groupby,apply和set不按预期行事......这是一个错误吗？

Question

要查看此问题,请考虑以下数据框

In [66]: dat = pandas.DataFrame(['a','b','c','d','e','f','g','h'],
                        columns=['letters'])

In [67]: dat['numbers'] = pandas.Series([1,2,3,4,5,6,7,8])

In [68]: dat['names'] = pandas.Series(['jim','jan','jerry','george'
                        ,'mary','mary','sue','sue'])

In [69]: dat
Out[69]:
  letters  numbers   names
0       a        1     jim
1       b        2     jan
2       c        3   jerry
3       d        4  george
4       e        5    mary
5       f        6    mary
6       g        7     sue
7       h        8     sue

按名称分组

In [78]: dat = dat.groupby(['names'])[['letters']]

现在我尝试连续写字会产生一个有趣的结果:

In [80]: dat.apply(lambda x: '|'.join(set(x)))
Out[80]:
names
george    letters|numbers|names
jan       letters|numbers|names
jerry     letters|numbers|names
jim       letters|numbers|names
mary      letters|numbers|names
sue       letters|numbers|names
dtype: object

以下黑客似乎有效,但为什么我需要再次选择"字母",为什么上面的输出看起来像呢？

In [84]: dat.apply(lambda x: '|'.join(set(x['letters'])))
Out[84]:
names
george      d
jan         b
jerry       c
jim         a
mary      e|f
sue       h|g
dtype: object

这可能是个错误吗？

安装版本

commit:无python:2.7.5.final.0 python-bits:64 OS:Darwin OS-release:13.1.0 machine:x86_64 processor:i386 byteorder:little LC_ALL:None LANG:en_US.UTF-8

pandas:0.13.1 Cython:0.20.1 numpy:1.6.2 scipy:0.11.0 statsmodels:0.5.0 IPython:2.0.0 sphinx:1.2.2 patsy:0.2.1 scikits.timeseries:None dateutil:1.5 pytz: 2012d瓶颈:无表:无numexpr:无matplotlib:1.1.1 openpyxl:无xlrd:无xlwt:无xlsxwriter:无sqlalchemy:无lxml:3.3.5 bs4:4.3.2 html5lib:无bq:无apiclient:无

Answer 1

这看起来有点奇怪,但正如你看到的一样,DataFrame是它的列:

In [11]: dat
Out[11]:
  letters  numbers   names
0       a        1     jim
1       b        2     jan
2       c        3   jerry
3       d        4  george
4       e        5    mary
5       f        6    mary
6       g        7     sue
7       h        8     sue

[8 rows x 3 columns]

In [12]: set(dat)
Out[12]: {'letters', 'names', 'numbers'}

这是由于您遍历DataFrame(通过列)的方式:

In [13]: for i in dat: print(i)
letters
numbers
names

这适用于SeriesGroupBy(迭代通过一系列迭代其元素):

In [21]: g = dat.groupby(['names'])['letters']

In [22]: g.apply(lambda x: '|'.join(set(x)))
Out[22]:
names
george      d
jan         b
jerry       c
jim         a
mary      e|f
sue       h|g
dtype: object

注意:您不需要设置或实际上是lambda:

In [23]: g.apply('|'.join)
Out[23]:
names
george      d
jan         b
jerry       c
jim         a
mary      e|f
sue       g|h
dtype: object