在数据处理中,不免会遇到数据之间的合并。学过关系数据库的童鞋应该都清楚数据表之间的连接。今天要说的数据合并其实和数据表之间的连接有很多相似之处。由于 pandas 库在数据分析中比较方便而且用者较多,我们就说pandas中的数据合并方式。
pandas 中数据合并常用的方法有三种:pandas.merge(), pandas.concat(), 以及实例方法 combine_first()
merge函数的参数:
left, right | 参与合并的左、右侧Dataframe |
how | inner、outer、left、right之一,默认是 inner |
on | 指定用于连接的列名。如果未指定,则会自动选取要合并数据中相同的列名 |
left_on, right_on | 左、右侧Dataframe用于连接键的列 |
left_index, right_index | 将左 \ 右侧的行索引用作连接键 |
sort | 根据连接键对合并后的数据进行排序,默认True。在处理大数据集的时候,禁用此功能能会有很好的性能提升 |
suffixes | 字符串元组,用于追加到重复列名的末尾,默认('_x', '_y') |
copy | 设置为False,可以在某些特殊情况下 避免将数据复制到结果数据结构中。默认总是复制。 |
看几个简单的例子:
In [10]: df1 = pd.DataFrame({'k':['s','s','w','x','x','n','f','c'],'data1':range(8)})
In [11]: df2 = pd.DataFrame({'k':['w','w','s','s','x','f'],'data2':range(6)})
In [12]: pd.merge(df1,df2) # 未指定合并的列,默认选取两者重复的列 k, 也可以指定 pd.merge(df1,df2,on='k')
Out[12]:
data1 k data2
0 0 s 2
1 0 s 3
2 1 s 2
3 1 s 3
4 2 w 0
5 2 w 1
6 3 x 4
7 4 x 4
8 6 f 5
从上面的结果中会发现有些行消失了,这是因为默认使用的是 inner 连接方式,结果做的交集。可以指定其他的连接方式(参见上面的 how 参数值)。例如:
In [13]: pd.merge(df1,df2,how='outer') Out[13]: data1 k data2 0 0 s 2.0 1 0 s 3.0 2 1 s 2.0 3 1 s 3.0 4 2 w 0.0 5 2 w 1.0 6 3 x 4.0 7 4 x 4.0 8 5 n NaN 9 6 f 5.0 10 7 c NaN合并的列名不同时,手动指定要合并的列:
In [14]: df2 = pd.DataFrame({'k_1':['w','w','s','s','x','f'] ,'data2':range(6)}) In [17]: pd.merge(df1,df2,left_on='k',right_on='k_1',how='outer') Out[17]: data1 k data2 k_1 0 0 s 2.0 s 1 0 s 3.0 s 2 1 s 2.0 s 3 1 s 3.0 s 4 2 w 0.0 w 5 2 w 1.0 w 6 3 x 4.0 x 7 4 x 4.0 x 8 5 n NaN NaN 9 6 f 5.0 f 10 7 c NaN NaN
根据多个键进行合并,只需要在 on 关键字传入一个列名组成的列表即可:
In [25]: left = pd.DataFrame({'key1': ['foo', 'foo', 'bar'],
...: 'key2': ['one', 'two', 'one'],'val': [1, 2, 3]})
In [26]: right = pd.DataFrame({'key1': ['foo', 'foo','bar','bar'],
...: 'key2': ['one', 'one', 'one','two'],'val': [4,5,6,7]})
In [27]: pd.merge(left, right, on=['key1', 'key2'], how='outer')
Out[27]:
key1 key2 val_x val_y
0 foo one 1.0 4.0
1 foo one 1.0 5.0
2 foo two 2.0 NaN
3 bar one 3.0 6.0
4 bar two NaN 7.0
从上面的例子中还可以看到,当合并的数据有相同的列时,结果会默认在后面添加_x, _y来区分。也可以手动指定:
In [31]: pd.merge(left, right, on='key1', how='outer',suffixes=('_left','_right')) Out[31]: key1 key2_left val_left key2_right val_right 0 foo one 1 one 4 1 foo one 1 one 5 2 foo two 2 one 4 3 foo two 2 one 5 4 bar one 3 one 6 5 bar one 3 two 7
有的时候我们要合并的连接键在索引中,这种情况我们就要通过 left_index = True \ right_index = True 来明确索引作为连接键。 In [35]: df3 = pd.DataFrame({'data3':[5,2,0]},index=list('sxn')) In [36]: df3 Out[36]: data3 s 5 x 2 n 0 In [37]: pd.merge(df1,df3,left_on='k',right_index=True,how='outer') Out[37]: data1 k data3 0 0 s 5.0 1 1 s 5.0 2 2 w NaN 3 3 x 2.0 4 4 x 2.0 5 5 n 0.0 6 6 f NaN 7 7 c NaN
合并层次化索引的数据,必须以列表的形式指明用作合并的列:
In [49]: lefth = pd.DataFrame({'key1': ['sxn', 'sxn', 'sxn', 'wfc', 'wfc'],
'key2': [2000, 2001, 2002, 2001, 2002],
...: 'data': np.arange(5.)})
In [50]: righth = pd.DataFrame(np.arange(12).reshape((6, 2)) ,
...: index=[['wfc', 'wfc', 'sxn', 'sxn', 'sxn', 'snx'],
...: [2001, 2000, 2000, 2000, 2001, 2002]],
...: columns=['event1', 'event2'])
In [51]: pd.merge(lefth, righth, left_on=['key1', 'key2'],right_index=True, how='outer')
Out[51]:
data key1 key2 event1 event2
0 0.0 sxn 2000 4.0 5.0
0 0.0 sxn 2000 6.0 7.0
1 1.0 sxn 2001 8.0 9.0
2 2.0 sxn 2002 NaN NaN
3 3.0 wfc 2001 0.0 1.0
4 4.0 wfc 2002 NaN NaN
4 NaN wfc 2000 2.0 3.0
4 NaN snx 2002 10.0 11.0
In [62]: df1 = pd.DataFrame({'data1':range(8)},index=['s','s','w','x','x','n','f','c']) In [63]: df2 = pd.DataFrame({'data2':range(4)},index=['s','n','f','c']) In [64]: df1.join(df2) Out[64]: data1 data2 c 7 3.0 f 6 2.0 n 5 1.0 s 0 0.0 s 1 0.0 w 2 NaN x 3 NaN x 4 NaN In [65]: df3 = pd.DataFrame({'data3':[5,2,0]},index=list('sx ...: n')) ...: In [66]: df1.join([df2,df3]) Out[66]: data1 data2 data3 c 7 3.0 NaN f 6 2.0 NaN n 5 1.0 0.0 s 0 0.0 5.0 s 1 0.0 5.0 w 2 NaN NaN x 3 NaN 2.0 x 4 NaN 2.0
这种连接也是轴向连接,也叫连接,绑定,堆叠。
numpy 中 concatenation函数。
In [68]: arr = np.arange(12).reshape(3,4) In [69]: np.concatenate([arr,arr],axis=1) # 指定连接轴,axis=1 Out[69]: array([[ 0, 1, 2, 3, 0, 1, 2, 3], [ 4, 5, 6, 7, 4, 5, 6, 7], [ 8, 9, 10, 11, 8, 9, 10, 11]]) In [70]: np.concatenate([arr,arr],axis=0) # 指定连接轴,axis=0 Out[70]: array([[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11], [ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]])
concat的参数:
objs | 参与连接的pandas的列表或字典,唯一必须的参数 |
axis | 指明连接的轴向,默认axis=0 |
join | inner、outer之一 |
join_axes | 指明用于其他n-1条轴上的索引,不执行并集交集运算 |
keys | 用于形成连接轴上的层次化索引 |
levels | 指定用于层次化索引各级别上的索引,如果设置了的keys的话 |
names | 用于创建分级别的名称,如果设置了levels和keys 的话 |
verify_integrity | 检查结果对象新轴上的重复情况,如果发现引发异常,默认可以重复 |
ignore_index | 不保留连接轴上的索引,产生一组新索引 |
下面是一些例子说明以上参数的使用情况:
In [77]: s1 = pd.Series([0,1],index=['a','b']) In [78]: s2 = pd.Series([2,3,4],index=['c','d','e']) In [79]: s3 = pd.Series([5,6],index=['f','g']) In [80]: pd.concat([s1,s2,s3]) # 默认 axis=0 Out[80]: a 0 b 1 c 2 d 3 e 4 f 5 g 6 dtype: int64 In [81]: pd.concat([s1,s2,s3],axis=1) # 按照 axis=1 进行连接,产生一个Dataframe 对象 Out[81]: 0 1 2 a 0.0 NaN NaN b 1.0 NaN NaN c NaN 2.0 NaN d NaN 3.0 NaN e NaN 4.0 NaN f NaN NaN 5.0 g NaN NaN 6.0
In [90]: s4 = pd.concat([s1*5,s3]) In [91]: pd.concat([s1,s4]) Out[91]: a 0 b 1 a 0 b 5 f 5 g 6 dtype: int64 In [94]: pd.concat([s1,s4],keys=['s1','s4']) # 可以区分合并后的结果 Out[94]: s1 a 0 b 1 s4 a 0 b 5 f 5 g 6 dtype: int64 In [97]: pd.concat([s1,s4],axis=1,keys=['s1','s4']) # 沿着 axis=1 合并,指定的 keys 就会变成Dataframe的列名 Out[97]: s1 s4 a 0.0 0 b 1.0 5 f NaN 5 g NaN 6
In [92]: pd.concat([s1,s4],axis=1) # 按照 axis=1 进行连接,产生一个Dataframe 对象
Out[92]:
0 1
a 0.0 0
b 1.0 5
f NaN 5
g NaN 6
In [93]: pd.concat([s1,s4],axis=1,join='inner') # join='inner' 产生交集
Out[93]:
0 1
a 0 0
b 1 5
In [96]: pd.concat([s1,s4],axis=1,join_axes=[['a','c','b','f']]) # 指定要在其他轴上使用的索引 Out[96]: 0 1 a 0.0 0.0 c NaN NaN b 1.0 5.0 f NaN 5.0
对于Dataframe的合并,逻辑差不多类似:
In [98]: df1 = pd.DataFrame(np.arange(6).reshape(3, 2), index=['a', 'b', 'c'],columns=['one', 'two']) In [99]: df2 = pd.DataFrame(5 + np.arange(4).reshape(2, 2), index=['a', 'c'],columns=['three', 'four']) In [100]: pd.concat([df1, df2], axis=1, keys=['level1', 'level2']) Out[100]: level1 level2 one two three four a 0 1 5.0 6.0 b 2 3 NaN NaN c 4 5 7.0 8.0
In [102]: pd.concat([df1, df2], axis=1) Out[102]: one two three four a 0 1 5.0 6.0 b 2 3 NaN NaN c 4 5 7.0 8.0
In [101]: pd.concat([df1, df2]) # 不指定轴向。默认axis=0
Out[101]:
four one three two
a NaN 0.0 NaN 1.0
b NaN 2.0 NaN 3.0
c NaN 4.0 NaN 5.0
a 6.0 NaN 5.0 NaN
c 8.0 NaN 7.0 NaN
In [103]: pd.concat([df1, df2],keys=['level1', 'level2'])
Out[103]:
four one three two
level1 a NaN 0.0 NaN 1.0
b NaN 2.0 NaN 3.0
c NaN 4.0 NaN 5.0
level2 a 6.0 NaN 5.0 NaN
c 8.0 NaN 7.0 NaN
In [104]: pd.concat([df1, df2], axis=1, keys=['level1', 'level2'],names=['up','down']) # 用 names 为各分层级别命名
Out[104]:
up level1 level2
down one two three four
a 0 1 5.0 6.0
b 2 3 NaN NaN
c 4 5 7.0 8.0
In [105]: s1 = pd.Series([np.nan, 2.5, np.nan, 3.5, 4.5, np.nan],index=['f', 'e', 'd', 'c', 'b', 'a']) In [107]: s2 = pd.Series(np.arange(len(s1), dtype=np.float64),index=['f', 'e', 'd', 'c', 'b', 'a']) In [108]: s2[-1] = np.nan In [110]: s1 Out[110]: f NaN e 2.5 d NaN c 3.5 b 4.5 a NaN dtype: float64 In [111]: s2 Out[111]: f 0.0 e 1.0 d 2.0 c 3.0 b 4.0 a NaN dtype: float64 In [113]: np.where(pd.isnull(s1),s2,s1) Out[113]: array([0. , 2.5, 2. , 3.5, 4.5, nan]) In [114]: s3 = pd.Series(np.where(pd.isnull(s1),s2,s1),index=s1.index) In [115]: s3 Out[115]: f 0.0 e 2.5 d 2.0 c 3.5 b 4.5 a NaN dtype: float64嗯,看到没,大概就是上面这个样子的......
combine_first() 也是实现一样的功能:
In [118]: s2[:-2].combine_first(s1[2:])
Out[118]:
a NaN
b 4.5
c 3.0
d 2.0
e 1.0
f 0.0
dtype: float64
看看Dataframe的 combine_first():
In [119]: df1 = pd.DataFrame({'a': [1., np.nan, 5., np.nan],'b': [np.nan, 2., np.nan, 6.], ...: 'c': range(2, 18, 4)}) In [120]: df2 = pd.DataFrame({'a': [5., 4., np.nan, 3., 7.],'b': [np.nan, 3., 4., 6., 8.]}) In [121]: df1.combine_first(df2) Out[121]: a b c 0 1.0 NaN 2.0 1 4.0 2.0 6.0 2 5.0 4.0 10.0 3 3.0 6.0 14.0 4 7.0 8.0 NaN