I have a very complex execution plan and below is part of it, opened in Plan Explorer.
The image is kind of faded, I guess it's because all operators produce 0 rows.
As you can see, the arrow 1 is a clustered index seek, which produces 0 record. Why the the hash match operator (arrow 3) still has 53% cost? As I understand, hash match builds the hash table based on the output of the upper operator (which is 0 row in this case). After that it get all rows from the lower operator (arrow 2) and probes in the hash table. But in this case, they are all 0 rows. Why it still has such a high cost 53.4%?