Hi Thomas,

>I have not checked the wording of the contest or the data to be input,
>but the code inside the scan looks mighty fragile.
>What happens if Table1.Name is something like "van de Mol" ?
>Or the street no. is something like "18 a" ?
>(I have done too many adress cleanings in the last month <g>)

I checked the real test data (500,000 records) for things like these and my sample code does what's necessary but not more.

>Have you measured alternative ways to get from Table1 to the deduped Table2 ?

Yes, several alternative ways. The current version of my sample code was the fastest.

>No chance of sidesteping the need to write Table1a ?

You could. But everything I tried was slower. If you find a faster way please let us know.

>Or would it be against the exact wording of the contest to
>write the deduped table *and* the csv at the same time ?

dotnetpro defined that this has to happen in two steps.

Markus