python - Fill missing dates by group in pandas -

i need fill missing date down group. here code create data frame. want add date of fill column down far when date of fill column changes , until group 'name' changes.

    data = {'tdate': [20080815,20080915,20081226,20090110,20090131,20080807,20080831,     20080918,20081023,20081114,20081207,20090117,20090203,20090219,20090305,20090318,20090501],         'name': ['a','a','a','a','a','b','b','b','b','b','b','b','b','b','b','b','b'],     'fill': [nan,nan,20080915,nan,nan,nan,nan,nan,nan,20081023,              nan,nan,nan,nan,20090219,nan,nan]}      df = pd.dataframe(data, columns=['tdate', 'name', 'fill'])     df 

current data frame

tdate   name    fill 0    20080815      nan 1    20080915      nan 2    20081226      20080915 3    20090110      nan 4    20090131      nan 5    20080807    b   nan 6    20080831    b   nan 7    20080918    b   nan 8    20081023    b   nan 9    20081114    b   20081023 10   20081207    b   nan 11   20090117    b   nan 12   20090203    b   nan 13   20090219    b   nan 14   20090305    b   20090219 15   20090318    b   nan 16   20090501    b   nan 

desired output

    tdate   name    fill 0    20080815      nan 1    20080915      nan 2    20081226      20080915 3    20090110      20080915 4    20090131      20080915 5    20080807    b   nan 6    20080831    b   nan 7    20080918    b   nan 8    20081023    b   nan 9    20081114    b   nan 10   20081207    b   20081023 11   20090117    b   20081023 12   20090203    b   20081023 13   20090219    b   20081023 14   20090305    b   20081023 15   20090318    b   20090219 16   20090501    b   20090219 

here code

df.groupby(df["name"])["fill"].fill() 

you pretty close, need forward-fill rather filling:

df.groupby('name')["fill"].ffill() out[42]:  0          nan 1          nan 2     20080915 3     20080915 4     20080915 5          nan 6          nan 7          nan 8          nan 9     20081023 10    20081023 11    20081023 12    20081023 13    20081023 14    20090219 15    20090219 16    20090219 dtype: float64 

or equivalently:

df.groupby('name')["fill"].fillna(method='ffill')