首页 > > 网络编程 > 其它 >

pandas的合并、连接、去重、替换

2019-05-08 07:29:58来源：博客园阅读 ()

  1 import pandas as pd
  2 import numpy as np
  3 
  4 # merge合并 ，类似于Excel中的vlookup
  5 
  6 df1 = pd.DataFrame({'key': ['K0', 'K1', 'K2', 'K3'],
  7                     'A': ['A0', 'A1', 'A2', 'A3'],
  8                     'B': ['B0', 'B1', 'B2', 'B3']})
  9 df2 = pd.DataFrame({'key': ['K0', 'K1', 'K2', 'K3'],
 10                     'C': ['C0', 'C1', 'C2', 'C3'],
 11                     'D': ['D0', 'D1', 'D2', 'D3']})
 12 df3 = pd.DataFrame({'key1': ['K0', 'K0', 'K2', 'K3'],
 13                     'key2': ['K0', 'K1', 'K0', 'K1'],
 14                     'A': ['A0', 'A1', 'A2', 'A3'],
 15                     'B': ['B0', 'B1', 'B2', 'B3']})
 16 df4 = pd.DataFrame({'key1': ['K0', 'K0', 'K2', 'K3'],
 17                     'key2': ['K0', 'K0', 'K0', 'K0'],
 18                     'C': ['C0', 'C1', 'C2', 'C3'],
 19                     'D': ['D0', 'D1', 'D2', 'D3']})
 20 print(pd.merge(df1,df2,on='key'))
 21 # 第一个DataFrame为拼接后左边的
 22 # 第二个DataFrame为拼接后右边的
 23 # on 为参考键
 24 '''
 25   key   A   B   C   D
 26 0  K0  A0  B0  C0  D0
 27 1  K1  A1  B1  C1  D1
 28 2  K2  A2  B2  C2  D2
 29 3  K3  A3  B3  C3  D3
 30 '''
 31 # 多个键连接
 32 print(pd.merge(df3, df4, on=['key1', 'key2']))
 33 # 当两个DataFrame中的key1和key2都相同时，才会连，否则不连
 34 '''
 35   key1 key2   A   B   C   D
 36 0   K0   K0  A0  B0  C0  D0
 37 1   K0   K0  A0  B0  C1  D1
 38 2   K2   K0  A2  B2  C2  D2
 39 '''
 40 # 参数how  ， 合并方式
 41 # 默认，取交集
 42 print(pd.merge(df3, df4, on=['key1', 'key2'], how='inner'))
 43 print('-' * 8)
 44 '''
 45   key1 key2   A   B   C   D
 46 0   K0   K0  A0  B0  C0  D0
 47 1   K0   K0  A0  B0  C1  D1
 48 2   K2   K0  A2  B2  C2  D2
 49 --------
 50 '''
 51 # 取并集，outer，数据缺失范围NaN
 52 print(pd.merge(df3, df4, on=['key1', 'key2'], how='outer'))
 53 print('-' * 8)
 54 '''
 55   key1 key2    A    B    C    D
 56 0   K0   K0   A0   B0   C0   D0
 57 1   K0   K0   A0   B0   C1   D1
 58 2   K0   K1   A1   B1  NaN  NaN
 59 3   K2   K0   A2   B2   C2   D2
 60 4   K3   K1   A3   B3  NaN  NaN
 61 5   K3   K0  NaN  NaN   C3   D3
 62 --------
 63 '''
 64 # 参照df3为参考合并，数据缺失范围NaN
 65 print(pd.merge(df3, df4, on=['key1', 'key2'], how='left'))
 66 print('-' * 8)
 67 '''
 68   key1 key2   A   B    C    D
 69 0   K0   K0  A0  B0   C0   D0
 70 1   K0   K0  A0  B0   C1   D1
 71 2   K0   K1  A1  B1  NaN  NaN
 72 3   K2   K0  A2  B2   C2   D2
 73 4   K3   K1  A3  B3  NaN  NaN
 74 --------
 75 '''
 76 # 参照df4为参考合并，数据缺失范围NaN
 77 print(pd.merge(df3, df4, on=['key1', 'key2'], how='right'))
 78 print('-' * 8)
 79 '''
 80   key1 key2    A    B   C   D
 81 0   K0   K0   A0   B0  C0  D0
 82 1   K0   K0   A0   B0  C1  D1
 83 2   K2   K0   A2   B2  C2  D2
 84 3   K3   K0  NaN  NaN  C3  D3
 85 --------
 86 '''
 87 # 参数left_on，right_on，left_index, right_index  ,当键不为一个列时，可以单独设置左键与右键
 88 df5 = pd.DataFrame({'lkey': list('bbacaab'),
 89                     'data1': range(7)})
 90 df6 = pd.DataFrame({'rkey': list('abd'),
 91                     'date2': range(3)})
 92 print(df5)
 93 print(df6)
 94 print(pd.merge(df5,df6,left_on='lkey',right_on='rkey'))
 95 '''
 96   lkey  data1
 97 0    b      0
 98 1    b      1
 99 2    a      2
100 3    c      3
101 4    a      4
102 5    a      5
103 6    b      6
104   rkey  date2
105 0    a      0
106 1    b      1
107 2    d      2
108   lkey  data1 rkey  date2
109 0    b      0    b      1
110 1    b      1    b      1
111 2    b      6    b      1
112 3    a      2    a      0
113 4    a      4    a      0
114 5    a      5    a      0
115 '''
116 
117 # concat() 连接，默认axis=0  行+行，当axis=1时，列+列  成为Dataframe
118 s1 = pd.Series([2, 3, 4])
119 s2 = pd.Series([1, 2, 3])
120 print(pd.concat([s1, s2]))
121 '''
122 0    2
123 1    3
124 2    4
125 0    1
126 1    2
127 2    3
128 dtype: int64
129 '''
130 print(pd.concat([s1,s2],axis=1))
131 '''
132    0  1
133 0  2  1
134 1  3  2
135 2  4  3
136 '''
137 snew = pd.concat([s1, s2], axis=1)
138 snew.reset_index(inplace=True)
139 print(snew)
140 '''
141    index  0  1
142 0      0  2  1
143 1      1  3  2
144 2      2  4  3
145 '''
146 snew2 = pd.concat([s1, s2], axis=1)
147 snew2.reset_index(inplace=True, drop=True)
148 print(snew2)
149 '''
150    0  1
151 0  2  1
152 1  3  2
153 2  4  3
154 '''
155 
156 # 去重  .duplicated()
157 s3 = pd.Series([1, 2, 2, 4, 4, 6, 7, 6, 87])
158 # 判断是否重复
159 print(s3.duplicated())
160 '''
161 0    False
162 1    False
163 2     True
164 3    False
165 4     True
166 5    False
167 6    False
168 7     True
169 8    False
170 dtype: bool
171 '''
172 # 取出重复的值
173 s4 = s3[s3.duplicated()]
174 print(s4)
175 # 取出唯一的元素
176 s5 = s3[s3.duplicated() == False]
177 print(s5)
178 '''
179 0     1
180 1     2
181 3     4
182 5     6
183 6     7
184 8    87
185 dtype: int64
186 '''
187 s5 = s3.drop_duplicates()
188 # 可以通过设置参数：inplace控制是否替换原先的值
189 print(s5)
190 '''
191 0     1
192 1     2
193 3     4
194 5     6
195 6     7
196 8    87
197 dtype: int64
198 '''
199 df7 = pd.DataFrame({'key1':['a','a',3,4,3],
200                     'key2':['a','a','b','b',5]})
201 print(df7.duplicated())
202 # 按行检测，第二次出现时，返回True
203 '''
204 0     1
205 1     2
206 3     4
207 5     6
208 6     7
209 8    87
210 dtype: int64
211 '''
212 # 今查看key2列
213 print(df7['key2'].duplicated())
214 '''
215 0    False
216 1     True
217 2    False
218 3     True
219 4    False
220 Name: key2, dtype: bool
221 '''
222 # 直接去重
223 print(df7.drop_duplicates())
224 '''
225   key1 key2
226 0    a    a
227 2    3    b
228 3    4    b
229 4    3    5
230 '''
231 print(df7['key2'].drop_duplicates())
232 '''
233 0    a
234 2    b
235 4    5
236 Name: key2, dtype: object
237 '''
238 
239 # 替换  .replace()
240 s6 = pd.Series(list('askjdghs'))
241 # 一次性替换一个值
242 # print(s6.replace('s','dsd'))
243 '''
244 0      a
245 1    dsd
246 2      k
247 3      j
248 4      d
249 5      g
250 6      h
251 7    dsd
252 dtype: object
253 '''
254 # 一次性替换多个值
255 print(s6.replace(['a','s'],np.nan))
256 '''
257 0    NaN
258 1    NaN
259 2      k
260 3      j
261 4      d
262 5      g
263 6      h
264 7    NaN
265 dtype: object
266 '''
267 # 通过字典的形式替换值
268 print(s6.replace({'a':np.nan}))
269 '''
270 0    NaN
271 1      s
272 2      k
273 3      j
274 4      d
275 5      g
276 6      h
277 7      s
278 dtype: object
279 
280 '''