最近做一个系列博客,跟着stackoverflow学Pandas。

以 pandas作为关键词,在stackoverflow中进行搜索,随后安照 votes 数目进行排序:

https://stackoverflow.com/questions/tagged/pandas?sort=votes&pageSize=15

Adding new column to existing DataFrame in Python pandas - Pandas 添加列

https://stackoverflow.com/questions/12555323/adding-new-column-to-existing-dataframe-in-python-pandas

pandas官方给出了对列的操作, 可以参考:

http://pandas.pydata.org/pandas-docs/stable/dsintro.html#column-selection-addition-deletion

  1. 数据准备

随机生成8*3的DataFrame df1,筛选 a 列大于0.5的行组成df2,作为我们的初始数据。

import numpy as np
import pandas as pd

print pd.__version__
#0.19.2
np.random.seed(0)
df1 = pd.DataFrame(np.random.randn(8, 3), columns=['a', 'b', 'c'])
print df1
          a         b         c
# 0  1.764052  0.400157  0.978738
# 1  2.240893  1.867558 -0.977278
# 2  0.950088 -0.151357 -0.103219
# 3  0.410599  0.144044  1.454274
# 4  0.761038  0.121675  0.443863
# 5  0.333674  1.494079 -0.205158
# 6  0.313068 -0.854096 -2.552990
# 7  0.653619  0.864436 -0.742165

df2 = df1[df1['a']> 0.5]
df3 = df2

sLength = len(df2['a'])
d = pd.Series(np.random.randn(sLength))

直接赋值

采用 df2['d'] = d 或者 df2.loc[:, 'd'] = d 直接进行赋值。

print df2
#           a         b         c
# 0  1.764052  0.400157  0.978738
# 1  2.240893  1.867558 -0.977278
# 2  0.950088 -0.151357 -0.103219
# 4  0.761038  0.121675  0.443863
# 7  0.653619  0.864436 -0.742165

print d
# 0    2.269755
# 1   -1.454366
# 2    0.045759
# 3   -0.187184
# 4    1.532779

print type(d)
#<class 'pandas.core.series.Series'>
# 下面的方法可以,但是会有SettingWithCopyWarning警告
df2['d'] = d
# /Library/Python/2.7/site-packages/ipykernel/__main__.py:1: SettingWithCopyWarning:
# A value is trying to be set on a copy of a slice from a DataFrame.
# Try using .loc[row_indexer,col_indexer] = value instead

# See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
#   if __name__ == '__main__':
# 为了避免警告我们可以采用这种方式来进行直接赋值
df2.loc[:, 'd'] = d
print df2
          a         b         c         d
# 0  1.764052  0.400157  0.978738  2.269755
# 1  2.240893  1.867558 -0.977278 -1.454366
# 2  0.950088 -0.151357 -0.103219  0.045759
# 4  0.761038  0.121675  0.443863  1.532779
# 7  0.653619  0.864436 -0.742165       NaN

df2.loc[:, 'd1'] = d.tolist() # 或者 d.values()
# d.tolist() 返回list
# d.values 返回 numpy.ndarray

print df2
#           a         b         c         d        d1
# 0  1.764052  0.400157  0.978738  2.269755  2.269755
# 1  2.240893  1.867558 -0.977278 -1.454366 -1.454366
# 2  0.950088 -0.151357 -0.103219  0.045759  0.045759
# 4  0.761038  0.121675  0.443863  1.532779 -0.187184
# 7  0.653619  0.864436 -0.742165       NaN  1.532779

我们可以发现,df2是5行数据, d 也是5个数据,但是赋值之后d列仅有4个值,深究发现,d是Series类型,df2['d'] = d 是根据index对其进行赋值,只有 0 1 2 4 等4个index在d中有对应, 7 没有对应所以为NaN.

如果忽略index影响,我们可以采用d.tolist() 或者 d.values()

同时,在 pandas 0.19.2 中,采用 df2['d'] = d, 提示SettingWithCopyWarning,尽量避免这种方式,采用df2.loc[:, 'd'] = d的方式进行列的增加。

assign 赋值

官方推荐,assign 为DataFrame增加新列。

pandas官方参考:

http://pandas.pydata.org/pandas-docs/stable/generated/pandas.DataFrame.assign.html

print df3
#           a         b         c
# 0  1.764052  0.400157  0.978738
# 1  2.240893  1.867558 -0.977278
# 2  0.950088 -0.151357 -0.103219
# 4  0.761038  0.121675  0.443863
# 7  0.653619  0.864436 -0.742165

print d
# 0    2.269755
# 1   -1.454366
# 2    0.045759
# 3   -0.187184
# 4    1.532779

# 对 d.values (numpy.ndarray)进行赋值
df3 = df3.assign(d = d.values)
print df3

#           a         b         c         d
# 0  1.764052  0.400157  0.978738  2.269755
# 1  2.240893  1.867558 -0.977278 -1.454366
# 2  0.950088 -0.151357 -0.103219  0.045759
# 4  0.761038  0.121675  0.443863 -0.187184
# 7  0.653619  0.864436 -0.742165  1.532779

# 对 d(Series) 进行赋值
df4 = df3.assign(d = d)
print df4

          a         b         c         d
# 0  1.764052  0.400157  0.978738  2.269755
# 1  2.240893  1.867558 -0.977278 -1.454366
# 2  0.950088 -0.151357 -0.103219  0.045759
# 4  0.761038  0.121675  0.443863  1.532779
# 7  0.653619  0.864436 -0.742165       NaN

可以发现 df3 采用 assign 进行赋值,可以得到跟loc直接赋值相同的结果, 区别在于赋值的类型是 Series还是 numpy.ndarray 或者是list。

同时,assign还可以进行多种操作,比如:

df4 = df3.assign(ln_A = lambda x: np.log(x['a']))
print df4

#           a         b         c         d      ln_A
# 0  1.764052  0.400157  0.978738  2.269755  0.567614
# 1  2.240893  1.867558 -0.977278 -1.454366  0.806875
# 2  0.950088 -0.151357 -0.103219  0.045759 -0.051200
# 4  0.761038  0.121675  0.443863 -0.187184 -0.273072
# 7  0.653619  0.864436 -0.742165  1.532779 -0.425231
05-11 16:01