父子关系分层答案

【问题标题】：Parent Child Relationship Hierarchical父子关系分层
【发布时间】：2021-07-16 03:47:14
【问题描述】：

Getting all descendants of a parent from a pandas dataframe parent child table

我想做与上面类似的事情，但是我不想按父母分组，而是希望对输出有更多的层次结构。所以 child_id 总是变成 parent_id 除非没有孩子；在这种情况下检查最后一个 parent_id

Current Output:
    parent_id  child_id
0        1000      2010
1        1000      2100
2        1000      2110
3        1000      3000
4        1000      3011
5        1000      3033
6        1000      3102
7        1000      3111

Preferred Output:
    parent_id  child_id
0        1000      2010
1        2010      3011
2        3011      3050
3        2010      3102
4        2010      4001
5        1000      3000
6        3000      3011
7        3011      3050
8        3000      3033
9        1000      3102
10       1000      3111

etc. etc.

【问题讨论】：

标签： python-3.x pandas

【解决方案1】：

我能够想出一些东西。我不知道它是否是最好/最快/最有效的，但它确实有效。

首先，使用上面的脚本创建父子关系（如果尚未存在），并添加一个名为 level 的列，用于描述部件在树 0 中的最高级别。那么……

#this part of the script will create a row for each branch of the tree
dfsort = df[df['level'] == 0][['parent_id','child_id']].rename(columns = {'parent_id':f'level 0', 'child_id':f'level 1'})


for i in sorted(df['level'].unique()[1:]):
    df1 = df[df['level'] == i][['parent_id','child_id']].rename(columns = {'parent_id':f'level {i}', 'child_id':f'level {i+1}'})

    dfsort = pd.merge(dfsort, 
                      df,                  
                      how = 'left', on = [f'level {i}']

dfsort = dfsort[sorted(dfsort.columns)]
#create a node column to drop duplicates on (in case any similar parent child relations are used across multiple higher level parts
dfsort['Node'] = dfsort.astype(str).apply(list, axis =1 )
dfsort['Node'] = dfsort['Node'].apply(lambda x: [i for i in x if i != 'nan'])

#now that you have this relationship you can break it out in the correct order using another for loop

dfsort2 = pd.DataFrame()

#append a new dataframe with the parent childs from above table one row at a time
for i in range(len(dfsort)):
    for l in range(len(dfsort.iloc[i][:-1])):
        df = dfsort.iloc[[i]][[f'level {i}',f'level {i-1}', 'Node']].rename(columns = {f'level {l}':'parent_id', f'level {l+1}'})
        df['level'] = l
        dfsort2 = pd.concat([dfsort2, df])

dfsort2 = dfsort[(dfsort2['parent_id'].notna()) & 
                 (dfsort2['child_id'].notna())]

dfsort2 ['order node index'] = dfsort2 .apply(lambda x: x['Node'].index(x['child_id']), axis = 1)
dfsort2 ['Query Node'] = dfsort2 .apply(lambda x: x['Node'][:x['order node index'] + 1], axis = 1).apply(lambda x: ",".join(x))
del dfsort2 ['order node index'], dfsort2 ['Node']
dfsort2 = dfsort2 .drop_duplicates()

【讨论】：