How to prevent insertion of value if value exists in other columns - sql

Question

Given 3 columns in a table tblEmails,
email1, nvarchar(50), NULLs not permitted
email2, nvarchar(50), NULLs permitted
email3, nvarchar(50), NULLs permitted
how do I prevent insertions or updates of a (non-NULL) value to any of the three columns if the value already exists in any of the other columns?
I was hoping to apply a CONSTRAINT by checking if the UNION ALL of the three tables contains the value to be inserted, but it seems count() can't be used in CONSTRAINTs.
Any solution implementable via the SSMS gui would be ideal.
I looked through at least a dozen SO posts, some SE posts, and articles online, but could not find a solution (or one that I could understand).

Answer 1

I would suggest creating a function which is then called by the check constraint. Following an example:
CREATE FUNCTION dbo.fn_chkMail(#mail nvarchar(100)) RETURNS INT AS
BEGIN
RETURN (SELECT COUNT(*) FROM mails WHERE mail1 = #mail) + (SELECT COUNT(*) FROM mails WHERE mail2 = #mail) + (SELECT COUNT(*) FROM mails WHERE mail3 = #mail)
END
and then
ALTER TABLE dbo.mails WITH CHECK ADD CONSTRAINT [CK_mails]
CHECK ((dbo.fn_chkMail([mail1]))+(dbo.fn_chkMail([mail2]))+(dbo.fn_chkMail([mail3]))=1)
See fiddle for details: http://sqlfiddle.com/#!18/6f375/7/1

Answer 2

You want to prevent modifying values using update when values are already in the table. Unfortunately, that suggests a trigger.
I think the logic looks like this:
CREATE TRIGGER trg_tblEmails_update ON tblEmails
AFTER UPDATE
AS BEGIN
IF (EXISTS (SELECT 1
FROM inserted i JOIN
deleted d
ON i.<primary key> = d.<primary key>
WHERE (d.email1 IS NOT NULL OR
d.email2 IS NOT NULL OR
d.email3 IS NOT NULL
) AND
(COALESCE(d.email1, '') <> COALESCE(i.email1, '') OR
COALESCE(d.email2, '') <> COALESCE(i.email2, '') OR
COALESCE(d.email3, '') <> COALESCE(i.email3, '')
)
)
)
BEGIN
RAISERROR('You can not update emails when value is already present', 16, 1);
ROLLBACK TRANSACTION;
END;
END;
I would suggest though that there might be a simpler data model. For instance, I would recommend storing the emails in a separate table with one row per email. You would use this table as:
When you insert a value into one email, you insert all three.
You have a unique index on the entity id and email number.
You don't allow updates on the table.
EDIT:
I suspect that you really want a unique constraint. You are not looking within a row but across all rows.
If that is the case, you simply have the wrong data model. You need a table with one email per row. This might require a column to identify which email, but something like this:
create table entity_emails (
entity_email_id int identity(1, 1) primary key,
which_email int,
email varchar(255)
);
Then you want the following constraints:
check (which_email in (1, 2, 3));
unique (entity_id, which_email);
unique (email);
The first two limits the number of emails to 3 per entity. The third insists that the email be unique over all rows and entities.
With the right data model, what you need to do may not require a trigger.