If I knew the order of the columns a priori, that would work, but sadly I don't. The number of columns, as well as the order, are random due to the fact that different sondes collect different data. In fact, my plan was to first determine the order of the columns in the source file, then I thought about using ALTER TABLE (in combination with SELECT column-of-interest) to put each column, one-by-one, into the new table in the proper order - last to first. If I select columns as you suggest, I still have to insert "empty" columns to maintain the proper column order, and I haven't found a way to insert a column into the middle of a table. (ALTER TABLE always creates a new column at the front of the table.)

I was also thinking of biting the bullet and learning about MS JET (ODBC provider) attributes/field names as a way of discovering what fields (columns) are available, but I'm trying to keep this as close to "token handling" as possible so I can easily cleave MS-specific file stuff from the meat of the program.

On 7/25/07, Chuck Anderson <cra@wpi.edu> wrote:

On Wed, Jul 25, 2007 at 09:29:43AM -0400, Stephen Daukas wrote:
> There are literally hundreds of files of data in CSV format, and I intend to
> process each by itself, preserving the original, and creating a new
> "adjusted" file for the next step.  (The next step would be to concatenate
> all files together for numeric processing.)  Original files must be
> preserved (chain of custody requirements), so I am opening the file,
> selecting records using SQL, and am now ready to do the rearranging/writing
> to a new file...

So you are selecting records out of a CSV file using SQL?  How about
this:

select columnB,columnG,columnA,columnC,columnF from foo;

That will take the columns and return them in the given order.
_______________________________________________
Wlug mailing list
Wlug@mail.wlug.org
http://mail.wlug.org/mailman/listinfo/wlug