flatten-df-columns

python

def flatten_dataframe(df, columns, fillna_value='', preserve_index=False):
    if (columns is not None \
        and len(columns) > 0 \
        and not isinstance(columns, (list, tuple, np.ndarray, pd.Series))):
        columns = [columns]

    diff_columns = df.columns.difference(columns)

    # get column's series of length
    series_length = df[columns[0]].str.len()

    length_gt_zero = (series_length > 0)
    length_zero = ~length_gt_zero

    indexes = np.repeat(df.index.values, series_length)

    data_need_to_fill = {col: np.repeat(df[col].values, series_length) for col in diff_columns}
    df_need_to_fill = pd.DataFrame(data_need_to_fill, index=indexes)

    data_origin = {col: np.concatenate(df.loc[length_gt_zero, col].values) for col in columns}
    df_flatten = df_need_to_fill.assign(**data_origin)

    # append rows has empty lists
    if length_zero.any():
        df_flatten = (df_flatten.append(df.loc[length_zero, diff_columns], sort=False).fillna(fillna_value))

    df_flatten = df_flatten.sort_index()

    if not preserve_index:
        df_flatten = df_flatten.reset_index(drop=True)

    return df_flatten