[dba-VB] Merge rows

Merging them into one flat John Colby record is truly ugly, John.  Why
not separate them into a polls table (unless, of course, that's where
you have them already) and use the John Colby record PK to connect them?

Charlotte Foust 

-----Original Message-----
From: dba-vb-bounces at databaseadvisors.com
[mailto:dba-vb-bounces at databaseadvisors.com] On Behalf Of jwcolby
Sent: Monday, September 21, 2009 2:55 PM
To: dwaters at usinternet.com; Discussion concerning Visual Basic and
related programming issues.
Subject: Re: [dba-VB] Merge rows

Dan,

In fact there are 50 million records in the table (though not that many
"duplicates") so there is no "making a decision".  Basically I have to
take row one and update it with row 2, 3, 4...  Delete row 2,3,4...

This is basically a table of "polls" that people have taken, with ~600
fields where there may or may not be information about doing, wanting or
having that thing.  Everything from information about the boat you own
to the electronics you own, to the medicine you take, whether you own a
dog or cat, hunt, fish etc.  In order to get in the database you have to
have volunteered the information somewhere along the line.  But you
might have taken several different polls at different times. 
Thus one record might have the electronics you own, another the medicine
you take, and you don't even own a boat so those fields contain nothing.

Like that.

I just want to take all the John Colby records and merge them into a
single John Colby record, deleting all the extras when I am done.

Of course this is one reason for my other question about getting rid of
any fields where there is no information in any record.  If I am going
to process 50 million records, it would be good to not have to process
any fields where there is no info to merge.

The other reason for the other question is that doing a table scan
against 50 million records with 640 fields is just plain going to take
longer than a table scan against 50 million records with only 400
fields.  It comes down to a matter of disk sectors to load.  And with
640 fields you cannot have indexes against all of them, so you are going
to be doing table scans.

I just did a test where I did a Count() where... on a non-indexed field
and it took 1:18.  So just doing these counts on all 640 fields could
take awhile.  I will be counting and logging the count in my data
dictionary table.

And off we go...

John W. Colby
www.ColbyConsulting.com