Here's the sequence I'd like to see for csv files - a checkbox in the connect dialog that uses the data structure that was already established in the tableau workbook, rather than re-guessing what the datatypes should be for each element. So, this would be an option only when refreshing an existing datasource. The problem this solves is that it typically takes some time to get the data types structured properly in a csv by establishing the first rows with dummy records. Once this task is done, however, subsequent csv data refreshes always have to reload the dummy records.
My suggestion is that a tableau workbook already has the intended structure and datatypes in its metadata. Sometimes you want to keep that structure as is and not have to load dummy records to get it all to work. In those instances, it would be nice to be able to check that a csv file data refresh (or even pointing to another csv datafile with the same structure) is to use the tableau workbook metadata for inference of datatypes rather than guessing from the first few rows.
What do you think? Do you experience similar challenges working with csv files? Would it be nice to create a datastructure csv using dummy records, then point that structure to the actual data csv? Would not having to worry about dummy records during production updates be something worthwhile to have?