How to fill values in a Dataframe depending on values around it

Starting from example

import pandas as pd
df=pd.DataFrame({"C":[3, 'X', 'X', 7, 'Y'], 'D':['String', '', '', 'String', '']}) # my own [mre]. You should have included that line in your question ;-)

So df is

   C       D
0  3  String
1  X        
2  X        
3  7  String
4  Y        

(don't worry, the X and Y have no influence on the result. I just included them to imitate your example)

What you are looking for is probably something like:

df['C'] = df.groupby((df['D']=='String').cumsum())['C'].transform('first')

Result:

   C       D
0  3  String
1  3        
2  3        
3  7  String
4  7        

To understand it, it is worth looking at what df['D']=='String').cumsum() does. df['D']=='String' is just a boolean series (True where last column is 'String', False elsewhere). But if you apply .cumsum on such series, it behaves as if True is 1 and False is 0. So what you get is a counter that is incremented each time there is a 'String' and stays as is otherwise. So

>>> (df['D']=='String').cumsum()
0    1
1    1
2    1
3    2
4    2
Name: D, dtype: int64

Which is exactly what you need to group your rows by, to have one group for each row with 'String' and all following rows without (till the next 'String').

Now, just transform C to take the first value of each group, and voila

df['C'] = df.groupby((df['D']=='String').cumsum())['C'].transform('first')

>>> df
   C       D
0  3  String
1  3        
2  3        
3  7  String
4  7        

Another possible solution:

df = df.assign(C = df['C'].where(df['D'].eq('String')).ffill().astype(int))

This creates a new version of df where column C is updated by forward-filling only the numeric values, leaving other values untouched. The df['D'].eq('String') method identifies which entries in column D is 'String'. The where() method replaces non-numeric entries with NaN, and then ffill() propagates the last valid numeric value downward, effectively filling the rows where C was not numeric with the most recent numeric value above it.

Output:

    A  B  C         D
0   1  2  3    String
1  ''  4  3        ''
2  ''  5  3        ''
3  ''  6  7    String
4  ''  1  7        ''

You can select the wanted rows with boolean indexing and reindex with method='ffill':

df['C'] = df.loc[df['D'].eq('String'), 'C'].reindex(df.index, method='ffill')

Alternatively, for fun, and assuming you have an ordered index, you could select the rows to propagate with boolean indexing and combine them to the original input with a merge_asof on the index:

df['C'] = pd.merge_asof(df[[]], df.loc[df['D'].eq('String'), 'C'],
                        left_index=True, right_index=True)

Or as a new DataFrame:

out = (pd.merge_asof(df.drop(columns='C'),
                     df.loc[df['D'].eq('String'), 'C'],
                     left_index=True, right_index=True)
         .reindex_like(df)
      )

Output:

   A  B  C       D
0  1  2  3  String
1     4  3        
2     5  3        
3     6  7  String
4     1  7