The catch I can see here is that SQL can be expressed on multiple lines and might have irrelevant variations (like one extra space after the SELECT) causing a statement to be entered under a separate count. Or what about 'WHERE A.A=C.A' being logically identical to 'WHERE C.A=A.A' - traversing output from a parser seems the only really reliable approach.

But that aside, once the parsing problem is resolved, the hash count is indeed the best way to finish off.