Question

This is probably a familiar problem but i didn’t find an answer in searches. I do have a solution (shown below) that involves COALESCE on every non-key column, I’m wondering if there’s a Dynamic TSQL routine, or perhaps a different way of forumulating the query that doesn’t involve explicitly naming all the columns.

I have two tables, “targetTable” and “deltaTable”, that have a the same field names and types. Both have a unique key, “id”.

Deltatable has data that has changed for a corresponding row in targetTable. Any column where the value hasn’t changed is null. I want to update the columns in the targetTable that have non-null values in the corresponding row of the deltaTable. DetlaTable.id is guaranteed to exist in TargetTable.id, and I never want to update a field to NULL.

I’m using SQL Server 2005 (which doesn’t have the merge statement, not that I know it would help).

So if a targetTable row looks like

  id | name    |  dob        | next_appt
  5    Bill       1-1-2012   | 3-3-2015

and a deltaTable row looks like

  id | name     | dob        | next_appt
  5  | NULL     | NULL       | 4-4-2015

I want targetTable updated to

  id | name    |  dob        | next_appt
  5    Bill       1-1-2012   | 4-4-2015

What I have now that I want to improve on is:

UPDATE target
SET target.colA = COALESCE(delta.colA, target.colA),
    target.colN = COALESCE(delta.colN, target.colN)
FROM  delta JOIN target on delta.id = target.id

Answer 1

I’m not sure it’s a great idea, but if you really want to do it, you could do something like this:

DECLARE @sql nvarchar(max)
SET @sql = N'UPDATE target SET
';
SELECT @sql = @sql + N'  target.'+QUOTENAME([name]) + ' = COALESCE(delta.' + QUOTENAME([name]) + ', target.' + QUOTENAME([name]) + '),
'
FROM sys.columns 
where object_id = object_id('target') AND [name] != 'id'
order by [column_id]

SET @sql = STUFF(@sql,LEN(@sql)-2,1,''); -- strip off the last commma
set @sql = @sql + 
'FROM delta 
JOIN target ON  delta.id = target.id'

EXECUTE sp_executesql @sql