每个分区的DateDiff和行号

Question

I have a dataset - example data below:

Username         | CreatedDate     |  Reference  |  Note
BOB                2014-05-23         Ref1          blah
BOB                2014-09-23         Ref1          blah
BOB                2014-05-24         Ref2          blah 
BOB                2014-06-23         Ref3          blah
BOB                2014-08-23         Ref3          blah
BOB                2014-01-01         Ref8          blah
BOB                2014-08-18         Ref8          blah
ERIC               2014-12-23         Ref13         blah
ERIC               2014-01-01         Ref18         blah
ERIC               2014-03-18         Ref18         blah

The data will have multiple users.

I have a requirement where I need to highlight any user and the reference, that have created a note more often than every 3 months.

So my results would be:

Username     |  Reference  |  
BOB             Ref3          
ERIC            Ref18         

Could anyone offer advice on how to go about this?

I 've added a:

ROW_NUMBER () OVER (PARTITION BY Usernanme, Reference, ORDER BY CreatedDate) as RowNum

But then I assume I will need to do a DATEDIFF between each RowNum for each Reference and that's where I'm stuck.

Answer 1

SQL Fiddle: http://sqlfiddle.com/#!6/d41d8/23459

declare @data table (
      Id int not null identity(1,1)
    , Username nvarchar(16)
    , CreatedDate date
    , Reference nvarchar(16)
    , Note nvarchar(16)
)

insert @data
      select 'BOB',                '2014-05-23',         'Ref1',          'blah'
union select 'BOB',                '2014-09-23',         'Ref1',          'blah'
union select 'BOB',                '2014-05-24',         'Ref2',          'blah'
union select 'BOB',                '2014-06-23',         'Ref3',          'blah'
union select 'BOB',                '2014-08-23',         'Ref3',          'blah'
union select 'BOB',                '2014-01-01',         'Ref8',          'blah'
union select 'BOB',                '2014-08-18',         'Ref8',          'blah'
union select 'ERIC',               '2014-12-23',         'Ref13',         'blah'
union select 'ERIC',               '2014-01-01',         'Ref18',         'blah'
union select 'ERIC',               '2014-03-18',         'Ref18',         'blah'

select Username, Reference
from @data a
where exists (
    select top 1 1 
    from @data b 
    where b.Username = a.Username 
    and b.Reference = a.Reference 
    and b.CreatedDate between dateadd(month,-3,a.CreatedDate) and dateadd(month,3,a.CreatedDate)
    and b.Id <> a.id --exclude the record itself (if you don't have an id column instead check for more than 1 match)
)
group by Username, Reference

Answer 2

Try this:

SELECT DISTINCT Username, Reference
FROM (SELECT Username, CreatedDate, Reference
, LAG(CreatedDate) OVER (PARTITION BY Username, Reference ORDER BY CreatedDate) as LastDate
FROM theTable) sub
WHERE DATEDIFF(month, sub.LastDate, sub.CreatedDate) < 3 and sub.LastDate is not null;