pyspark df.select(*) is disordered after df.sort()

Question

This is my original pyspark dataframe.

+----+----+----+
|col1|col2|col3|
+----+----+----+
|   1|   1|   2|
|   1|   2|   2|
|   1|   3|   2|
|   1|   2|   1|
|