Python Sunburst Chart - 将数据框转换为嵌套列表格式

Question

我已经按照下面提到的链接在 Python 中创建 Sunburst 图表： How to make a sunburst plot in R or Python?

附上笔记本供参考。

但是，创建图表的函数需要特定格式的数据（按级别嵌套的列表）。示例：

data = [
    ('/', 100, [
        ('home', 70, [
            ('Images', 40, []),
            ('Videos', 20, []),
            ('Documents', 5, []),
        ]),
        ('usr', 15, [
            ('src', 6, [
                ('linux-headers', 4, []),
                ('virtualbox', 1, []),

            ]),
            ('lib', 4, []),
            ('share', 2, []),
            ('bin', 1, []),
            ('local', 1, []),
            ('include', 1, []),
        ]),
    ]),
]
sunburst(data)

对于同一个例子，如果有人给我一个以节点层次结构为级别的 excel 文件中的决策树输出，有没有办法将此 excel 输出（在下面找到）转换为上面的列表，以便我可以使用给定函数创建 Sunburst .

Excel 输出：

Level0,Level1,Level2,Level3,Volume
/,,,,15
/,home,Images,,40
/,home,Videos,,20
/,home,Documents,,5
/,home,,,5
/,usr,src,linux-headers,4
/,usr,src,virtualbox,1
/,usr,src,,1
/,usr,lib,,4
/,usr,share,,2
/,usr,bin,,1
/,usr,local,,1
/,usr,include,,1

Answer 1

您可以使用 pandas DataFrame 和递归来做到这一点：

import pandas as pd

def df_to_nested(dataframe, _groupby, level, col):
    """
    - dataframe: source data
    - _groupby: groupby columns
    - level: start from this level (0 by default)
    - col: value to aggregate
    """
    if len(dataframe) == 1:
        return [] # Reached max depth
    else:
        result = []
        df = dataframe.groupby(_groupby[level])
        level += 1 # Level0 -> Level1 (increase level)
        for key, val in df: # Iterate through groups
            result.append(tuple([key, val[col].sum(), df_to_nested(val, _groupby, level, col)]))
        level -= 1 # Level1 -> Level0 (decrease level)
        return result

df = pd.read_csv('test.csv') # Read your file

_groupby = ['Level0', 'Level1', 'Level2', 'Level3'] # Group by cols

result = df_to_nested(df, _groupby, 0, 'Volume')

print(result)

示例输出：

[
    ('/', 100, [
        ('home', 70, [
            ('Documents', 5, []),
            ('Images', 40, []),
            ('Videos', 20, [])
        ]),
        ('usr', 15, [
            ('bin', 1, []),
            ('include', 1, []),
            ('lib', 4, []),
            ('local', 1, []),
            ('share', 2, []),
            ('src', 6, [
                ('linux-headers', 4, []),
                ('virtualbox', 1, [])
            ])
        ])
    ])
]

Answer 2

Nam Nguyen 的回答非常好，但是在特定级别只有一条记录的情况下有一个小错误，然后此语句 len(dataframe) == 1 变为 True 并且特定级别的一个值不包含在结果。我也更新了他的答案以计算这种情况：

def df_to_nested(dataframe, _groupby, level, col):
"""
- dataframe: source data
- _groupby: groupby columns
- level: start from this level (0 by default)
- col: value to aggregate
"""
result = []
if len(dataframe) == 1:        
    try:
        df = dataframe.groupby(_groupby[level])                    
        for key, val in df: # Iterate through groups                
            result.append(tuple([key, val[col].sum(), []]))
    except Exception: # Reached max depth
        pass
else:
    df = dataframe.groupby(_groupby[level])
    level += 1 # Level0 -> Level1 (increase level)
    for key, val in df: # Iterate through groups
        result.append(tuple([key, val[col].sum(), df_to_nested(val, _groupby, level, col)]))
    level -= 1 # Level1 -> Level0 (decrease level)

return result

Answer 3

我想用这个问题的答案在 Javascript 中创建 Plotly Sunburst。但是，输出与these examples 中的数据格式不同。我想要一个更像answer 的输出，所以我稍微更改了代码，现在我可以将它用于前端的 Javascript。

我会把它留在这里，以防它对某人有用。

def df_to_nested(dataframe, _groupby, level=0, col='count'):

    result = []

    if level == (len(_groupby) - 1):
        df = dataframe.groupby(_groupby[level])
        parent_cols = _groupby[:level]
        for key, val in df:  # Iterate through groups
            row = val.head(1)
            parents = "-".join(str(list(row[p])[0]) for p in parent_cols)
            
            result.append({'labels':key,'values':int(val[col].sum()),'parents':parents,
                           'ids':parents+"-"+str(key)})

    else:
        df = dataframe.groupby(_groupby[level])
        parent_cols = _groupby[:level]
        level += 1  # Level0 -> Level1 (increase level)
        for key, val in df:  # Iterate through groups
            if level==1:
                parents = ""
                ids=key
            else:
                row = val.head(1)
                parents = "-".join(str(list(row[p])[0]) for p in parent_cols)
                ids = parents+"-"+str(key)

            result.append({'labels':key,'values':int(val[col].sum()),'parents':parents,
           'ids':ids})
            result.extend(df_to_nested(val, _groupby, level, col))

        level -= 1  # Level1 -> Level0 (decrease level)

    return result

def get_sunburst_format(df,path):
    # path is the list of columns in dataframe that you want a sunburst from
    tmp = df.copy()
    tmp['count'] = 1
    sunburst_data = df_to_nested(tmp,path)
    sunburst_data = pd.DataFrame(sunburst_data)
    return {column:list(sunburst_data[column]) for column in sunburst_data.columns}