Question

例如:

lst = [[ PF2 ,  E1 , -500, -127, 199971, 200164, True, True], 
       [ PR2 ,  E1 , -500, -167, 199655, 200124, True, True],
       [ PF2 ,  E1 , -500, -167, 199645, 200124, False, True],
       [ PF2 ,  E1 , -400, -127, 199971, 200564, True, True], 
       [ PR2 ,  E1 , -400, -167, 199155, 200324, True, True]]

df = pd.DataFrame(lst, columns=["Name", "Part", "Rel_s", 
                                        "Rel_e", "Abs_s", "Abs_e",
                                        "Quality_Start", "Quality_End"])

I want to modify this dataframe to change the values of Abs_s to the smallest for each combination of Part and Rel_s (and the same thing for Abs_e with Part and Rel_e with the largest value). This part works well with this code:

df[ Abs_s ] = df.groupby(["Part", "Rel_s"])[ Abs_s ].transform( min )

df[ Abs_e ] = df.groupby(["Part", "Rel_e"])[ Abs_e ].transform( max )

如同这一解决办法一样,因为这种办法似乎简单易懂;然而,我也想考虑质量价值,以便我在<条码>、Abs_s值><条码_Start:和最大<条码>>Abs_e><条码>>上填写<条码>。因此,此处为E1,-500, 更正Abs_s。页: 1 如果没有良好的质量(<>Tru)值以取代,则应保持Bad Quality(False)。

我能否补充这些条件? 我如何做这一转变?

Answer 1

您可以将价值替换为<代码>。正确的s in Quality_Start and Quality_End to Disappearance first, here in Series. where :

g = (df.assign(Abs_s=df[ Abs_s ].where(df[ Quality_Start ]),
               Abs_e=df[ Abs_e ].where(df[ Quality_End ]))
       .groupby(["Part", "Rel_s"]))
    
df[ Abs_s ] = g[ Abs_s ].transform( min )
df[ Abs_e ] = g[ Abs_e ].transform( max )

print (df)    
  Name Part  Rel_s  Rel_e     Abs_s   Abs_e  Quality_Start  Quality_End
0  PF2   E1   -500   -127  199655.0  200564           True         True
1  PR2   E1   -500   -167  199655.0  200324           True         True
2  PF2   E1   -500   -167  199655.0  200324          False         True
3  PF2   E1   -400   -127  199155.0  200564           True         True
4  PR2   E1   -400   -167  199155.0  200324           True         True

www.un.org/spanish/ecosoc 如何工作:

print (df.assign(Abs_s=df[ Abs_s ].where(df[ Quality_Start ]),
                 Abs_e=df[ Abs_e ].where(df[ Quality_End ])))  

  Name Part  Rel_s  Rel_e     Abs_s   Abs_e  Quality_Start  Quality_End
0  PF2   E1   -500   -127  199971.0  200164           True         True
1  PR2   E1   -500   -167  199655.0  200124           True         True
2  PF2   E1   -500   -167       NaN  200124          False         True
3  PF2   E1   -400   -127  199971.0  200564           True         True
4  PR2   E1   -400   -167  199155.0  200324           True         True

如果需要电离层的分类,可按<代码>将具有缺失值的一栏转换成。 Int64 :

g = (df.assign(Abs_s=df[ Abs_s ].where(df[ Quality_Start ]).astype( Int64 ),
               Abs_e=df[ Abs_e ].where(df[ Quality_End ]).astype( Int64 ))
       .groupby(["Part", "Rel_s"]))

df[ Abs_s ] = g[ Abs_s ].transform( min )
df[ Abs_e ] = g[ Abs_e ].transform( max )

print (df)    
  Name Part  Rel_s  Rel_e   Abs_s   Abs_e  Quality_Start  Quality_End
0  PF2   E1   -500   -127  199655  200164           True         True
1  PR2   E1   -500   -167  199655  200164           True         True
2  PF2   E1   -500   -167  199655  200164          False         True
3  PF2   E1   -400   -127  199155  200564           True         True
4  PR2   E1   -400   -167  199155  200564           True         True

www.un.org/spanish/ecosoc 如何工作:

print (df.assign(Abs_s=df[ Abs_s ].where(df[ Quality_Start ]).astype( Int64 ),
                 Abs_e=df[ Abs_e ].where(df[ Quality_End ]).astype( Int64 )))  

  Name Part  Rel_s  Rel_e   Abs_s   Abs_e  Quality_Start  Quality_End
0  PF2   E1   -500   -127  199655  200164           True         True
1  PR2   E1   -500   -167  199655  200164           True         True
2  PF2   E1   -500   -167    <NA>  200164          False         True
3  PF2   E1   -400   -127  199155  200564           True         True
4  PR2   E1   -400   -167  199155  200564           True         True

a. 测试,不匹配:

lst = [[ PF2 ,  E1 , -500, -127, 199971, 200164, False, True], 
       [ PR2 ,  E1 , -500, -167, 199655, 200124, False, True],
       [ PF2 ,  E1 , -500, -167, 199645, 200124, False, True],
       [ PF2 ,  E1 , -400, -127, 199971, 200564, True, True], 
       [ PR2 ,  E1 , -400, -167, 199155, 200324, True, True]]

df = pd.DataFrame(lst, columns=["Name", "Part", "Rel_s", 
                                        "Rel_e", "Abs_s", "Abs_e",
                                        "Quality_Start", "Quality_End"])


g = (df.assign(Abs_s=df[ Abs_s ].where(df[ Quality_Start ]).astype( Int64 ),
               Abs_e=df[ Abs_e ].where(df[ Quality_End ]).astype( Int64 ))
       .groupby(["Part", "Rel_s"]))

df[ Abs_s ] = g[ Abs_s ].transform( min )
df[ Abs_e ] = g[ Abs_e ].transform( max )

print (df)    
  Name Part  Rel_s  Rel_e   Abs_s   Abs_e  Quality_Start  Quality_End
0  PF2   E1   -500   -127    <NA>  200164          False         True
1  PR2   E1   -500   -167    <NA>  200164          False         True
2  PF2   E1   -500   -167    <NA>  200164          False         True
3  PF2   E1   -400   -127  199155  200564           True         True
4  PR2   E1   -400   -167  199155  200564           True         True

print (df.assign(Abs_s=df[ Abs_s ].where(df[ Quality_Start ]).astype( Int64 ),
                 Abs_e=df[ Abs_e ].where(df[ Quality_End ]).astype( Int64 )))  
  Name Part  Rel_s  Rel_e   Abs_s   Abs_e  Quality_Start  Quality_End
0  PF2   E1   -500   -127    <NA>  200164          False         True
1  PR2   E1   -500   -167    <NA>  200164          False         True
2  PF2   E1   -500   -167    <NA>  200164          False         True
3  PF2   E1   -400   -127  199155  200564           True         True
4  PR2   E1   -400   -167  199155  200564           True         True

EDIT:

#one groupby with created 2 temporary columns
g = (df.assign(Abs_s_t=df[ Abs_s ].where(df[ Quality_Start ]).astype( Int64 ),
               Abs_e_t=df[ Abs_e ].where(df[ Quality_End ]).astype( Int64 ))
       .groupby(["Part", "Rel_s"])) 
 
#if need replace by original values
df[ Abs_s ] = g[ Abs_s_t ].transform( min ).fillna(df[ Abs_s ]) 
df[ Abs_e ] = g[ Abs_e_t ].transform( max ).fillna(df[ Abs_e ])

#if need replace by min/max values per groups
df[ Abs_s1 ] = g[ Abs_s_t ].transform( min ).fillna(g[ Abs_s ].transform( min )) 
df[ Abs_e1 ] = g[ Abs_e_t ].transform( max ).fillna(g[ Abs_e ].transform( max )) 

print (df)    
  Name Part  Rel_s  Rel_e   Abs_s   Abs_e  Quality_Start  Quality_End  Abs_s1  Abs_e1
0  PF2   E1   -500   -127  199971  200164          False         True  199645  200164
1  PR2   E1   -500   -167  199655  200164          False         True  199645  200164
2  PF2   E1   -500   -167  199645  200164          False         True  199645  200164
3  PF2   E1   -400   -127  199155  200564           True         True  199155  200564
4  PR2   E1   -400   -167  199155  200564           True         True  199155  200564

Answer 2

某些组别的过滤器将作trick,只有情况确实变化的数值。质量法的所有其他价值保持不变。

start = df[ Quality_Start ]
end = df[ Quality_End ]
df.loc[start,  Abs_s ] = df[start].groupby(["Part", "Rel_s"])[ Abs_s ].transform( min )
df.loc[end,  Abs_e ] = df[end].groupby(["Part", "Rel_s"])[ Abs_e ].transform( max )

End result:

	Name	Part	Rel_s	Rel_e	Abs_s	Abs_e	Quality_Start	Quality_End
0	PF2	E1	-500	-127	199655	200164	True	True
1	PR2	E1	-500	-167	199655	200164	True	True
2	PF2	E1	-500	-167	199645	200164	False	True
3	PF2	E1	-400	-127	199155	200564	True	True
4	PR2	E1	-400	-167	199155	200564	True	True

友情链接