Denormalize SQL Server Tables

Question

I have two normalized SQL Server 2008 tables, one for names and another for emails:

create table Name (NameId int, Name varchar(50))
create table Email (NameId int, Email varchar(50))
go

insert into Name values (1, 'JOHN SMITH')
insert into Name values (2, 'MARY SMITH')
insert into Name values (3, 'PHILL TAYLOR')
go

insert into Email values (1, 'john@hotmail.com')
insert into Email values (1, 'john@gmail.com')
insert into Email values (1, 'john.smith@hotmail.com')
insert into Email values (2, 'mary@hotmail.com')
insert into Email values (3, 'phill@hotmail.com')
insert into Email values (3, 'phill@gmail.com')
insert into Email values (3, 'taylor.phill@hotmail.com')
insert into Email values (3, 'taylor.phill@gmail.com')
go

When I join those tables I have several rows, which row with just one email:

Select name, email
from Name inner join Email on name.NameId=email.NameId


NAME            EMAIL
-------------   --------------------------
JOHN SMITH      john@hotmail.com
JOHN SMITH      john@gmail.com
JOHN SMITH      john.smith@hotmail.com
MARY SMITH      mary@hotmail.com
PHILL TAYLOR    phill@hotmail.com
PHILL TAYLOR    phill@gmail.com
PHILL TAYLOR    taylor.phill@hotmail.com
PHILL TAYLOR    taylor.phill@gmail.com

But I need to have all the emails in the same row to satisfy a predefined layout:

NAME            EMAIL1              EMAIL2          EMAIL3                      EMAIL4
-------------   -----------------   --------------- -----------------------     -----------------
JOHN SMITH      john@hotmail.com    john@gmail.com  john.smith@hotmail.com
MARY SMITH      mary@hotmail.com
PHILL TAYLOR    phill@hotmail.com   phill@gmail.com taylor.phill@hotmail.com    taylor.phill@hotmail.com

Is it possible using a single select?

Any ideas would be very appreciated

TIA

Ricardo Nogueira

Answer 1

It can be done in 1 select. If you only want the first 4 email addresses and use them as specific columns. Use the RowNumber function on the email table to number the mailaddresses per user. This example assumes the email addresses should be sorted alphabetically:

with m as (
  select 
    ROW_NUMBER() over (partition by nameId order by email) as Nr, 
    email.NameId, email.Email 
  from email
)
select 
  name.NameId, 
  name.Name, 
  m1.Email as Mail1,
  m2.Email as Mail2,
  m3.Email as Mail3,
  m4.Email as Mail4
from name
left join m m1 on (m1.Nr=1 and m1.NameId=name.NameId)
left join m m2 on (m2.Nr=2 and m2.NameId=name.NameId)
left join m m3 on (m3.Nr=3 and m3.NameId=name.NameId)
left join m m4 on (m4.Nr=4 and m4.NameId=name.NameId)

Answer 2

It's not pretty and has usage problems, etc. but if you must.. you must.. So, using your schema, this:-

create function dbo.emails(@NameId int, @Ordinal int)
returns varchar(1024)
as begin
    declare @returnValue varchar(1024)
    select @returnValue=email
    from (
        select 
            ROW_NUMBER() over (order by e.Email) as ordinal,
            e.Email
        from Email e
        where e.NameId=@NameId
    ) p1 
    where p1.ordinal=@Ordinal
    return @returnValue
end
go
select e.Name,
    dbo.emails(n.NameId,1) as email1,
    dbo.emails(n.NameId,2) as email2,
    dbo.emails(n.NameId,3) as email3,
    dbo.emails(n.NameId,4) as email4
from Name n
go

produces:-

Name          email1                  email2             email3                  email4
JOHN SMITH    john.smith@hotmail.com  john@gmail.com     john@hotmail.com        NULL
MARY SMITH    mary@hotmail.com        NULL               NULL                    NULL
PHILL TAYLOR  phill@gmail.com         phill@hotmail.com  taylor.phill@gmail.com  taylor.phill@hotmail.com