Combine rows if value is blank - sql

Question

I'm using SQL-Server 2008. I need to combine rows with the same Name and increase counter when:
1 or more Id's for the same Name is blank
NOT merge rows if Id is NULL!
NOT merge rows if have the same Name, but different Ids
Output for now:
Name Id Cnt
John 1 1
Peter 2 2 -- This Peter with the same Id have 2 entries so Cnt = 2
Peter 3 1 -- This is other Peter with 1 entry so Cnt = 1
Lisa 4 1
Lisa NULL 1
David 5 1
David 1 -- here Id is blank ''
Ralph 2 -- Ralph have both rows with blank Id so Cnt = 2
Desired output:
Name Id Cnt
John 1 1
Peter 2 2
Peter 3 1
Lisa 4 1
Lisa NULL 1 -- null still here
David 5 2 -- merged with blank '' so Cnt = 2
Ralph 2 -- merged both blanks '' so Cnt = 2
SQL-Query:
This is sample query what I'm using for now:
SELECT Name,
Id,
COUNT(Id) AS Cnt
FROM Employees
WHERE Condition = 1
GROUP BY Name, Id
What I have tried:
Added aggregate MAX to Id in SELECT clause and grouped by Name only, but in this case merged rows with NULL values and with the same names with different Id's what is wrong for me.
SELECT Name,
MAX(Id), -- added aggregate
COUNT(Id) AS Cnt
FROM Employees
WHERE Condition = 1
GROUP BY Name -- grouped by Name only
Have you any ideas? If anything is not clear about problem - ask me, I will provide more details.
UPDATE:
DDL
CREATE TABLE Employees
(
Name NVARCHAR(40),
Id NVARCHAR(40)
);
DML
INSERT INTO Employees VALUES
('John' , '1')
,('Peter', '2')
,('Peter', '2')
,('Peter', '3')
,('Lisa' , '4')
,('Lisa' , NULL)
,('David', '5')
,('David', '')
,('Ralph', '')
,('Ralph', '')
DEMO: SQL FIDDLE

Answer 1

Edit
DECLARE #Data table (Name varchar(10), Id varchar(10)) -- Id must be varchar for blank value
INSERT #Data VALUES
('John', '1'),
('Peter', '2'),('Peter', '2'),
('Peter', '3'),--('Peter', ''), --For test
('Lisa', '4'),
('Lisa', NULL),
('David', '5'),
('David', ''),
('Ralph', ''), ('Ralph', '')
SELECT
Name,
Id,
COUNT(*) + ISNULL(
(SELECT COUNT(*) FROM #data WHERE Name = d.Name AND Id = '' AND d.Id <> '')
, 0) AS Cnt
FROM #data d
WHERE
Id IS NULL
OR Id <> ''
OR NOT EXISTS(SELECT * FROM #data WHERE Name = d.Name AND Id <> '')
GROUP BY Name, Id

Answer 2

You can use CASE statement inside your SELECT. It allows you to set Id = [some value] for employees where it is blank. Query can be something like this:
SELECT E.Name,
CASE
WHEN E.Id = ''
THEN
(Select Employees.Id from Employees where Employees.Id <> '' and E.Name = Employees.Name)
ELSE E.Id
END as Idx,
COUNT(Id) AS Cnt
FROM Employees as E
WHERE Condition = 1
GROUP BY Name, Idx

Answer 3

A version with window functions:
SELECT Name,ID, Cnt from
( select *, sum(1-AmtBlank) over (partition by Name, ID) + sum(case id when 0 then 1 else 0 end) over (partition by Name) Cnt,
rank() over (partition by Name order by AmtBlank ) rnk,
row_number() over (partition by Name, ID order by AmtBlank) rnr
FROM (select * , case id when '' then 1 else 0 end AmtBlank from Employees /*WHERE Condition = 1*/ ) e
) c where rnr=1 and rnk = 1
This uses case id when '' then 1 else 0 end AmtBlank to keep an amount for the blank amounts per row (making the amount for non blanks 1-AmtBlank) and 2 window functions, one with id for a count per name and id (sum(1-AmtBlank) over (partition by Name, ID)) and a count for all blanks in a name section (sum(case id when 0 then 1 else 0 end) over (partition by Name))
The row_number is used to subsequently fetch only the first rows of a group and rank is used to only include the blank records when there are no records with an id.

Answer 4

Try this. using cte and joins
;with cte as (
SELECT Name,
Id,
COUNT(*) AS Cnt
FROM Employees
WHERE isnull(Id,1)<>''
GROUP BY Name, Id
),
cte2 as (SELECT Name,id, COUNT(*) AS Cnt FROM Employees WHERE Id='' GROUP BY Name,id)
select cte.Name,cte.Id,(cte.cnt + ISNULL(cte2.Cnt,0)) as cnt
from cte
left JOIN cte2
on cte.Name = cte2.Name
union all
select cte2.Name,cte2.Id,cte2.cnt
from cte2
left JOIN cte
on cte.Name = cte2.Name
where cte.Name is null

Answer 5

You could try something like this.
;WITH NonBlanks AS
(
SELECT Name,
Id,
COUNT(ISNULL(Id, 1)) AS Cnt
FROM Employees
WHERE ISNULL(Id,0) <> ''
GROUP BY Name, Id
)
,Blanks AS
(
SELECT Name,
Id,
COUNT(ISNULL(Id, 1)) AS Cnt
FROM Employees
WHERE ID = ''
GROUP BY Name, Id
)
SELECT CASE WHEN nb.NAME IS NULL THEN b.NAME ELSE nb.NAME END NAME,
CASE WHEN nb.NAME IS NULL THEN b.Id ELSE nb.Id END Id,
(ISNULL(nb.Cnt,0) + ISNULL(b.Cnt,0)) Cnt
FROM NonBlanks nb FULL JOIN Blanks b
ON nb.Name = b.Name

Answer 6

This simple syntax is compatible with older versions or other RDBMSs
-- Self explained on comments
edited:
select name, id, count(*) from (
-- adds "normal" records
select name, id from Employees where id is null or id <> ''
-- adds one record to each name-notBlankId for each blank id (David, Peter if you add 'Peter','')
-- uncomment /*or id is null*/ if you want even null ids to recieve merged blanks
union all
select e1.name, e1.id
from (select distinct name, id from Employees where id <> '' /*or id is null*/ ) as e1
inner join (select name, id from Employees where id = '') as e2 on e1.name = e2.name
-- adds records that can't be merged (Ralph)
union all
select name, id from Employees e1
where e1.id = ''
and not exists(select * from Employees e2 where e1.name = e2.name and e2.id <> '')
) as fullrecords
group by name, id