#!/usr/bin/env python3 # You must be this tall to enjoy this ride ######################################################################################## """Do a quick performance comparison between dict and treap.""" import sys import time import treap # we define our own random function, because the treap module uses the random module, so just using a fixed seed doesn't give # consistent results between dictionaries and treaps, because dictionaries don't need random values. So this test uses its own # random function def my_random(n): """Generate simple LCG random numbers.""" # three largish consecutive primes. No, I didn't check Knuth for how good these values are - we don't need cryptographic # strength randomness this time a = 99999959 b = 99999971 c = 99999989 r = n while True: r = (r * a + b) % c yield r def test(description, dict1, n): """Test the dict-like.""" create_t0 = time.time() for counter, key, value in zip(range(n), my_random(n), my_random(n)): # print counter, key, value dict1[key] = value create_t1 = time.time() # retrieve_t0 = time.time() # for counter, key in zip(range(n), my_random(n)): # if key in dict1: # dummy = dict1[key] # retrieve_t1 = time.time() sort_t0 = time.time() keys = list(dict1.keys()) if description == 'dictionary': keys.sort() sort_t1 = time.time() return dict1, create_t1 - create_t0, sort_t1 - sort_t0 def main(): """Do the comparison.""" full_compare_top = 20 # print 'legend:' # print ' cre: create (low is good)' # print ' trp: treap' # print ' {}: dictionary' # print ' quo: quotient: dictionary time over treap time (low is good for dictionary, high is good for treap)' # print ' rtr: retrieve (low is good)' # print ' srt: srt (low is good)' # print( 'exponent|' + 'number of elements|' + 'dict create (low is good)|' + 'treap create (low is good)|' + 'creation quotient (low is good for dictionaries, high is good for treaps)|' + 'dict sort (low is good)|' + 'treap sort (low is good)|' + 'sort quotient (low is good for dictionaries, high is good for treaps)') for exponent in range(0, full_compare_top): num_elements = pow(2, exponent) dictionary_result_collection, dictionary_create_duration, dictionary_sort_duration = \ test('dictionary', {}, num_elements) if exponent > full_compare_top: del dictionary_result_collection treap_result_collection, treap_create_duration, treap_sort_duration = test('treap', treap.treap(), num_elements) if exponent > full_compare_top: del treap_result_collection else: dict_result = list(dictionary_result_collection.items()) dict_result.sort() treap_result = list(treap_result_collection.items()) treap_result.sort() # we really want zip_longest here. On well, check the length if len(dict_result) != len(treap_result): sys.stderr.write('%s: lengths do not match: %d and %d\n' % (sys.argv[0], len(dict_result), len(treap_result))) sys.exit(1) paired_up = zip(dict_result, treap_result) for (d, t) in paired_up: dict_key, dict_value = d treap_key, treap_value = t if dict_key != treap_key or dict_value != treap_value: print('((%d, %d), (%d, %d))' % (dict_key, dict_value, treap_key, treap_value)) sys.stderr.write('%s: Hmmm, bad comparison\n' % sys.argv[0]) sys.exit(1) # avoid division by 0 try: create_quotient = '%4.2f' % (dictionary_create_duration / treap_create_duration) except ZeroDivisionError: create_quotient = '%4s' % 'Inf' try: sort_quotient = '%4.2f' % (dictionary_sort_duration / treap_sort_duration) except ZeroDivisionError: sort_quotient = '%4s' % 'Inf' print('%d|%d|%.4f|%.4f|%s|%.4f|%.4f|%s' % ( exponent, num_elements, dictionary_create_duration, treap_create_duration, create_quotient, dictionary_sort_duration, treap_sort_duration, sort_quotient, )) sys.stdout.flush() # dictionary_retrieve_duration, treap_retrieve_duration, retrieve_quotient, \ main()