then I found it was because I had converted the CSV file to an Excel file. I had a number of different expressions and they were still fine after converting to an Excel file except for the expressions that included ={''}. These no longer worked at all. They weren't seeing the empty fields as empty any longer. I tried ={'*'} and this would see the cells with something in them, but -={'*'} would not work. It was still seeing the empty cells as having something in them. I went to the source file and made searches for fields with nothing in them and this showed them as empty and there are no formulas in them either. I reverted back to a CSV file for the source and all the set analysis expressions worked again.
Without a need to adjust or to add data you shouldn't convert a txt/csv-file into an excel-file and even then it might be useful to convert it back into txt/csv-file again. The handling from txt-files to excel-files differs by formatings and NULL's and for this are also various null-variables available to adjust the null-handling. A very good starting point could you find here: NULL handling in QlikView.
Beside them your above used approach from -={'*'} should work. Maybe these fieldvalues aren't NULL else contain any invisible content. You could check for example with functions like len() or ord() which content is really there.
