python:HDF和CSV存储优劣对比分析

import pandas as pd
import numpy as np
from wja.wja_tool import test_time as tt
from wja import wja_tool as tool

df = tool.generate_sampleDF(row, col)

tt().run()
df.to_csv('try.csv')
tt().end()

tt().run()
df.to_hdf('try.h5','df',mode='w')
tt().end() 

tt().run()
df1 = pd.read_csv('try.csv')
tt().end()

tt().run()
df2 = pd.read_hdf('try.h5')
tt().end()

# csv保存
# hdf保存
# csv读取
# hdf读取
程序用时:0.015
程序用时:0.9985
程序用时:0.009
程序用时:0.0369

# csv保存
# hdf保存
# csv读取
# hdf读取
程序用时:0.017
程序用时:1.1016
程序用时:0.01
程序用时:0.013

# csv保存
# hdf保存
# csv读取
# hdf读取
程序用时:0.2383
程序用时:1.0308
程序用时:0.0499
程序用时:0.016

# csv保存
# hdf保存
# csv读取
# hdf读取
程序用时:2.0895
程序用时:1.0073
程序用时:0.4055
程序用时:0.0169

# csv保存
# hdf保存
# csv读取
# hdf读取
df = tool.generate_sampleDF(10000,1000)

# csv保存
# hdf保存
# csv读取
# hdf读取
程序用时:23.5693
程序用时:2.2057
程序用时:3.3697
程序用时:0.0619

n,m = 5, 10
road = [[1, 2, 2], [1, 3, 3], [1, 4, 7], [1, 5, 2],
    [2, 3, 4], [2, 4, 9], [2, 5, 5], [3, 4, 4],
    [3, 5, 5], [4, 5, 3]]

def hold(list1, list2):
  jiaoji = list(set(list1)&set(list2))
  need = [i for i in set(list1+list2) if i not in jiaoji]
  need.sort()
  return need

def get(road):
  option = {}
  for i in range (m):
    option[(road[i][0],road[i][1])] = [road[i][2]]
  for i in range (m):
    for j in range(i+1,m):
      dot = hold(road[i][:2], road[j][:2])
      if len(dot)==2:
        if (dot[0],dot[1]) in option.keys():
          option[(dot[0],dot[1])].append(max([road[i][2],road[j][2]]))
        else:
          option[(dot[0],dot[1])] = []
          option[(dot[0],dot[1])].append(max([road[i][2],road[j][2]]))
  road_new = []
  for i in option.items():
    road_new.append(list(i[0])+[min(i[1])])
  if road==road_new:
    print(road_new)
    return road_new
  return get(road_new)