answer 2

If using SQLServer > 2016 then I would just do it like you would think about it without thinking about coding first. Breaking down dataset using "with" statements allows the solution to flow naturally.
Solution:
Group all children under their parents for both tables using the STRING_AGG function and with statements
with ParentNameFamilies as
(
SELECT ParentName, STRING_AGG(CONVERT(NVARCHAR(max), ChildName), ',') AS ChildNames
, STRING_AGG(CONVERT(NVARCHAR(max), ChildAge), ',') AS ChildAges
FROM dbo.ExistingParentsAndChildren
group by ParentName
), ParentIDFamilies as
(
SELECT ParentID, STRING_AGG(CONVERT(NVARCHAR(max), ChildName), ',') AS ChildNames
, STRING_AGG(CONVERT(NVARCHAR(max), ChildAge), ',') AS ChildAges
FROM dbo.ParentsToSearch
group by ParentID
)
Then compare the two types of families on both child names and child ages values
select a.ParentName, b.ParentID
from ParentNameFamilies a
inner join ParentIDFamilies b on a.ChildNames=b.ChildNames and b.ChildAges=b.ChildAges
and so all together in one view or query:
with ParentNameFamilies as
(
SELECT ParentName, STRING_AGG(CONVERT(NVARCHAR(max), ChildName), ',') AS ChildNames
, STRING_AGG(CONVERT(NVARCHAR(max), ChildAge), ',') AS ChildAges
FROM dbo.ExistingParentsAndChildren
group by ParentName
), ParentIDFamilies as
(
SELECT ParentID, STRING_AGG(CONVERT(NVARCHAR(max), ChildName), ',') AS ChildNames
, STRING_AGG(CONVERT(NVARCHAR(max), ChildAge), ',') AS ChildAges
FROM dbo.ParentsToSearch
group by ParentID
)
select a.ParentName, b.ParentID
from ParentNameFamilies a
inner join ParentIDFamilies b on a.ChildNames=b.ChildNames and b.ChildAges=b.ChildAges