Sql-server – Identify changes across records for the field ID

rowsql server

Consider the following table

TabName     ID      Gender    FirstName 
Tab1        C1      M         John   
Tab2        C1      M         John   
Tab3        C1      F         John   
Tab1        C2      F         Mary  
Tab2        C2      F         Mary  
Tab3        C2      F         Mary

I am looking to identify individual IDs where there is inconsistency across the 'tabs'. For example, ID 'C1' is not the same in tabs 'Tab1' and 'Tab2', however, has a change in the Gender field in 'Tab3'. On the other hand, 'Mary' is consistent across all three tabs.

A potential approach may be to find 'ID' where the count of duplicates records (each row is exactly the same) is not equal to the count of occurrences of each 'ID'.

However, I am not sure how I would write the above SQL. Additionally, I think there is probably a better and cleaner approach to this problem.

Best Answer

SELECT id
FROM table
GROUP BY id
HAVING COUNT(DISTINCT gender) != 1
    OR COUNT(DISTINCT FirstName) != 1

Related Solutions

MySQL – Selecting Records with Zero Dues

SELECT DISTINCT ID FROM tbl WHERE DUES = 0

will get the IDs of the naughty people, correct? So now you need to do a "self-JOIN" back to get the rest of the data:

SELECT DISTINCT a.*
    FROM tbl AS a
    JOIN tbl AS b  USING(ID)
    WHERE b.DUES = 0
    ORDER BY a.ID, a.DATE

Or...

SELECT *
    FROM tbl AS a
    WHERE EXISTS ( SELECT ID FROM tbl
                      WHERE DUES = 0
                      AND ID = a.ID
                 )
    ORDER BY ID, DATE

SQL Server – Create Hierarchy with Random Number of Children

^{(O.P.'s note: preferred solution is the 4th / final code block)}

XML seems to me to be the obvious choice of data structure to use here.

with N as
(
  select T.N
  from (values(1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),
              (12),(13),(14),(15),(16),(17),(18),(19),(20)) as T(N)
)

select top(5 + abs(checksum(newid())) % 15)
  N1.N as '@Value',
  (
  select top(1 + abs(checksum(newid())) % 10)
    N2.N as '@Value',
    (
    select top(1 + abs(checksum(newid())) % 5)
      N3.N as '@Value'
    from N as N3
    where N2.N > 0
    for xml path('Level3'), type
    )
  from N as N2
  where N1.N > 0
  for xml path('Level2'), type
  )
from N as N1
for xml path('Level1'), root('Root');

The trick to make SQL Server use different values for top() for each node is to make the sub-queries correlated. N1.N > 0 and N2.N > 0.

Flatteing the XML:

declare @X xml;

with N as
(
  select T.N
  from (values(1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),
              (12),(13),(14),(15),(16),(17),(18),(19),(20)) as T(N)
)
select @X  = (
             select top(5 + abs(checksum(newid())) % 15)
               N1.N as '@Value',
               (
               select top(1 + abs(checksum(newid())) % 10)
                 N2.N as '@Value',
                 (
                 select top(1 + abs(checksum(newid())) % 5)
                   N3.N as '@Value'
                 from N as N3
                 where N2.N > 0
                 for xml path('Level3'), type
                 )
               from N as N2
               where N1.N > 0
               for xml path('Level2'), type
               )
             from N as N1
             for xml path('Level1')
             );


select L1.X.value('@Value', 'varchar(10)')+'\'+
       L2.X.value('@Value', 'varchar(10)')+'\'+
       L3.X.value('@Value', 'varchar(10)')
from @X.nodes('/Level1') as L1(X)
  cross apply L1.X.nodes('Level2') as L2(X)
  cross apply L2.X.nodes('Level3') as L3(X);

And a version totally void of XML.

with N as
(
  select T.N
  from (values(1),(2),(3),(4),(5),(6),(7),(8),(9),(10),(11),
              (12),(13),(14),(15),(16),(17),(18),(19),(20)) as T(N)
)
select cast(N1.N as varchar(10))+'\'+
       cast(N2.N as varchar(10))+'\'+
       cast(N3.N as varchar(10))
from (
     select top(5 + abs(checksum(newid())) % 15)
       N.N
     from N
     ) as N1
  cross apply
     (
     select top(1 + abs(checksum(newid())) % 10)
       N.N
     from N
     where N1.N > 0
     ) as N2
  cross apply
     (
     select top(1 + abs(checksum(newid())) % 5)
       N.N
     from N
     where N2.N > 0
     ) as N3;

Correlation N1.N > 0 and N2.N > 0 is still important.

A version using a table with 20 names to be used instead of just integers.

declare @Elements table
(
  Name nvarchar(50) not null
);

insert into @Elements(Name)
select top(20) C.name 
from sys.columns as C
group by C.name;

select N1.Name + N'\' + N2.Name + N'\' + N3.Name
from (
     select top(5 + abs(checksum(newid())) % 15)
       E.Name
     from @Elements as E
     ) as N1
  cross apply
     (
     select top(1 + abs(checksum(newid())) % 10)
       E.Name
     from @Elements as E
     where N1.Name > ''
     ) as N2
  cross apply
     (
     select top(1 + abs(checksum(newid())) % 5)
       E.Name
     from @Elements as E
     where N2.Name > ''
     ) as N3;

Best Answer

Related Solutions

MySQL – Selecting Records with Zero Dues

SQL Server – Create Hierarchy with Random Number of Children

Related Question