Pandas ffill and bfill in the same column - conditional

Question

I have a column with chemistries and at some point I have the change of them. I want to ffill until the "changepoint" and bfill after it.

I have several change points like this:

date0   NaN
date1   chem1
date2   NaN
date3   NaN
date4   change
date5   NaN
date6   chem2
date7   chem2
date8   NaN
date9   change
date10   NaN
date11   chem3
date12   NaN

I would like to have it:

date0   chem1
date1   chem1
date2   chem1
date3   chem1
date4   change
date5   chem2
date6   chem2
date7   chem2
date8   chem2
date9   change
date10   chem3
date11   chem3
date12   chem3

any suggestions on the best approach here?

Answer 1

Assuming 'col' your second column, use a custom group:

df['col'] = df.groupby(df['col'].eq('change').cumsum())['col'].apply(lambda s:s.bfill().ffill())

Output;

      date     col
0    date0   chem1
1    date1   chem1
2    date2   chem1
3    date3   chem1
4    date4  change
5    date5   chem2
6    date6   chem2
7    date7   chem2
8    date8   chem2
9    date9  change
10  date10   chem3
11  date11   chem3
12  date12   chem3

Answer 2

df = pd.read_clipboard(header = None, names = ['a', 'b'])

Run an ffill on the second column (I relabelled the columns as a and b):

step1 = df.b.ffill()

Run a condition on step1, and do a bfill after:

step1 = step1.mask(step1.eq('change') & df.b.isna()).bfill()
df.assign(b = step1)

         a       b
0    date0   chem1
1    date1   chem1
2    date2   chem1
3    date3   chem1
4    date4  change
5    date5   chem2
6    date6   chem2
7    date7   chem2
8    date8   chem2
9    date9  change
10  date10   chem3
11  date11   chem3
12  date12   chem3