条件数据帧选择的矢量化解决方案 [英] Vectorized solution to conditional dataframe selection

问题描述

我最近问了一个已回答的问题-当列为一系列列表时，如何有条件地将其添加到pandas数据框列中的单元格选择中？，但是我相信我遇到了一个新问题

在以下数据框中，我需要两个条件才能导致对列 d 进行更改。 d 列中的每个值都是一个列表。

• 其中 a == b 时，d中的最后一个整数加1。



• 在 a！= b 的情况下，扩展整数列表，并将值 1 插入到 list 在列 d 中。

    abcd 
  开开[0] [0,3] 
  开关[0] [0,1] 
  开开[0] [2] 
  开开[0] [0,4,4] 
  开关[0] [0] 
    




• 结果，数据框将如下所示：

    abcd 
  开开[0] [ 0,4] 
  开开[0] [0,1,1] 
  开开[0] [3] 
  开开[0] [0,4,5] 
  开关[0] [0,1] 
    

我意识到可以使用 pd.Series.apply 方法和预定义函数或使用 lambda来完成此操作但是，数据帧包含100000行，我希望可以存在针对这两个条件的矢量化解决方案。

解决方案

作为Edchum ，

一个非矢量化解决方案，其 应用 自定义功能：

  df ['e'] = df ['d'] 
 
 def扩展（lst）：
返回lst + [1] 
 
 def增量（lst）：
 lst [-1] = lst [ -1] + 1 
 return lst 
 
 df.loc [df.a！= df.b，'d'] = df.e.apply（exten）
 df .loc [df.a == df.b，'d'] = df.e.apply（incre）
 df = df.drop（'e'，axis = 1）
 print df 
abcd 
 0开开[0] [0，4] 
 1开关[0] [0，1，1] 
 2开开[0] [3] 
 3开开[0] [0，4，5] 
 4开关[0] [0，1] 
  

I recently asked a question which was answered - How do I add conditionally to a selection of cells in a pandas dataframe column when the the column is a series of lists?, but I believe have a new problem which I had not previously considered.

In the following dataframe I need two conditions to result in a change to column d. Each value in column d is a list.

• Where a == b, the final integer in d is incremented by one.

• Where a != b, the list of integers is extended and the value 1 is inserted at the end of the list in column d.

  a       b       c           d           
  On      On      [0]         [0,3]       
  On      Off     [0]         [0,1]
  On      On      [0]         [2]         
  On      On      [0]         [0,4,4]         
  On      Off     [0]         [0]
  

• As a result, the dataframe would like this:

  a       b       c       d       
  On      On      [0]     [0,4]       
  On      Off     [0]     [0,1,1]     
  On      On      [0]     [3]
  On      On      [0]     [0,4,5] 
  On      Off     [0]     [0,1]
  

I realise that this can be done using pd.Series.apply method in conjunction with a predefined function or use of lambda however the data frame consists of 100000 rows and I was hoping that a vectorized solution to these two conditions may exist.

解决方案

As Edchum says, vecorised solution can be problematic.

One non vectorized solution with apply custom functions:

df['e'] = df['d']

def exten(lst):
    return lst + [1]

def incre(lst):
    lst[-1] = lst[-1] + 1
    return lst

df.loc[df.a != df.b, 'd'] = df.e.apply(exten)
df.loc[df.a == df.b, 'd'] = df.e.apply(incre)
df = df.drop('e', axis=1)
print df
    a    b    c          d
0  On   On  [0]     [0, 4]
1  On  Off  [0]  [0, 1, 1]
2  On   On  [0]        [3]
3  On   On  [0]  [0, 4, 5]
4  On  Off  [0]     [0, 1]

这篇关于条件数据帧选择的矢量化解决方案的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！