Adrian Klaver <adrian.klaver@aklaver.com> writes:
> Where I am going with this, is that it is not clear to me how you are
> matching the two sets of records to determine whether they are different
> or not.
He's not. The query is forming the cartesian product of the two tables
and then dropping join rows where the tables match ... but every B row is
going to have multiple A rows where it doesn't match, and those join rows
will all survive the WHERE. Then "select distinct" gets rid of the
duplicates, and since nothing from A is presented in the result, it's not
very obvious what's happening.
This is a great example of "select distinct" being used as a band-aid
over a fundamental misunderstanding of SQL. It's good advice to never use
"distinct" unless you know exactly why your query is generating duplicate
rows in the first place.
regards, tom lane
