process_hdf.py 文件源码

def nan_interpolate(df):
    """

    Reference:
    https://stackoverflow.com/questions/29007830/identifying-consecutive-nans-with-pandas
    """
    sum_nan = df.isnull().sum()
    df_null_int = df.isnull().astype(int)
    for col in df.columns:
        max_run = df[col].isnull().astype(int).groupby(df[col].notnull().astype(int).cumsum()).sum()
        if sum_nan[col]:
            # BELOW IS BROKEN!!!
            pass
            # logger.warning('column {} has {} NaNs ({} max consecutive run)'.format(col,
                                                                                   # sum_nan[col],
                                                                                   # max_run))
    df.interpolate(inplace=True)
    return df