Re: unoptimized nested loops

c_db=>select count(id) from metadata where author like '%Kelly%';

  count
-------
   3558
(1 row)

I do not see evidence that the nested loop is trying to reduce overhead
by using the smaller set.  It seems to want to scan on data first either
way.

I have run vacuum analyze, in the hopes that the optimizer is
miscalculating, to no avail.