Question

我正在尝试将csv中的多个文件导入单个DataFrame，并在尝试添加第三个DataFrame时收到以下错误。

AssertionError: cannot create BlockManager._ref_locs because block [ObjectBlock: [CompletionDate, Categories, DateEntered_x, <lots more columns here>...], dtype=object)] does not have _ref_locs set

代码是：

project = pandas.read_csv(read_csv('dbo_Project.csv')
project = pandas.read_csv(read_csv('dbo_ProjectEnergy.csv')
project = pandas.read_csv(read_csv('dbo_BuildingDescription.csv')
part_merged = pandas.merge(project, project_energy,
    on='ProjectID',
    how='outer')
part_merged = pandas.merge(part_merged, project_energy_data,
    on='ProjectEnergyID',
    how='outer')
part_merged = pandas.merge(part_merged, building_description,
    on='ProjectEnergyID',
    how='outer')

我应该如何加入这些DataFrames以避免此问题？

根据Stefan Jansen的回答编辑：

直到发生新错误的新代码为：

project = pandas.read_csv(read_csv('dbo_Project.csv')
project = pandas.read_csv(read_csv('dbo_ProjectEnergy.csv')
part_merged = pandas.concat([project, project_energy],
    axis=1,
    join='outer')
part_merged.set_index(['ProjectEnergyID'])
part_merged = pandas.concat([self.part_merged,
    project_energy_data],
    axis=1,
    join='outer')

Answer 1

一个很简单的答案。

问题是重复的列。导致问题的列并不重要，因此我在合并之前将它们删除了。

def remove_clashes(df):
    unwanted_cols = ['DataCompleteness', 'DeletedFlag','DateEntered', 'EnteredBy',
        'LastModified', 'MandatoryDataInput', 'ModifiedBy']
    return df.drop([col for col in unwanted_cols if col in df.columns], axis=1)

Answer 2

我更喜欢将pandas.concat()用于多个帧。还有'outer'选项 - see documentation.

如果要合并的列是索引列，这可以很好地工作，您可以使用pandas.set_index()来实现，可能先于.reset_index()。

合并Pandas中的数据帧会引发'AssertionError：无法创建BlockManager._ref_locs'

2 个答案: