Python 使用重复的标题值读取excel

我有一个excel表格，我想读入熊猫多索引数据框。复杂的是excel工作表包含重复的标题值。读取时，pandas会将.x添加到第二级标题的末尾，而不是第一级标题的末尾。有没有办法重命名顶级标题而不是第二级标题

excel文件示例：

阅读脚本：

从pathlib导入路径
作为pd进口熊猫
def main（）：
xl_file=Path（'.'）/“pandasExample.xlsx”
df=pd.read\u excel（xl\u文件，工作表名称='Sheet1'，页眉=[
0,1]，skiprows=[0]）
打印（df）
如果uuuu name uuuuuu='\uuuuuuu main\uuuuuuu'：
main（）

输出：

  Rectangle        Ellipse    Rectangle
      Width Height       a  b   Width.1 Height.1 Width.2 Height.2
0        10     20       1  2        20       30      40       50

期望输出：

  Rectangle        Ellipse    Rectangle.1        Rectangle.2       
      Width Height       a  b      Width Height      Width Height
0        10     20       1  2         20     30         40     50

---

取消堆叠数据帧，然后将级别0重新指定给唯一标签。我是手工做的，但是你可以通过在每两列中添加一个后缀来编程。设置多重索引，然后堆叠结果。元组中有三个值：级别0、级别1和0

 df=pd.read_excel('dup_header.xls',skiprows=2,nrows=10)
 unstack_df=df.stack()
 unstack_df=unstack_df.reset_index()
 unstack_df['level_0']=['Rectangle1','Rectangle1','Ellipse','Ellipse','Rectangle2','Rectangle2','Rectangle3','Rectangle3']
 unstack_df=unstack_df.set_index(['level_0','level_1'])
 stack_series=unstack_df.stack()

 df=stack_series.to_frame()
 df.columns=['value']
 #print(df.index)
 #print(df.values)
 print(df)

输出：

                             value
  level_0     level_1    
  Rectangle1  Width     0    10
              Height    0    20
  Ellipse     a         0     1
              b         0     2
  Rectangle2  width     0    20
              height    0    30
  Rectangle3  width.1   0    40
              height.1  0    50

---

取消堆叠数据帧，然后将级别0重新指定给唯一标签。我是手工做的，但是你可以通过在每两列中添加一个后缀来编程。设置多重索引，然后堆叠结果。元组中有三个值：级别0、级别1和0

 df=pd.read_excel('dup_header.xls',skiprows=2,nrows=10)
 unstack_df=df.stack()
 unstack_df=unstack_df.reset_index()
 unstack_df['level_0']=['Rectangle1','Rectangle1','Ellipse','Ellipse','Rectangle2','Rectangle2','Rectangle3','Rectangle3']
 unstack_df=unstack_df.set_index(['level_0','level_1'])
 stack_series=unstack_df.stack()

 df=stack_series.to_frame()
 df.columns=['value']
 #print(df.index)
 #print(df.values)
 print(df)

输出：

                             value
  level_0     level_1    
  Rectangle1  Width     0    10
              Height    0    20
  Ellipse     a         0     1
              b         0     2
  Rectangle2  width     0    20
              height    0    30
  Rectangle3  width.1   0    40
              height.1  0    50

---

这里有一个不同的答案，可以产生问题中列出的精确的期望输出

from pathlib import Path
import pandas as pd
from typing import List


def rename_headers(headers: List[str]) -> List[str]:
    header_dict = {}
    new_headers = []
    for header in headers:
        header_prefix = header.split('.')[0]
        header_occurance = header_dict.get(header_prefix, 0)
        if header_occurance > 0:
            new_header = header_prefix + f'.{header_occurance}'
        else:
            new_header = header_prefix
        new_headers.append(new_header)
        header_occurances[header_prefix] = header_occurance + 1
    return new_headers

def main():
    xl_file = Path('.') / 'pandasExample.xlsx'

    # Read first level headers
    header_df = pd.read_excel(xl_file, sheet_name='Sheet1', header=[
        0], skiprows=[0], nrows=1)
    headers = list(filter(lambda x: not x.startswith(
        'Unnamed'), list(header_df.columns)))

    # Generate the desired headers
    new_headers = rename_headers(headers)

    # Read in the full dataframe
    df = pd.read_excel(xl_file, sheet_name='Sheet1', header=[
        0, 1], skiprows=[0])

    # Create a dictionary that identifies the parameters for each unique header
    unique_headers = pd.unique(pd.Index(df.columns.get_level_values(0)))
    parameters = {}
    for header in unique_headers:
        parameters[header] = pd.unique(
            [column.split('.')[0] for column in df[header].columns])


    unstack_df = df.head(1).stack()
    # Keep order of the original index after stack
    index = df.head(1).unstack().index.get_level_values(1)
    unstack_df = unstack_df.reindex(zip([0] * len(index), index))
    unstack_df = unstack_df.reset_index()

    # Create the new level 0 and level 1 headers
    level_0 = []
    for header in new_headers:
        level_0 += [header] * len(parameters[header.split('.')[0]])
    level_1 = [parameter.split('.')[0] for parameter in unstack_df['level_1']]

    # Rename level 0 and level 1 columns for the dataframe
    df.columns = pd.MultiIndex.from_tuples(zip(level_0, level_1))
    print(df)


if __name__ == '__main__':
    main()

输出：

  Rectangle        Ellipse    Rectangle.1        Rectangle.2       
      Width Height       a  b       Width Height       Width Height
0        10     20       1  2          20     30          40     50

---

这里有一个不同的答案，可以产生问题中列出的精确的期望输出

from pathlib import Path
import pandas as pd
from typing import List


def rename_headers(headers: List[str]) -> List[str]:
    header_dict = {}
    new_headers = []
    for header in headers:
        header_prefix = header.split('.')[0]
        header_occurance = header_dict.get(header_prefix, 0)
        if header_occurance > 0:
            new_header = header_prefix + f'.{header_occurance}'
        else:
            new_header = header_prefix
        new_headers.append(new_header)
        header_occurances[header_prefix] = header_occurance + 1
    return new_headers

def main():
    xl_file = Path('.') / 'pandasExample.xlsx'

    # Read first level headers
    header_df = pd.read_excel(xl_file, sheet_name='Sheet1', header=[
        0], skiprows=[0], nrows=1)
    headers = list(filter(lambda x: not x.startswith(
        'Unnamed'), list(header_df.columns)))

    # Generate the desired headers
    new_headers = rename_headers(headers)

    # Read in the full dataframe
    df = pd.read_excel(xl_file, sheet_name='Sheet1', header=[
        0, 1], skiprows=[0])

    # Create a dictionary that identifies the parameters for each unique header
    unique_headers = pd.unique(pd.Index(df.columns.get_level_values(0)))
    parameters = {}
    for header in unique_headers:
        parameters[header] = pd.unique(
            [column.split('.')[0] for column in df[header].columns])


    unstack_df = df.head(1).stack()
    # Keep order of the original index after stack
    index = df.head(1).unstack().index.get_level_values(1)
    unstack_df = unstack_df.reindex(zip([0] * len(index), index))
    unstack_df = unstack_df.reset_index()

    # Create the new level 0 and level 1 headers
    level_0 = []
    for header in new_headers:
        level_0 += [header] * len(parameters[header.split('.')[0]])
    level_1 = [parameter.split('.')[0] for parameter in unstack_df['level_1']]

    # Rename level 0 and level 1 columns for the dataframe
    df.columns = pd.MultiIndex.from_tuples(zip(level_0, level_1))
    print(df)


if __name__ == '__main__':
    main()

输出：

  Rectangle        Ellipse    Rectangle.1        Rectangle.2       
      Width Height       a  b       Width Height       Width Height
0        10     20       1  2          20     30          40     50

---

这不适用于多索引头，如我的示例所示<代码>值错误：指定多索引标题时无法指定名称跳过标题级别0并将其替换为唯一的标签，然后在数据帧中设置多索引，然后堆叠值，然后将结果转换为数据帧，请参见上文，我想这是可行的。必须编写一些额外的代码来计算标题值。这不适用于多索引标题，如我的示例所示<代码>值错误：指定多索引标题时无法指定名称跳过标题级别0并将其替换为唯一的标签，然后在数据帧中设置多索引，然后堆叠值，然后将结果转换为数据帧，请参见上文，我想这是可行的。必须编写一些额外的代码来计算标题值。