Makes sense and I agree, however how do you setup the data catalog if the input file should be the same as the output file, since it is dependent on the output of the same node (but executed a day before) and the data catalog is time-independent.