pandas為我們提供了多種切片方法,而要是不太了解這些方法,就會經常容易混淆。下面舉例對這些切片方法進行說明。
數據介紹
先隨機生成一組數據:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
|
In [5]: rnd_1 = [random.randrange(1,20) for x in xrange(1000)] ...: rnd_2 = [random.randrange(1,20) for x in xrange(1000)] ...: rnd_3 = [random.randrange(1,20) for x in xrange(1000)] ...: fecha = pd.date_range('2012-4-10', '2015-1-4') ...: ...: data = pd.DataFrame({'fecha':fecha, 'rnd_1': rnd_1, 'rnd_2': rnd_2, 'rnd_3': rnd_3}) In [6]: data.describe() Out[6]: rnd_1 rnd_2 rnd_3 count 1000.000000 1000.000000 1000.000000 mean 9.946000 9.825000 9.894000 std 5.553911 5.559432 5.423484 min 1.000000 1.000000 1.000000 25% 5.000000 5.000000 5.000000 50% 10.000000 10.000000 10.000000 75% 15.000000 15.000000 14.000000 max 19.000000 19.000000 19.000000 |
[]切片方法
使用方括號能夠對DataFrame進行切片,有點類似于python的列表切片。按照索引能夠實現行選擇或列選擇或區塊選擇。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
|
# 行選擇 In [7]: data[1:5] Out[7]: fecha rnd_1 rnd_2 rnd_3 1 2012-04-11 1 16 3 2 2012-04-12 7 6 1 3 2012-04-13 2 16 7 4 2012-04-14 4 17 7 # 列選擇 In [10]: data[['rnd_1', 'rnd_3']] Out[10]: rnd_1 rnd_3 0 8 12 1 1 3 2 7 1 3 2 7 4 4 7 5 12 8 6 2 12 7 9 8 8 13 17 9 4 7 10 14 14 11 19 16 12 2 12 13 15 18 14 13 18 15 13 11 16 17 7 17 14 10 18 9 6 19 11 15 20 16 13 21 18 9 22 1 18 23 4 3 24 6 11 25 2 13 26 7 17 27 11 8 28 3 12 29 4 2 .. ... ... 970 8 14 971 19 5 972 13 2 973 8 10 974 8 17 975 6 16 976 3 2 977 12 6 978 12 10 979 15 13 980 8 4 981 17 3 982 1 17 983 11 5 984 7 7 985 13 14 986 6 19 987 13 9 988 3 15 989 19 6 990 7 11 991 11 7 992 19 12 993 2 15 994 10 4 995 14 13 996 12 11 997 11 15 998 17 14 999 3 8 [1000 rows x 2 columns] # 區塊選擇 In [11]: data[:7][['rnd_1', 'rnd_2']] Out[11]: rnd_1 rnd_2 0 8 17 1 1 16 2 7 6 3 2 16 4 4 17 5 12 19 6 2 7 |
不過對于多列選擇,不能像行選擇時一樣使用1:5這樣的方法來選擇。
1
2
3
4
5
|
In [12]: data[['rnd_1':'rnd_3']] File "< ipython-input-13-6291b6a83eb0 >", line 1 data[['rnd_1':'rnd_3']] ^ SyntaxError: invalid syntax |
loc
loc可以讓你按照索引來進行行列選擇。
1
2
3
4
5
6
7
8
|
In [13]: data.loc[1:5] Out[13]: fecha rnd_1 rnd_2 rnd_3 1 2012-04-11 1 16 3 2 2012-04-12 7 6 1 3 2012-04-13 2 16 7 4 2012-04-14 4 17 7 5 2012-04-15 12 19 8 |
這里需要注意的是,loc與第一種方法不同之處在于會把第5行也選擇進去,而第一種方法只會選擇到第4行為止。
1
2
3
4
5
6
|
data.loc[2:4, ['rnd_2', 'fecha']] Out[14]: rnd_2 fecha 2 6 2012-04-12 3 16 2012-04-13 4 17 2012-04-14 |
loc能夠選擇在兩個特定日期之間的數據,需要注意的是這兩個日期必須都要在索引中。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
|
In [15]: data_fecha = data.set_index('fecha') ...: data_fecha.head() Out[15]: rnd_1 rnd_2 rnd_3 fecha 2012-04-10 8 17 12 2012-04-11 1 16 3 2012-04-12 7 6 1 2012-04-13 2 16 7 2012-04-14 4 17 7 In [16]: # 生成兩個特定日期 ...: fecha_1 = dt.datetime(2013, 4, 14) ...: fecha_2 = dt.datetime(2013, 4, 18) ...: ...: # 生成切片數據 ...: data_fecha.loc[fecha_1: fecha_2] Out[16]: rnd_1 rnd_2 rnd_3 fecha 2013-04-14 17 10 5 2013-04-15 14 4 9 2013-04-16 1 2 18 2013-04-17 9 15 1 2013-04-18 16 7 17 |
更新:如果沒有特殊需求,強烈建議使用loc而盡量少使用[],因為loc在對DataFrame進行重新賦值操作時會避免chained indexing問題,使用[]時編譯器很可能會給出SettingWithCopy的警告。
具體可以參見官方文檔:http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy
iloc
如果說loc是按照索引(index)的值來選取的話,那么iloc就是按照索引的位置來進行選取。iloc不關心索引的具體值是多少,只關心位置是多少,所以使用iloc時方括號中只能使用數值。
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
|
# 行選擇 In [17]: data_fecha[10: 15] Out[17]: rnd_1 rnd_2 rnd_3 fecha 2012-04-20 14 6 14 2012-04-21 19 14 16 2012-04-22 2 6 12 2012-04-23 15 8 18 2012-04-24 13 8 18 # 列選擇 In [18]: data_fecha.iloc[:,[1,2]].head() Out[18]: rnd_2 rnd_3 fecha 2012-04-10 17 12 2012-04-11 16 3 2012-04-12 6 1 2012-04-13 16 7 2012-04-14 17 7 # 切片選擇 In [19]: data_fecha.iloc[[1,12,34],[0,2]] Out[19]: rnd_1 rnd_3 fecha 2012-04-11 1 3 2012-04-22 2 12 2012-05-14 17 10 |
at
at的使用方法與loc類似,但是比loc有更快的訪問數據的速度,而且只能訪問單個元素,不能訪問多個元素。
1
2
3
4
5
6
7
8
|
In [20]: timeit data_fecha.at[fecha_1,'rnd_1'] The slowest run took 3783.11 times longer than the fastest. This could mean that an intermediate result is being cached. 100000 loops, best of 3: 11.3 µs per loop In [21]: timeit data_fecha.loc[fecha_1,'rnd_1'] The slowest run took 121.24 times longer than the fastest. This could mean that an intermediate result is being cached. 10000 loops, best of 3: 192 µs per loop In [22]: data_fecha.at[fecha_1,'rnd_1'] Out[22]: 17 |
iat
iat對于iloc的關系就像at對于loc的關系,是一種更快的基于索引位置的選擇方法,同at一樣只能訪問單個元素。
1
2
3
4
5
6
7
|
In [23]: data_fecha.iat[1,0] Out[23]: 1 In [24]: timeit data_fecha.iat[1,0] The slowest run took 6.23 times longer than the fastest. This could mean that an intermediate result is being cached. 100000 loops, best of 3: 8.77 µs per loop In [25]: timeit data_fecha.iloc[1,0] 10000 loops, best of 3: 158 µs per loop |
ix
以上說過的幾種方法都要求查詢的秩在索引中,或者位置不超過長度范圍,而ix允許你得到不在DataFrame索引中的數據。
1
2
3
4
5
6
7
8
9
10
11
|
In [28]: date_1 = dt.datetime(2013, 1, 10, 8, 30) ...: date_2 = dt.datetime(2013, 1, 13, 4, 20) ...: ...: # 生成切片數據 ...: data_fecha.ix[date_1: date_2] Out[28]: rnd_1 rnd_2 rnd_3 fecha 2013-01-11 19 17 19 2013-01-12 10 9 17 2013-01-13 15 3 10 |
如上面的例子所示,2013年1月10號并沒有被選擇進去,因為這個時間點被看作為0點0分,比8點30分要早一些。
以上這篇淺談pandas中Dataframe的查詢方法([], loc, iloc, at, iat, ix)就是小編分享給大家的全部內容了,希望能給大家一個參考,也希望大家多多支持服務器之家。
原文鏈接:https://blog.csdn.net/wr339988/article/details/65446138