ENH: DataFrame.interpolate limit to support all-or-none filling

[aaronsl-hku]

Currently with df.interpolate(limit, limit_direction) , I must choose 1 or both sides to fill when I am limiting the interpolation. What I find more useful is a all-or-none strategy rather than only fill up to the limit count, so I can fill up some short-term missing data and keep long-term missing data to be filtered after. Demonstrated as here:

>>> df = pd.DataFrame([[0,1,2,3],[1,np.nan,np.nan,np.nan],[np.nan,np.nan,np.nan,5],[3,4,5,6]],columns=list('abcd'))
>>> df
     a    b    c    d
0  0.0  1.0  2.0  3.0
1  1.0  NaN  NaN  NaN
2  NaN  NaN  NaN  5.0
3  3.0  4.0  5.0  6.0
>>> # Current options
>>> df.interpolate(axis=0,limit=1,limit_direction='forward')
     a    b    c    d
0  0.0  1.0  2.0  3.0
1  1.0  2.0  3.0  4.0
2  2.0  NaN  NaN  5.0
3  3.0  4.0  5.0  6.0
>>> df.interpolate(axis=0,limit=1,limit_direction='backward')
     a    b    c    d
0  0.0  1.0  2.0  3.0
1  1.0  NaN  NaN  4.0
2  2.0  3.0  4.0  5.0
3  3.0  4.0  5.0  6.0
>>> df.interpolate(axis=0,limit=1,limit_direction='both')
     a    b    c    d
0  0.0  1.0  2.0  3.0
1  1.0  2.0  3.0  4.0
2  2.0  3.0  4.0  5.0
3  3.0  4.0  5.0  6.0
>>> # What is desired
>>> interpolated_df = pd.DataFrame([[0,1,2,3],[1,np.nan,np.nan,4],[2,np.nan,np.nan,5],[3,4,5,6]],columns=list('abcd'))
>>> interpolated_df # NaNs at column b and c not filtered for exceeding limit 1
   a    b    c  d
0  0  1.0  2.0  3
1  1  NaN  NaN  4
2  2  NaN  NaN  5
3  3  4.0  5.0  6

[joAschauer]

Hi there, have a look at #36352 and #25141.

[aaronsl-hku]

Hi there, have a look at #36352 and #25141.

Hello, thanks I just read into them. I find this different from #36352 as that was reported as BUG while I am suggesting it as an ENH (as extra functionality).

#25141 was quite align with my thoughts... but was it closed to give way to #36352 ?