A Spark DataFrame is a description of a table-shaped computation. Reading spark.table('main.bronze.orders') does not copy all its records into a Python list. Transformations such as select and filter build a plan. Actions, including show, count and collect, ask Spark to execute that plan.
Use display(df) for the workspace result grid. df.show() prints a text table, which is convenient while debugging. Calling show on a column is an error because a column describes one expression and cannot display a table. Start by selecting a few named columns instead of collecting the whole feed. In this browser, the familiar API builds SQL for a local engine. You practice the operations and their results here; distributed execution and production capacity belong to a real Spark environment.