pandas26 combine通過函式組合2個數據幀（ tcy）
阿新 • • 發佈：2019-01-01
combine通過函式組合2個數據幀2018/12/31 =================================================================================== df1.combine(other, func, fill_value=None, overwrite=True) 用途：通過函式組合2個數據幀 # 不傳播NaN值，如果1個數據缺少列用其他資料的列值替代（也可能是NaN） 引數 # other：DataFrame # func：函式#將兩個系列作為輸入並返回一個Series或一個標量 # # fill_value：標量值;表示na值或補齊的資料長度 # overwrite=True：如為True則覆蓋呼叫幀中公共鍵的值 （本引數實驗發現無什麼變化歡迎指正） 返回：組合DataFrame ==================================================================================== # 例項1.1：df1 if df1>=df2 else df2 df1 = pd.DataFrame({'A': [10, 11], 'B': [12, 13]}) df2 = pd.DataFrame({'A': [21, -22], 'B': [23, -24]}) result=df1[df1>=df2].append(df2[df1<df2]) # df1 df2 result A B A B A B 0 10 12 0 21 23 0 NaN NaN 1 11 13 1 -22 -24 1 11.0 13.0 0 21.0 23.0 1 NaN NaN # 例項1.df1 if df1>=df2 else df2 df1.combine(df2, lambda s1, s2: s1 if s1.any() < s2.any() else s2) # df1 df2 result A B A B A B 0 10 12 0 21 23 0 21 23 1 11 13 1 -22 -24 1 -22 -24 # 例項1.3： result=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2])) # df1 df2 result A B A B A B 0 10 12 0 21 23 0 21 23 1 11 13 1 -22 -24 1 11 13 =================================================================================== # 例項2.1： df1 = pd.DataFrame({'A': [10, 11], 'B': [12, 13],'C': [14, 15]}) df2 = pd.DataFrame({'A': [21, -22], 'B': [23, -24]}) result=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2])) result1=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2]),fill_value=-1) result2=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2]),fill_value=30) #df1 df2 result result1 result2 A B C A B A B C A B C A B C 0 10 12 14 0 21 23 0 21 23 NaN 0 21 23 14 0 21 23 30 1 11 13 15 1 -22 -24 1 11 13 NaN 1 11 13 15 1 11 13 30 # 例項2.2：fill_value補齊資料長度 df1 = pd.DataFrame({'A': [10, 11], 'B': [12, 13]}) df2 = pd.DataFrame({'A': [21, -22], 'B': [23, -24],'C': [25, 26]}) result=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2])) result1=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2]),fill_value=-1) result2=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2]),fill_value=30) #df1 df2 result result1 result2 A B A B C A B C A B C A B C 0 10 12 0 21 23 25 0 21 23 NaN 0 21 23 25.0 0 21 23 30.0 1 11 13 1 -22 -24 26 1 11 13 NaN 1 11 13 26.0 1 11 13 30.0 ================================================================================== # 例項3：na值 df1 = pd.DataFrame({'A': [10, 11], 'B': [12, np.nan],'C': [14, np.nan]}) df2 = pd.DataFrame({'A': [21, -22], 'B': [23, np.nan],'C': [25, 26]}) df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2])) #df1 df2 result A B C A B C A B C 0 10 12.0 14.0 0 21 23.0 25 0 21 23.0 25.0 1 11 NaN NaN 1 -22 NaN 26 1 11 NaN NaN # 例項4：na值 fill_value 替代na值 result=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2]),fill_value=-1) result1=df1.combine(df2, lambda s1, s2: s1[s1>=s2].append(s2[s1<s2]),fill_value=30) #df1 df2 result result1 A B C A B C A B C A B C 0 10 12.0 14.0 0 21 23.0 25 0 21 23.0 25.0 0 21 23.0 25.0 1 11 NaN NaN 1 -22 NaN 26 1 11 -1.0 26.0 1 11 30.0 30.0 =================================================================================