Hi,
To be able to perform this denormalization you must make sure that you know the names of the columns you want in your output.
If these column names can be variable you should not try the approach I'm about to explain. In that case you need some custom tJavaRow code.
If you do know the exact column names however, you can use a tMap in combination with a tAggregateRow as seen in my screenshots.
I've copied your data into a tab delimited file instead of a table and use a log row in stead of an output table, but the idea is the same.
Hope this helps!
Best regards,
Arno