Question

I am to create a function to read a table containing some 65 tables and find all columns containing dates, which are stored mostly as int, sometimes as char or datetime. Column names are unknown but I assume ending in ‘%Date’.

I need to find the MAX ie latest date for each client based on his/her SSN. However, the SSN is also called different names, e.g. SSN, SSN_Nr, SSN_No, etc. My code below is hopelessly wrong:

CREATE FUNCTION dbo.LastActivityDate (@SSN varchar(10))
RETURNS datetime
AS
BEGIN
DECLARE @Result datetime

DECLARE @SSN varchar(10)
DECLARE @tableSSN varchar(10)
DECLARE @table varchar(100)
DECLARE @column_name nvarchar(100)
DECLARE @tableDate datetime
DECLARE @LastActivityDate datetime

SET @column_name='%DATE' --only ends in date, so as not to pick up 'update_office',etc

if @SSN IS NULL OR @SSN = '' OR @SSN = ' '  
begin
    BREAK
end 
else 
begin
  WHILE select Table_Name,Field1 from dbo.MERGE_TABLES where [Enabled]='Y' AND Field1=@SSN 
  BEGIN
      SET @table = Table_Name
      SET @tableSSN = Field1
      declare @column_name nvarchar(100)
      set @table = 'e_client'
      set @column_name='%DATE' --only ends in date, so as not to pick up 'update_office',etc
      SELECT column_name FROM INFORMATION_SCHEMA.COLUMNS WHERE table_name=@table AND column_name LIKE @column_name ORDER BY ordinal_position 
      IF ISDATE(SELECT .... FROM INFORMATION_SCHEMA.COLUMNS WHERE table_name=@table AND column_name LIKE @column_name)=1
      begin
          CONVERT(datetime,(SELECT .... FROM INFORMATION_SCHEMA.COLUMNS WHERE table_name=@table AND column_name LIKE @column_name))
      end
.
.
.
      IF (Select MAX([@tableDate]) from @table where @tableSSN=@SSN) > @LastActivityDate --column like FinishDate, proposedFinishDate  or startdate if finish is empty

      begin
         @LastActivityDate=@tableDate
      end
  END

end--@SSN
SET @Result=@LastActivityDate
RETURN @Result
END
GO

Answer 1

• You’ve got an undocumented database.
• You don’t know the column names.
• You don’t know which columns contain dates.
• The dates are stored using at least three different data types.
• You don’t know which columns contain SSANs.
• Columns are inconsistently named.

Some problems cry out for automation. I don’t think this is one of them.

I wouldn’t trust a script to get it right in this database. I think you should look at each table, and determine by eye which columns contain dates and which columns contain SSANs. Even if it takes you 5 minutes a table, that’s still less than a day’s work.

If you have to do it a second time, then you can automate it, based on your new-found knowledge of the tables and columns.

If you worked here, you could use a text tool to find all the SSAN columns, because they’d all be based on a CREATE DOMAIN statement, like (PostgreSQL)

CREATE DOMAIN SSAN AS CHAR(9) CHECK (VALUE ~ '^\\d{9}$');
CREATE TABLE users (
    user_id INTEGER PRIMARY KEY,
    ssan SSAN NOT NULL UNIQUE,     -- These *are* unique where I work. YMMV.
    ...
);