Pandas DataFrame Repeat Value Based on a Condition

Question

I'm trying to repeat row values in a DataFrame based on conditions in a column. If the value in column Change = 1, then I'd like to repeat the values in columns A, B, and C until the next Change = 1.

index = pandas.date_range('20000131', periods=5)
columns = ['A', 'B', 'C', 'Change']

data = {'A': pandas.Series([False, True, False, True, False], index=index)
    , 'B': pandas.Series([True, True, False, False, False], index=index)
    , 'C': pandas.Series([True, False, True, True, True], index=index)
    , 'Change' : pandas.Series([1,0,0,1,0], index=index)}

Results:

                A      B      C  Change
2000-01-31  False   True   True       1
2000-02-01   True   True  False       0
2000-02-02  False  False   True       0
2000-02-03   True  False   True       1
2000-02-04  False  False   True       0

Desired results:

                A      B      C  Change
2000-01-31  False   True   True       1
2000-02-01  False   True   True       0
2000-02-02  False   True   True       0
2000-02-03   True  False   True       1
2000-02-04   True  False   True       0

This is the closest I've been able to get using shift(), but it only persists for one row. I need it to persist for N number of rows. It breaks down in row three (or row 2 with the 0 base) in the example below.

print pandas.DataFrame(numpy.where(pandas.DataFrame(df['Change']==1)
    , df, df.shift()))

Results:

       0      1      2  3
0  False   True   True  1
1  False   True   True  1
2  False   True  False  0
3   True  False   True  1
4   True  False   True  1

Thank you.

Answer 1

You could fill in the Change == 0 rows with NaN and ffill:

In [11]: df.loc[df.Change != 1, ['A', 'B', 'C']] = numpy.nan

In [12]: df
Out[12]:
             A   B   C  Change
2000-01-31   0   1   1       1
2000-02-01 NaN NaN NaN       0
2000-02-02 NaN NaN NaN       0
2000-02-03   1   0   1       1
2000-02-04 NaN NaN NaN       0

In [13]: df.ffill()
Out[13]:
            A  B  C  Change
2000-01-31  0  1  1       1
2000-02-01  0  1  1       0
2000-02-02  0  1  1       0
2000-02-03  1  0  1       1
2000-02-04  1  0  1       0

If you need these to be bool columns, then use astype(bool) on each column.

As an aside you can nearly this with a resample (except for the last missing rows and Changed column):

In [14]: df[df.Change == 1].resample('D', fill_method='ffill')
Out[14]:
            A  B  C  Change
2000-01-31  0  1  1       1
2000-02-01  0  1  1       1
2000-02-02  0  1  1       1
2000-02-03  1  0  1       1