I am quite new to Qlikveiw coding and want to better understand how to build the most efficient QVDs possible.
Am I correct in assuming that any compression benefit would be lost if I make a truly unique column? It seems in testing that when I only include data/date/ID columns I get a great compression rate. When I include an integer ID column I seem to get ok compression. But now that I have created a few fairly unique %Key columns I am basically getting horrible compression.
And on the subject of Keys. I have a fairly unique data set in that it is occurances of compter names over time (maybe millions of events). I want to have the ability to count them various ways. Distinct machines over a week, a month, a day, etc. I frequently count and group by month but occasionally do need the data by day. When I create my unique keys to use in my link tables - should I be creating 2 keys, one for the daily models and one for the monthly models? Or would the daily key be enough if I also include in the data set a new column for 'StartOfMonth' or something?