Compare PL/R implementation of an average to PostgreSQL `avg()`

gistfile1.sql

CREATE OR REPLACE FUNCTION r_mean(float8) RETURNS FLOAT AS $$
  return(mean(arg1))
$$ language 'plr';

CREATE AGGREGATE mean (
  sfunc = plr_array_accum,
  basetype = float8,
  stype = _float8,
  finalfunc = r_mean
);

CREATE UNLOGGED TABLE things (floats float8);
INSERT INTO things SELECT generate_series(1, 1e5::int);

SELECT avg(floats) from things;
-- 30 ms

SELECT mean(floats) from things;
-- 7 sec, and grows exponentially with number of rows !