Sql for distinct record comparison - sql

Question

I am comparing a table to itself trying to determine whether an email in one record is being used in any one of four other columns in another record.
To make this easier, lets look at an example (simplified):
Name: Bob
Office Email: bob#aaa.com
Home Email: bob#home.com
Mobile Email: bobster#gmail.com
.
Name: Rob
Office Email: rob#bbb.com
Home Email: bob#home.com
Mobile Email: robert#gmail.com
Now I have a sql statement like this:
select c1.ContactId id1, c1.FullName Name1, 'Office Email 1' EmailType1, c1.EMailAddress1 Email,
c2.ContactId id2, c2.FullName Name2,
CASE c1.EmailAddress1
WHEN c2.EMailAddress1 THEN 'Office Email 1'
WHEN c2.Si_OfficeEmail2nd THEN 'Office Email 2'
WHEN c2.EMailAddress2 THEN 'Mobile Email'
WHEN c2.pc_hmemail THEN 'Home Email'
ELSE '?'
END EmailType2,
CASE c1.EmailAddress1
WHEN c2.EMailAddress1 THEN c2.EMailAddress1
WHEN c2.Si_OfficeEmail2nd THEN c2.Si_OfficeEmail2nd
WHEN c2.EMailAddress2 THEN c2.EMailAddress2
WHEN c2.pc_hmemail THEN c2.pc_hmemail
ELSE '?'
END DuplicateEmail
from Contact c1, Contact c2
where (
LTRIM(RTRIM(c1.EMailAddress1 )) = LTRIM(RTRIM(c2.EMailAddress1))
Or LTRIM(RTRIM(c1.EMailAddress1 )) = LTRIM(RTRIM(c2.EMailAddress2))
Or LTRIM(RTRIM(c1.EMailAddress1 )) = LTRIM(RTRIM(c2.pc_hmemail))
Or LTRIM(RTRIM(c1.EMailAddress1 )) = LTRIM(RTRIM(c2.Si_OfficeEmail2nd))
)
And c1.ContactId <> c2.ContactId
And c1.StateCode = 0
and c2.StateCode = 0
order by c1.FullName, c2.FullName
Unfortunately, because Bob and Rob have the same email 'type' (Home Email) that is duplicated due to a typo, my query returns two records, one which shows that Bobs email is duplicated in Robs email, and a second that Robs email is duplicated in Bobs email.
I only need one record. I'm sure this is a common problem but I don't quite know how to describe this problem well enough to have a search engine return something useful.
Perhaps there is a better way of going about this? If not, other than jumping through a bunch of intermediate temporary tables to eliminate these equivalent records, is there a way to write a single query for this?

Answer 1

The solution to your problem is to add the condition: c1.contactId < c2.ContactId. This limits the pairs you are looking at.
If you are looking at emails, you might find a faster approach to look directly at emails. Something like the following will return all emails (on separate rows) that are duplicated:
select e.*
from (select e.*, COUNT(*) over (partition by email) as NumTimes
from ((select contactId, 'Office' as which, EmailAddress1 as email
from Contact
) union all
(select contactId, 'Office2', Si_OfficeEmail2nd
from Contact
) union all
(select contact_id, 'Home', pc_hmemail
from Contact
) union all
(select contact_id, 'Mobile', EmailAddress2
from Contact
)
) e
where email is not null and email <> ''
) e
where NumTimes > 1
order by email

Answer 2

I'd first suggest to continue to normalise your datastructure. A person may have several types of contact information. Therefore the personID, typeID and value can be placed into another table. From this table you can create another relation with a type table, where you keep track of the different contact types (e.g. Home E-mail, Work e-mail, Twitter, linkedIn, Facebook etc). It does not only improve the extendibility of your system but also enables to run these types of queries much more efficiently.
SELECT user.username FROM user u LEFT JOIN contactinfo ci ON u.user_id=ci.user_id LEFT JOIN contacttype ct ON ci.type_id=ct.type_id GROUP BY ci.value HAVING count(value)>1 would be the query to find any duplicate source