Interpolate a Pandas.DataFrame to an input index, drop duplicated index, with scipy.interpolate methods.

output.png

reindex_and_interpolate_with_pandas.py

def reindex_and_interpolate(df,
                            new_index,
                            method,
                            extrapolation_warnings=True):
    """Interpolate an input Pandas.DataFrame to an input index.
        Uses scipy.interpolate methods > https://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html
        Handles extrapolation before and after the original X.
        Works only for pandas version >=0.23.0.
        
        Adapted from https://stackoverflow.com/a/51535881/1435167

    Parameters
    ----------
    df : pandas.DataFrame
        df input pandas.DataFrame, index need to be numbers. 
    new_index : Index or array-like
        new index, anthing that pandas.Index.union accepts.
    method : str
        one of scipy.interpolate method 
    extrapolation_warnings : Bool
        raise a warnings if extrapolation happens (new_index "bigger" than df.index)
    """
    kw = dict(method=method, fill_value="extrapolate",limit_area=None,limit_direction='both')
    # add new index, interpolate on Nan values on new_index,select only new_index
    out_df = df.reindex(df.index.union(new_index)).interpolate(**kw).loc[new_index]
    # drop duplicated indexes 
    out_df = out_df.loc[out_df.index.drop_duplicates(keep='first')]
    return out_df

## simple reindexing example
# test dataframe
f = pd.DataFrame(data=[0.0,3.0,2.0,7.0,3.5,6.0],columns=['a'],index=np.arange(6))
ax = f.rename(columns={'a':'original'}).plot(marker='.',markersize=30,linewidth=6,color='black',figsize=[10,4])

new_index = np.concatenate([[-0.25],f.index.to_numpy()+0.25])
f_reindex_ls = [
                reindex_and_interpolate(f, new_index ,method=method).rename(columns={'a':method}) 
                for method in ['nearest',  'zero', 'slinear', 'quadratic', 'cubic']
               ]

f_reindex = pd.concat(f_reindex_ls,axis=1)
f_reindex.plot.line(marker='.',ax=ax,markersize=20,linewidth=4)
print(f)
print(f_reindex)
plt.tight_layout()
plt.savefig('output.png',dpi=100,facecolor='white', transparent=False)