多头 pandas 数据帧

Question

是否可以使用 pandas.DataFrame 复制此 csv 的结构？

所有数据都从一个 HDF5 文件中提取，然后将属性解析到 pd.DataFrame

我的担忧 是 meta header 和 meta data（csv 中的第 1 行和第 2 行）与 attribute header 和attribute data 长度或形状。

我是这样称呼 pd.DataFrame:

    # Meta Pandas DataFrame
    meta_df = pd.DataFrame(index=range(0, 8760, 24), columns=['source', 'location_id', 'state', 'country', 'latitude', 
                                                              'longitude', 'time_zone', 'elevation', 'clearsky_dhi', 
                                                              'clearsky_dni', 'clearsky_ghi', 'dewpoint_unit', 
                                                              'temperature_unit'])
    # Meta Header & Data
    meta_df['source'] = source
    meta_df['location_id'] = location_id
    meta_df['state'] = state
    meta_df['country'] = country
    meta_df['latitude'] = latitude
    meta_df['longitude'] = longitude
    meta_df['time_zone'] = local_time
    meta_df['elevation'] = elevation
    meta_df['clearsky_dhi'] = clearsky_dhi
    meta_df['clearsky_dni'] = clearsky_dni
    meta_df['clearsky_ghi'] = clearsky_ghi
    meta_df['dewpoint_unit'] = dewpoint_unit
    meta_df['temperature_unit'] = temperature_unit

    # Attribute Pandas DataFrame
    att_df = pd.DataFrame(index=range(0, 8760, 24), columns=['dhi', 'dni', 'ghi', 'source', 'dew_point', 'temperature'])

    # Attribute Header & Data
    att_df['year'] = year
    att_df['month'] = month
    att_df['day'] = day
    att_df['hour'] = hour
    att_df['minute'] = minute
    att_df['dhi'] = dhi
    att_df['dni'] = dni
    att_df['ghi'] = ghi
    att_df['dew_point'] = dew_point
    att_df['temperature'] = temperature

    # Make one DataFrame with multiple headers?
    # Do something, then export to csv.
    df.to_csv(ndir_root + ndir + '/' + fname + '.csv', index=False)

是否最好创建两个单独的数据帧，然后将它们垂直堆叠以创建第三个数据帧并将最后一个数据帧导出为 csv？

布勒?

Answer 1

我认为您可以通过 .to_csv() 执行此操作，因为此方法接受文件路径（如您所做的那样）或缓冲区。我假设您知道元 header、元数据和属性 header 字符串的顺序，因此您可以选择将它们写入文件的方式。您缺少的部分如下所示。

with open('output.csv','w') as fid:
    # write your meta header etc., here assumed to be a list of strings
    fid.write(','.join(meta_header) + '\n')
    fid.write(','.join(meta_data) + '\n')
    fid.write(','.join(attribute_header) + '\n')

    # now write attr_df to a csv by passing data to your fid buffer
    attr_df.to_csv(fid, sep=',', header=False, index=False)

多头 pandas 数据帧

multiheader pandas dataframe

python

csv

hdf5

python-2.7

pandas