I've got a pipe-delimited file to which I connected via Tableau Desktop and then created an extract (.tde file). When the original extract and metadata were created, I noticed that some of the columns were incorrectly set as a number data type and placed in the measures pane. As an example, one of the columns is a member ID value. Many of the ID's are numeric only but most are a mix of alpha-numeric. I'm guessing Tableau reads a subset of rows to determine the best-fit data type. I've since changed the attribute to a dimension and converted it to a string data type.
The issue is that when Tableau first loaded the data, it excluded the member ID values that were mixed alpha-numeric and left them null. If I refresh the extract with the updated metadata (member ID is now set as string), it still only loads the values that are numeric, leaving the majority of them null. How can I refresh the data and successfully load all of the values for this attribute?
So after some additional research and testing, I found the solution to this problem is to modify the Schema.ini file. Within this file I was able to force the data type of each column, including setting the member ID as text. When I refresh the extract, it now loads all of the member ID values successfully.