I believe sql server optimizer will choose loop join if one of the input table will be relatively small . like few rows. but in the following example show that it use hash join and using the smaller table as to build hash column in memory . so the query optimizer make a wrong choice ?
