Remove duplicates with less null values - sql

Question

I have a table of employees which contains about 25 columns. Right now there are a lot of duplicates and I would like to try and get rid of some of these duplicates.
First, I want to find the duplicates by looking for multiple records that have the same values in first name, last name, employee number, company number and status.
SELECT
firstname,lastname,employeenumber, companynumber, statusflag
FROM
employeemaster
GROUP BY
firstname,lastname,employeenumber,companynumber, statusflag
HAVING
(COUNT(*) > 1)
This gives me duplicates but my goal is to find and keep the best single record and delete the other records. The "best single record" is defined by the record with the least amount of NULL values in all of the other columns. How can I do this?
I am using Microsoft SQL Server 2012 MGMT Studio.
EXAMPLE:
Red: DELETE
Green: KEEP
NOTE: There are a lot more columns in the table than what this table shows.

Answer 1

You can use the sys.columns table to get a list of columns and build a dynamic query. This query will return a 'KeepThese' value for every record you want to keep based on your given criteria.
-- insert test data
create table EmployeeMaster
(
Record int identity(1,1),
FirstName varchar(50),
LastName varchar(50),
EmployeeNumber int,
CompanyNumber int,
StatusFlag int,
UserName varchar(50),
Branch varchar(50)
);
insert into EmployeeMaster
(
FirstName,
LastName,
EmployeeNumber,
CompanyNumber,
StatusFlag,
UserName,
Branch
)
values
('Jake','Jones',1234,1,1,'JJONES','PHX'),
('Jake','Jones',1234,1,1,NULL,'PHX'),
('Jake','Jones',1234,1,1,NULL,NULL),
('Jane','Jones',5678,1,1,'JJONES2',NULL);
-- get records with most non-null values with dynamic sys.column query
declare #sql varchar(max)
select #sql = '
select e.*,
row_number() over(partition by
e.FirstName,
e.LastName,
e.EmployeeNumber,
e.CompanyNumber,
e.StatusFlag
order by n.NonNullCnt desc) as KeepThese
from EmployeeMaster e
cross apply (select count(n.value) as NonNullCnt from (select ' +
replace((
select 'cast(' + c.name + ' as varchar(50)) as value union all select '
from sys.columns c
where c.object_id = t.object_id
for xml path('')
) + '#',' union all select #','') + ')n)n'
from sys.tables t
where t.name = 'EmployeeMaster'
exec(#sql)

Answer 2

Try this.
;WITH cte
AS (SELECT Row_number()
OVER(
partition BY firstname, lastname, employeenumber, companynumber, statusflag
ORDER BY (SELECT NULL)) rn,
firstname,
lastname,
employeenumber,
companynumber,
statusflag,
username,
branch
FROM employeemaster),
cte1
AS (SELECT a.firstname,
a.lastname,
a.employeenumber,
a.companynumber,
a.statusflag,
Row_number()
OVER(
partition BY a.firstname, a.lastname, a.employeenumber, a.companynumber, a.statusflag
ORDER BY (CASE WHEN a.username IS NULL THEN 1 ELSE 0 END +CASE WHEN a.branch IS NULL THEN 1 ELSE 0 END) )rn
-- add the remaining columns in case statement
FROM cte a
JOIN employeemaster b
ON a.firstname = b.firstname
AND a.lastname = b.lastname
AND a.employeenumber = b.employeenumber
AND a.companynumbe = b.companynumber
AND a.statusflag = b.statusflag)
SELECT *
FROM cte1
WHERE rn = 1

Answer 3

I test with MySQL and use NULL String concat to found the best record. Because LENGTH ( NULL || 'data') is 0. Only if all column not NULL some length exists. Maybe this is not perfekt.
create table EmployeeMaster
(
Record int auto_increment,
FirstName varchar(50),
LastName varchar(50),
EmployeeNumber int,
CompanyNumber int,
StatusFlag int,
UserName varchar(50),
Branch varchar(50),
PRIMARY KEY(record)
);
INSERT INTO EmployeeMaster
(
FirstName, LastName, EmployeeNumber, CompanyNumber, StatusFlag, UserName, Branch
) VALUES ('Jake', 'Jones', 1234, 1, 1, 'JJONES', 'PHX'), ('Jake', 'Jones', 1234, 1, 1, NULL, 'PHX'), ('Jake', 'Jones', 1234, 1, 1, NULL, NULL), ('Jane', 'Jones', 5678, 1, 1, 'JJONES2', NULL);
My query idea looks like this
SELECT e.*
FROM employeemaster e
JOIN ( SELECT firstname,
lastname,
employeenumber,
companynumber,
statusflag,
MAX( LENGTH ( username || branch ) ) data_quality
FROM employeemaster
GROUP BY firstname, lastname, employeenumber, companynumber, statusflag
HAVING count(*) > 1
) g
ON LENGTH ( username || branch ) = g.data_quality