What is the difference between an anti join and a left anti join?

An anti join in Spark returns rows from the left DataFrame that have no matching key in the right DataFrame. It is essentially a filter that removes any rows that have a match. A left anti join is a specific type of anti join that preserves the order of the left DataFrame and returns all rows from the left that do not have a match in the right. The difference is subtle: a plain anti join may not guarantee the original order of the left side, whereas a left anti join guarantees that the result is a subset of the left DataFrame in the same order. In practice, both produce the same set of rows, but left anti join is often used when you want to maintain the left DataFrame's ordering or when you need to preserve the original schema. Performance-wise, both are implemented using a broadcast hash join or shuffle hash join depending on the size of the right DataFrame. The key takeaway is that an anti join filters out matches, while a left anti join does the same but explicitly indicates that the left side is the source of truth.