The problem doesn't appear to be specific to the JDBC driver. Tried a quick version of this in Python for grins with a database that was already populated by the Java code (sadly, the
psycopg2 library doesn't directly support prepared statements):
import psycopg2
import time
conn = psycopg2.connect('dbname=test')
cur = conn.cursor()
cur.execute('PREPARE myplan AS '
'SELECT col2 FROM test WHERE col1 = $1 AND col2 LIKE $2 ORDER BY col2')
times = []
for i in range(0, 20):
start_time = time.time()
cur.execute('EXECUTE myplan (%s, %s)', ('xyz', '%'))
cur.fetchall()
end_time = time.time()
times.append(int((end_time - start_time) * 1000))
print(times)
The output looks similar to the pattern in the Java test code, though it gets slow after 5 iterations rather than 9:
[7, 6, 6, 5, 6, 102, 104, 111, 107, 109, 108, 114, 102, 107, 107, 134, 102, 106, 108, 103]