What are the methods for Dataframe query in pandas

This time I will bring you the DataframeQuerymethods in pandas, and theNotesof Dataframe query in pandas. The following is a practical case, let’s take a look.

pandas provides us with a variety of slicing methods, but if you don't know much about these methods, it's often easy to get confused. The following examples illustrate these slicing methods.

Data introduction

First randomly generate a set of data:

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

[]Slicing method

Use square brackets to slice DataFrame, somewhat similar to python's list slicing. Row selection or column selection or block selection can be achieved according to the index.

# 行选择
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
# 列选择
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]
# 区块选择
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

However, for multi-column selection, you cannot use the 1:5 method like when selecting rows.

In [12]: data[['rnd_1':'rnd_3']]
 File "<ipython-input-13-6291b6a83eb0>", line 1
  data[['rnd_1':'rnd_3']]
         ^
SyntaxError: invalid syntax</ipython-input-13-6291b6a83eb0>

loc

loc allows you to select rows and columns based on index.

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

It should be noted here that the difference between loc and the first method is that it will also select the 5th row, while the first method will only select the 4th row.

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 can select data between two specific dates. It should be noted that both dates must be in the index.

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]: # 生成两个特定日期
  ...: fecha_1 = dt.datetime(2013, 4, 14)
  ...: fecha_2 = dt.datetime(2013, 4, 18)
  ...: 
  ...: # 生成切片数据
  ...: 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

Update:If there are no special needs, it is strongly recommended to use loc and use [] as little as possible, because loc is reprocessing the DataFrame. Chained indexing problems will be avoided during assignment operations. When using [], the compiler is likely to give a warning about SettingWithCopy.

For details, please refer to the official documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy

iloc

If loc is selected according to the value of the index, then iloc is selected according to the position of the index. iloc doesn't care about the specific value of the index, it only cares about the position, so when using iloc, only numerical values ​​can be used in square brackets.

# 行选择
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
# 列选择
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
# 切片选择
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

The usage of at is similar to loc, but it has faster access to data than loc, and can only access A single element, multiple elements cannot be accessed.

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 is to iloc what at is to loc, a faster option based on index position Method, like at, can only access a single element.

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

The methods mentioned above all require that the rank of the query is in the index, or the position does not exceed the length range, and ix allows you to get data that is not in the DataFrame index.

In [28]: date_1 = dt.datetime(2013, 1, 10, 8, 30)
  ...: date_2 = dt.datetime(2013, 1, 13, 4, 20)
  ...: 
  ...: # 生成切片数据
  ...: 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

As shown in the above example, January 10, 2013 was not selected because this time point is regarded as 0:00, which is earlier than 8:30.

I believe you have mastered the method after reading the case in this article. For more exciting information, please pay attention to other related articles on the php Chinese website!

Recommended reading:

How python implements the steps of Baidu speech recognition API

How python calls the API to achieve intelligence Reply function

The above is the detailed content of What are the methods for Dataframe query in pandas. For more information, please follow other related articles on the PHP Chinese website!