There is a huge misconception in this post and that is that in the post it is claimed that Parquet is columnar file format. This is simply not true. The Parquet format is actually a hybrid format because the data is first split horizontally and then each column chunk is compressed and written independently... And this is a key design decision in order to take full advantage of modern CPUs but also to be able to skip some data based on computed statistics in the Metadata part in the footer of Parquet files...

Missing to mention RowGroups and the Footer is a huge mistake when talking about the Parquet file format. Even if the goal is to explain it in a simple way because those two things are just so key for the Parquet format.