フリー問題

問題文

A DataFrame from a retail feed has 60 columns whose names contain spaces, and the team wants every name normalized in one pass. Which approach is idiomatic in PySpark?

選択肢

1. Convert to a pandas DataFrame, rename there, and convert back.
2. Write the data out and read it back with an explicit schema that has the new names, and the columns are renamed if the schema lists the new names.
3. Build a list of aliased column expressions from the existing column names and pass that list to a single select.
4. Call the rename function once per column in a loop, because each call is planned separately and Spark therefore needs one call per column to keep the lineage of the renames in the correct order for the optimizer.