python - Python 3 字典迭代中的性能 : dict[key] vs. dict.items()

Question

哪个更快，为什么？或者它们是一样的？答案是否因任何条件(字典大小、数据类型等)而异？

传统的:

for key in dict:
    x = dict[key]
    x = key

潮人:

for key, value in dict.items():
    y = value
    y = key

我还没有看到完全相同的副本，但如果有，我很乐意指出它。

Answer 1

事实证明实际上存在数量级的差异。

我不太了解性能测试，但我尝试做的是创建 3 个不同大小的字典，每个较小的字典都是较大字典的子集。然后，我通过两个函数(传统与时髦)运行了所有三个命令。然后我做了 100 次。

dict1、dict2 和 dict3 的字典大小(键值对的数量)分别为 1000、50000、500000。

似乎有显着差异，d.items() 通常更快 而 d.items() WAY 在较小的词典上更快。这符合预期(Python 通常奖励“pythonic”代码)。

结果:

--d[key]--
dict1 -- mean: 0.0001113555802294286, st. dev: 1.9951038526222054e-05
dict2 -- mean: 0.01669296698019025, st. dev: 0.019088713496142
dict3 -- mean: 0.2553815016898443, st. dev: 0.02778986771642094

--d.items()--
dict1 -- mean: 6.005059978633653e-05, st. dev: 1.1960199272812617e-05
dict2 -- mean: 0.00507106617995305, st. dev: 0.009871762371401046
dict3 -- mean: 0.07369932165958744, st. dev: 0.023440325168927384

提供结果的代码(repl.it):

import timeit
import random
import statistics

def traditional(dicty):

  for key in dicty:
    x = dicty[key]
    x = key

def hipster(dicty):

  for key, value in dicty.items():
    y = value
    y = key

def generate_random_dicts():
  random_dict1, random_dict2, random_dict3 = {}, {}, {}

  for _ in range(1000):
    key = generate_random_str_one_to_ten_chars()
    val = generate_random_str_one_to_ten_chars()
    random_dict1[key] = val
    random_dict2[key] = val
    random_dict3[key] = val

  for _ in range(49000):
    key = generate_random_str_one_to_ten_chars()
    val = generate_random_str_one_to_ten_chars()
    random_dict2[key] = val
    random_dict3[key] = val

  for _ in range(450000):
    key = generate_random_str_one_to_ten_chars()
    val = generate_random_str_one_to_ten_chars()
    random_dict3[key] = val

  return [random_dict1, random_dict2, random_dict3]

def generate_random_str_one_to_ten_chars():
  ret_str = ""
  for x in range(random.randrange(1,10,1)):
    ret_str += chr(random.randrange(40,126,1))
  return ret_str

dict1, dict2, dict3 = generate_random_dicts()

test_dicts = [dict1, dict2, dict3]

times = {}
times['traditional_times'] = {}
times['hipster_times'] = {}

for _ in range(100):

  for itr, dictx in enumerate(test_dicts):
    start = timeit.default_timer() 
    traditional(dictx)
    end = timeit.default_timer() 
    time = end - start
    try:
      times['traditional_times'][f"dict{itr+1}"].append(time)
    except KeyError:
      times['traditional_times'][f"dict{itr+1}"] = [time]

    start = timeit.default_timer() 
    hipster(dictx)
    end = timeit.default_timer() 
    time = end - start
    try:
      times['hipster_times'][f"dict{itr+1}"].append(time)
    except KeyError:
      times['hipster_times'][f"dict{itr+1}"] = [time]

print("--d[key]--")
for x in times['traditional_times'].keys():
  ltimes = times['traditional_times'][x]
  mean = statistics.mean(ltimes)
  stdev = statistics.stdev(ltimes)
  print(f"{x} -- mean: {mean}, st. dev: {stdev}\n\n")

print("--d.items()--")
for x in times['hipster_times'].keys():
  ltimes = times['hipster_times'][x]
  mean = statistics.mean(ltimes)
  stdev = statistics.stdev(ltimes)
  print(f"{x} -- mean: {mean}, st. dev: {stdev}")