Retrieve column names from a different table?

Question

I have a "datadump" table that has a bunch of mixed performance-related data. Something like:

MachID  TestDate  MachType  Value1   Value2    ... 
00001   01/01/09  Server    15        48       
00001   01/02/09  Server    16        99       
19999   01/01/09  Switch    32        4.9880   
19999   01/02/09  Switch    32        5.8109

The trick is that the "values" columns MEAN different things for different types of machines. So we have a "xRef" table that looks like:

MachType       Column   Description
Server         Value1   Users Connected
Server         Value2   % CPU _total
Switch         Value1   Number of Ports
Switch         Value2   packets/ms 
...

I know, weird structure, but I didn't make it, and can't change it.

I'd like to somehow "inner join" these so I can query the appropriate column headers based on the type of data. Something like this for the servers:

MachID  TestDate  MachType  Users Connected     % CPU _total    Total RAM
00001   01/01/09  Server    15                  48              4096
00001   01/02/09  Server    16                  99              4096

and this for the switches:

MachID  TestDate  MachType  Number of Ports   packets/ms  Total Cumulative kb
19999   01/01/09  Switch    32                4.9880      1024547
19999   01/02/09  Switch    32                5.8109      1029450

Is there a way to do this without doing individual hard-coded queries for each type?

Note: I will only need to query one type of object at a time. Most likely, I'll only be looking at all results between particular dates for a single MachID, if that helps. This is MS SQL 2000.

Thanks!

Answer 1

Create a table storing the header-name for each value for that individual type of query.

Then, create a stored procedure and use Dynamic SQL to fill in the column name as drawn from that table.

Answer 2

Since you can't change the data model I would suggest putting the presentation code into the presentation layer of your application. Have a table that gives you the column headings to use based on the results being requested and go from there.

Answer 3

This will do them all together - you can modify as appropriate if you want them all split up.

DECLARE @template AS varchar(max)
DECLARE @sql AS varchar(max)
DECLARE @column_list AS varchar(max)
SELECT  @column_list = COALESCE(@column_list + ', ', '')
        + QUOTENAME([Description])
FROM    xRef

SET @template = ';
WITH    up
          AS (
              SELECT    MachID
                       ,TestDate
                       ,MachType
                       ,[COLUMN]
                       ,[Value]
              FROM      datadump UNPIVOT ( [Value] FOR [Column] IN ([Value1], [Value2]) ) AS unpvt
             )
             ,ready AS (
    SELECT  machID
           ,TestDate
           ,up.MachType
           ,[Description]
           ,up.[Value]
    FROM    up
    INNER JOIN xRef
            ON xRef.machType = up.MachType
               AND xRef.[Column] = up.[Column]
)
SELECT * FROM ready
PIVOT (SUM([Value]) FOR [Description] IN ({@column_list})) AS pvt
'

machID TestDate                MachType Users Connected                         % CPU _total                            Number of Ports                         packets/ms
------ ----------------------- -------- --------------------------------------- --------------------------------------- --------------------------------------- ---------------------------------------
00001  2009-01-01 00:00:00.000 Server   15.000000000000000                      48.000000000000000                      NULL                                    NULL
00001  2009-01-02 00:00:00.000 Server   16.000000000000000                      99.000000000000000                      NULL                                    NULL
19999  2009-01-01 00:00:00.000 Switch   NULL                                    NULL                                    32.000000000000000                      4.988000000000000
19999  2009-01-02 00:00:00.000 Switch   NULL                                    NULL                                    32.000000000000000                      5.810900000000000

Answer 4

A dynamic sql option would be (written out as a query rather than made into a proc):

declare @machtype varchar(40) --stored proc parameter?
set @machtype = 'Switch' --or 'Server'

declare @sql nvarchar(4000)

set @sql = 'select
    MachID,
    TestDate,
    MachType,
    Value1 as ''' + (select [Description] from dbo.xref where machtype = @machtype and [Column] = 'Value1') + ''',
    Value2 as ''' + (select [Description] from dbo.xref where machtype = @machtype and [Column] = 'Value2') + ''',
    Value3 as ''' + (select [Description] from dbo.xref where machtype = @machtype and [Column] = 'Value3') + '''
from 
    dbo.datadump
where
    machtype = ''' + @machtype + ''''

exec sp_executesql @sql

If you find that simply too ugly for you then wrapping the logic for getting the column name in a function would tidy it up:

create function dbo.ColNameForDataDump(
    @machtype varchar(40),
    @column varchar(40)
)
RETURNS varchar(40)
as
begin

    declare @col_desc varchar(40)
    select
     @col_desc = [description]
    from
     dbo.xref
    where
     machtype = @machtype
     and [column] = @column

    return @col_desc
end

Then your dynamic SQL will look more like:

declare @machtype varchar(40) --stored proc parameter?
set @machtype = 'Switch' --or 'Server'

declare @sql nvarchar(4000)

set @sql = 'select
    MachID,
    TestDate,
    MachType,
    Value1 as ''' + dbo.ColNameForDataDump(@machtype, 'Value1') + ''',
    Value2 as ''' + dbo.ColNameForDataDump(@machtype, 'Value2') + ''',
    Value3 as ''' + dbo.ColNameForDataDump(@machtype, 'Value3') + '''
from 
    dbo.datadump
where
    machtype = ''' + @machtype + ''''

exec sp_executesql @sql

Finally a passing point / comment on the code above: you mentioned that you are on SQL Server 2000 so make sure when you do have to write some dynamic sql to define it as an nvarchar and use sp_executesql to call it...thereby negating some of the performance pain of having to go dynamic.