Re: Pg uses non-unique index instead of pkey index

"Thalis A. Kalfigopoulos" <thalis@cs.pitt.edu> writes:
> Why exactly is experimentsc_expid_i chosen over experimentsc_pkey?

There is no reason to prefer either one over the other.  So you get
a quasi-random choice (whichever one the optimizer happens to consider
first, I think).

            regards, tom lane