Coalesce essentially groups multiple partitions into a larger partitions. So use coalesce when you want to reduce the number of partitions (and also tasks) without impacting sort order. Ex:- when you want to write-out a single CSV file output instead of multiple parts
Use repartition when you want to cause a shuffle that changes the number of partitions. A common use-case for repartition is to remove skew in file sizes or to start out with a smaller/different number of partitions than the default in Spark
