Chris Webb performs some experiments:
Recently I was involved in an interesting discussion on
datasetsemantic model can affect compression and therefore its size and query performance. You can read the thread here. It got me thinking: is there a way to get more detail on how well compression, and in particular run-length encoding (RLE from hereon), is working for a column when you’re using Import mode or Direct Lake? After a bit of research I found out there is, so let’s see some examples that illustrate what I learned.
The results are interesting and Chris explains why we see what we do.