No meaningful performance overhead: using withColumns for a single column performs the same single projection/plan analysis as withColumn, provided you’re on Spark 3.3+ (the release that introduced withColumns). The only practical difference is readability.