T-sql – GETDATE – Current date, Month Start and Month End in yyyymmdd

The date type can't store partial dates, like only year, or only day and month.

In one project I had a need to store incomplete dates and I used three separate nullable columns for day (tinyint), month (tinyint) and year (smallint). I don't see in what cases you'd need to use varchar instead of integer types.

In my case it was information about customer birth dates and quite often salespeople didn't know the year, but knew month or day. With this approach it is very flexible and allows the user to enter any bit of information he has. It was also important to know that certain part of the date is missing, such as "we don't know the exact birth date, but we know that it is in June".

Obviously, there were a bunch of custom checks in the application (not at the database level) trying to prevent entering things like 31st of June. In that particular project it was OK to perform the checks in the application. Most likely it would have been possible to implement them in T-SQL as well. In general, it is better to have constraints implemented at the database level, because applications come and go, but data remains.

It was more difficult to use these incomplete dates in queries as well. Different queries/reports required different approaches to dealing with the missing bits.

30K rows is not much at all and performance should not be a problem. So, I'd focus on smart parsing and validation of your Excel data and storing all information that you can extract.

First, get your list of dates from the dimension. Just specify the start and end you need - let's say all of January, since you should have at least a gap at the end:

DECLARE @Start date = '20170101',
  @End date = DATEADD(MONTH,1,'20170101');

SELECT [Date] FROM dbo.DateDimension
  WHERE [Date] >= @Start AND [Date] < @End;

You should see 31 rows here. Now, use an outer join to pull in the rows you have aggregates for:

DECLARE @Start date = '20170101',
  @End date = DATEADD(MONTH,1,'20170101');

SELECT created = d.[Date], total = COALESCE(s.total, 0)
  FROM dbo.DateDimension AS d
  LEFT OUTER JOIN
  (
     SELECT [Date] = CONVERT(date, CreatedAt), 
            total  = COUNT(*)
     FROM dbo.whatever
     WHERE CreatedAt >= @Start
       AND CreatedAt < @End
     GROUP BY CONVERT(date, CreatedAt)
  ) AS s ON d.[Date] = s.[Date]
  WHERE d.[Date] >= @Start 
    AND d.[Date] < @End
  ORDER BY d.[Date];