Hi,
DI 3.4 may not support my approach; I haven't checked. What I do in 4.x, in a nutshell, is to
1) define the CSV in the metadata with a macro variable reference as the filename (like &csvfile). make certain the generated code wraps this in double quotes or you're stuffed. This may actually be the breaking point for 3.4.
2) define and load a table that holds files present in your input location and keeps track of their processed status, eg through a timestamp.
3) Then in a job process that table in the first step to find out which file must be processed. Stick that value in a macro var. As the next step process the external file you defined in the metadata. Give that step a parameter to transfer the result from step 1 into the macro variable that matches the filename of the CSV file from 1). As a last step (or a subsequent job) mark the file as processed in the table as defined in 2).
Good luck, Jan.
