Computing a column based on comparisons of other columns in table - sql-server

Question

Ok I am working on a database table with 4 columns, lets say a first name, middle name, last name and a group id. I want to group people based on the fact that they have the same first and last names regardless of their middle name. I also want to, if a new entry comes in, give that entry the correct group id.
Here is an example of the result:
----------------------------------------------------------
| First_Name | Middle_Name | Last_Name | Group_ID |
----------------------------------------------------------
| Jon | Jacob | Schmidt | 1 |
----------------------------------------------------------
| William | B. | Schmidt | 1 |
----------------------------------------------------------
| Sally | Anne | Johnson | 2 |
----------------------------------------------------------
I'm not sure whether or not this falls under the jurisdiction of a computed column, some kind of join or something far less obscure, Please help!

Answer 1

If you only need to enumerate the groups within a query then row_number() will work for you:
declare #Names table ( First_Name varchar(10), Middle_Name varchar(10), Last_Name varchar(10))
insert into #Names
select 'Jon', 'Jacob', 'Schmidt' union all
select 'William', 'B.', 'Schmidt' union all
select 'Sally', 'Anne', 'Johnson' union all
select 'Jon', 'Two', 'Schmidt'
;with Groups (First_Name, Last_Name, Group_ID) as
( select First_Name, Last_Name, row_number()over(order by Last_Name)
from #Names
group
by First_Name, Last_Name
)
select n.First_Name, n.Middle_Name, n.Last_Name, g.Group_Id
from #Names n
join Groups g on
n.First_Name = g.First_Name and
n.Last_Name = g.Last_Name;
Be aware the Group_ID value will change as new nameGroups are introduced.
If you want to assign and persist a Group_ID then I would suggest creating an ancillary table and assign the Group_IDs there.
By storing the mapping outside of the #Names table you are allowing users to change their names and not have to worry about re-evaluating the group assignment. It also allows you to modify the grouping logic without re-assigning names. You also have the ability to map similar enough values to the same grouping (John, Jon, Jonny).
Group_ID is composed of a First_Name and Last_Name. So, store it that way.
declare #Names table ( First_Name varchar(10), Middle_Name varchar(10), Last_Name varchar(10))
insert into #Names
select 'Jon', 'Jacob', 'Schmidt' union all
select 'William', 'B.', 'Schmidt' union all
select 'Sally', 'Anne', 'Johnson' union all
select 'Jon', 'Two', 'Schmidt'
declare #NameGroup table (Group_Id int identity(1,1), First_Name varchar(10), Last_Name varchar(10) unique(First_Name, Last_Name));
insert into #NameGroup (First_Name, Last_Name)
select 'Jon', 'Schmidt' union all
select 'Sally', 'Johnson';
declare #Group_ID int;
declare #First_Name varchar(10),
#Middle_Name varchar(10),
#Last_Name varchar(10)
select #First_Name = 'Jon',
#Middle_Name = 'X',
#Last_Name = 'Schmidt'
--be sure the Id has already been assigned
insert into #NameGroup
select #First_Name, #Last_Name
where not exists(select 1 from #NameGroup where First_Name = #First_Name and Last_Name = #Last_Name)
--resolve the id
select #Group_ID = Group_ID
from #NameGroup
where First_Name = #First_Name and
Last_Name = #Last_Name;
--store the name
insert into #Names (First_Name, Middle_Name, Last_Name)
values(#First_Name, #Middle_Name, #Last_Name);
select n.First_Name, n.Middle_Name, n.Last_Name, ng.Group_Id
from #Names n
join #NameGroup ng on
n.First_Name = ng.First_Name and
n.Last_Name = ng.Last_Name;