SQL: how to select a single id (“row”) that meets multiple criteria from a single column

Question

I have a very narrow table: user_id, ancestry.

The user_id column is self explanatory.

The ancestry column contains the country from where the user\'s ancest

Answer 1

one of the approach if you want to get all user_id that satisfies all conditions is:

SELECT DISTINCT user_id FROM table WHERE ancestry IN ('England', '...', '...') GROUP BY user_id HAVING count(*) = <number of conditions that has to be satisfied> 

etc. If you need to take all user_ids that satisfies at least one condition, then you can do

SELECT DISTINCT user_id from table where ancestry IN ('England', 'France', ... , '...')

I am not aware if there is something similar to IN but that joins conditions with AND instead of OR

Answer 2

I was having a similar issue like yours, except that I wanted a specific subset of 'ancestry'. Hong Ning's query was a good start, except it will return combined records containing duplicates and/or extra ancestries (e.g. it would also return someone with ancestries ('England', 'France', 'Germany', 'Netherlands') and ('England', 'France', 'England'). Supposing you'd want just the three and only the three, you'd need the following query:

SELECT Src.user_id
FROM yourtable Src
WHERE ancestry in ('England', 'France', 'Germany')
    AND EXISTS (
        SELECT user_id
        FROM dbo.yourtable
        WHERE user_id = Src.user_id
        GROUP BY user_id
        HAVING COUNT(DISTINCT ancestry) = 3
        )
GROUP BY user_id
HAVING COUNT(DISTINCT ancestry) = 3

Answer 3

This question is some years old but i came via a duplicate to it. I want to suggest a more general solution too. If you know you always have a fixed number of ancestors you can use some self joins as already suggested in the answers. If you want a generic approach go on reading.

What you need here is called Quotient in relational Algebra. The Quotient is more or less the reversal of the Cartesian Product (or Cross Join in SQL).

Let's say your ancestor set A is (i use a table notation here, i think this is better for understanding)

ancestry
-----------
'England'
'France'
'Germany'

and your user set U is

user_id
--------
   1
   2
   3

The cartesian product C=AxU is then:

user_id  |  ancestry
---------+-----------
   1     | 'England'
   1     | 'France'
   1     | 'Germany'
   2     | 'England'
   2     | 'France'
   2     | 'Germany'
   3     | 'England'
   3     | 'France'
   3     | 'Germany'

If you calculate the set quotient U=C/A then you get

user_id
--------
   1
   2
   3

If you redo the cartesian product UXA you will get C again. But note that for a set T, (T/A)xA will not necessarily reproduce T. For example, if T is

user_id  |  ancestry
---------+-----------
   1     | 'England'
   1     | 'France'
   1     | 'Germany'
   2     | 'England'
   2     | 'France'

then (T/A) is

user_id
--------
   1

(T/A)xA will then be

user_id  |  ancestry
---------+------------
   1     | 'England'
   1     | 'France'
   1     | 'Germany'

Note that the records for user_id=2 have been eliminated by the Quotient and Cartesian Product operations.

Your question is: Which user_id has ancestors from all countries in your ancestor set? In other words you want U=T/A where T is your original set (or your table).

To implement the quotient in SQL you have to do 4 steps:

1. Create the Cartesian Product of your ancestry set and the set of all user_ids.
2. Find all records in the Cartesian Product which have no partner in the original set (Left Join)
3. Extract the user_ids from the resultset of 2)
4. Return all user_ids from the original set which are not included in the result set of 3)

So let's do it step by step. I will use TSQL syntax (Microsoft SQL server) but it should easily be adaptable to other DBMS. As a name for the table (user_id, ancestry) i choose ancestor

CREATE TABLE ancestry_set (ancestry nvarchar(25))
INSERT INTO ancestry_set (ancestry) VALUES ('England')
INSERT INTO ancestry_set (ancestry) VALUES ('France')
INSERT INTO ancestry_set (ancestry) VALUES ('Germany')

CREATE TABLE ancestor ([user_id] int, ancestry nvarchar(25))
INSERT INTO ancestor ([user_id],ancestry) VALUES (1,'England')
INSERT INTO ancestor ([user_id],ancestry) VALUES(1,'Ireland')
INSERT INTO ancestor ([user_id],ancestry) VALUES(2,'France')
INSERT INTO ancestor ([user_id],ancestry) VALUES(3,'Germany')
INSERT INTO ancestor ([user_id],ancestry) VALUES(3,'Poland')
INSERT INTO ancestor ([user_id],ancestry) VALUES(4,'England')
INSERT INTO ancestor ([user_id],ancestry) VALUES(4,'France')
INSERT INTO ancestor ([user_id],ancestry) VALUES(4,'Germany')
INSERT INTO ancestor ([user_id],ancestry) VALUES(5,'France')
INSERT INTO ancestor ([user_id],ancestry) VALUES(5,'Germany')

1) Create the Cartesian Product of your ancestry set and the set of all user_ids.

SELECT a.[user_id],s.ancestry
FROM ancestor a, ancestry_set s
GROUP BY a.[user_id],s.ancestry

2) Find all records in the Cartesian Product which have no partner in the original set (Left Join) and

3) Extract the user_ids from the resultset of 2)

SELECT DISTINCT cp.[user_id]
FROM (SELECT a.[user_id],s.ancestry
      FROM ancestor a, ancestry_set s
      GROUP BY a.[user_id],s.ancestry) cp
   LEFT JOIN ancestor a ON cp.[user_id]=a.[user_id] AND cp.ancestry=a.ancestry
WHERE a.[user_id] is null

4) Return all user_ids from the original set which are not included in the result set of 3)

SELECT DISTINCT [user_id]
FROM ancestor
WHERE [user_id] NOT IN (
   SELECT DISTINCT cp.[user_id]
   FROM (SELECT a.[user_id],s.ancestry
         FROM ancestor a, ancestry_set s
         GROUP BY a.[user_id],s.ancestry) cp
   LEFT JOIN ancestor a ON cp.[user_id]=a.[user_id] AND cp.ancestry=a.ancestry
   WHERE a.[user_id] is null
   )