I use an SQL query to get certain records from my database (using DAO w/ jet). Then I copy the data returned into a user-defined type array.

For example:
My database contains these columns...
Code:
"Name", "Age", "Height", "Weight", "Sex"
I define my user-defined type:
Code:
  Type myType
    Name as String
    Age as Integer
    Height as Double
    Weight as Double
    Sex as String
  End Type
Thus the "Name" column would be assigned to myType(1).Name for example.

I have a lot of columns (approx 20) in my real project. I also have about 9,000 records. To make things faster I chose to only "pull" data out of five columns (instead of all twenty) and assign them to my array.

THE QUESTION: (finally!) would there be a large speed/memory usage differential if I just dumped all 20 columns into my array (so I could work with all the values at my leisure?). Note, as of right now I use the data I got (the five columns) to access the database again (later on) to get any other info I need (say the user's "Origin for example).

To sum up, I know that getting all the records right off the bat would be faster and just to give you an idea of my data, the other 15 columns are basically double precision numbers or booleans. I just wanted to get an expert opinion.