检查 dataframe 中的每一行和每一列，并用用户定义的 function 替换值

Question

df=pd.DataFrame({'0':["qwa-abc","abd-xyz","abt-Rac","xyz-0vc"],'1':['axc-0aa',"abd-xyz","abt-Rac","xyz-1avc"],
                '3':['abc-aaa',"NaN","abt-9ac","xyz-9vc"]})

I have this DataFrame, I want to check each row and each column for a specific value. for example index 0 there are 4 values "qwa-abc","abd-xyz","abt-Rac","xyz-0avc".

for every value I want to check if xxx- any number xx.

example:

qua-abc has a at the position 4, so do nothing. when it reach to xyz- 0 ac there is number 0 at position 4. hence it should run user define function to replace whole value(xyz-0vc) to whatever the user define function get.

NOTE: I tried running str.replace but it only supports specific user define string. here user function will connect to different system and get a string. hence it's not predefine .

Answer 1

If you want to change all the cells in your Dataframe you need to use pd.apply over the row axis, so your custom function needs to take a pd.Series as one of the parameters. In this example row is the series.

This generator function iterates over each cell in the row, checks if the character at index 4 is numeric. If true returns the value to replace string with, otherwise it will return the value of the cell itself.

def replace_value(row, value):
    for cell in row:
        if pd.notna(cell) and cell[4].isnumeric():
            yield value
        else:
            yield cell

df.apply(lambda x: pd.Series(replace_value(x, 'myvalue')), axis=1)

You then apply your custom function row wise, ( axis=1 ) and wrap it in a lambda so you can pass additional arguments ( value in this case) and then call pd.Series on the iterator returned by the function.

Hope it makes sense.

Answer 2

You don't need a separate method, try this:

In [1200]: df.loc[df['0'].str[4].str.isdigit(), '0'] = 'myvalue'                                                                                                                                            

In [1201]: df                                                                                                                                                                                               
Out[1201]: 
         0         1        3
0  qwa-abc   axc-0aa  abc-aaa
1  abd-xyz   abd-xyz      NaN
2  abt-Rac   abt-Rac  abt-9ac
3  myvalue  xyz-1avc  xyz-9vc

For doing this in all columns, do this:

In [1242]: def check_digit(cols,new_val): 
      ...:     for i in cols: 
      ...:         df.loc[(df[i].str[4].str.isdigit()) & (df[i].notna()), i] = new_val 
      ...:  

In [1243]: df.apply(lambda x: check_digit(df.columns, 'myval'), 1)

In [1244]: df                                                                                                                                                                                               
Out[1244]: 
         0        1        3
0  qwa-abc    myval  abc-aaa
1  abd-xyz  abd-xyz      NaN
2  abt-Rac  abt-Rac    myval
3    myval    myval    myval

Answer 3

This answer is based on @NomadMonad

string_replacer() is a function that will change value based on input value that satisfies condition

def replace_value(row, value): for cell in row: try: if pd.notna(cell) and cell[4].isnumeric(): value=string_replacer(cell) yield value else: yield cell except: print(row,value) if_df.apply(lambda x: pd.Series(replace_value(x,value)), axis=1)