趣学Python数据分析：轴和索引

In [1]: import datetime as dt
In [3]: import pandas as pd
In [4]: df = pd.DataFrame([[1,'gz',dt.date(2019,4,13)],[3,'lg',dt.date(2019,4,14   ...: )],[5,'jh',dt.date(2019,4,15)]])
In [5]: dfOut[5]:    0   1           20  1  gz  2019-04-131  3  lg  2019-04-142  5  jh  2019-04-15

In [6]: df.index  # 展示行索引Out[6]: RangeIndex(start=0, stop=3, step=1)
In [7]: df.columns  # 展示列索引Out[7]: RangeIndex(start=0, stop=3, step=1)
In [8]: df.columns=['id','nickname','register']  # 为列索引重新赋值
In [9]: dfOut[9]:    id nickname    register0   1       gz  2019-04-131   3       lg  2019-04-142   5       jh  2019-04-15
In [10]: df.index=['p1','p2','p3']   # 为行索引重新赋值
In [11]: dfOut[11]:     id nickname    registerp1   1       gz  2019-04-13p2   3       lg  2019-04-14p3   5       jh  2019-04-15

In [12]: df.dtypes  # 展示每列的数据类型Out[12]: id           int64nickname    objectregister    objectdtype: object
In [13]: df.ftypes   # 展示每列为稠密(dense)或者稀疏(sparse)Out[13]: id           int64:densenickname    object:denseregister    object:densedtype: object
In [14]: df.get_dtype_counts()  # 统计列类型Out[14]: int64     1object    2dtype: int64
In [15]: df.select_dtypes(include=['int'])  # 根据列类型(如为int,或者object)筛选数据Out[15]:     idp1   1p2   3p3   5

In [16]: df.values  # 返回NumPy数据对象Out[16]: array([[1, 'gz', datetime.date(2019, 4, 13)],       [3, 'lg', datetime.date(2019, 4, 14)],       [5, 'jh', datetime.date(2019, 4, 15)]], dtype=object)
In [17]: df.get_values()  # 稀疏值转稠密后，返回NumPy数据对象Out[17]: array([[1, 'gz', datetime.date(2019, 4, 13)],       [3, 'lg', datetime.date(2019, 4, 14)],       [5, 'jh', datetime.date(2019, 4, 15)]], dtype=object)
In [18]: df.axesOut[18]: [Index(['p1', 'p2', 'p3'], dtype='object'), Index(['id', 'nickname', 'register'], dtype='object')]
In [19]: df.sizeOut[19]: 9
In [20]: df.ndimOut[20]: 2
In [21]: df.shapeOut[21]: (3, 3)

In [22]: df.loc[:,'gender']=['Male',None,None]   # 稠密列
In [23]: dfOut[23]:     id nickname    register genderp1   1       gz  2019-04-13   Malep2   3       lg  2019-04-14   Nonep3   5       jh  2019-04-15   None
In [24]: dfspa = df.copy()                                                      
In [25]: dfspaOut[25]:     id nickname    register genderp1   1       gz  2019-04-13   Malep2   3       lg  2019-04-14   Nonep3   5       jh  2019-04-15   None
In [27]: dfspa.loc[:,'gender']=pd.SparseArray(['Male',None,None])   # 稀疏列
In [28]: dfspaOut[28]:     id nickname    register genderp1   1       gz  2019-04-13   Malep2   3       lg  2019-04-14    NaNp3   5       jh  2019-04-15    NaN
In [30]: dfspa.dtypes        # 查看稀疏列的类型Out[30]: id                        int64nickname                 objectregister                 objectgender      Sparse[object, nan]dtype: object
In [31]: df.dtypesOut[31]: id           int64nickname    objectregister    objectgender      objectdtype: object
In [32]: df.memory_usage()                                                      Out[32]: Index       24id          24nickname    24register    24gender      24dtype: int64
In [33]: dfspa.memory_usage()  # 稀疏列更加节省内存空间Out[33]: Index       104id           24nickname     24register     24gender       12dtype: int64

In [35]: df.emptyOut[35]: False
In [41]: dfemp = pd.DataFrame([])In [42]: dfempOut[42]: Empty DataFrameColumns: []Index: []
In [43]: dfemp.emptyOut[43]: True
In [44]: dfemp = pd.DataFrame([[]],index=['a','b'])In [45]: dfempOut[45]: Empty DataFrameColumns: []Index: [a, b]

In [47]: df                                                                     Out[47]:     id nickname    register genderp1   1       gz  2019-04-13   Malep2   3       lg  2019-04-14   Nonep3   5       jh  2019-04-15   None
In [48]: df.get('id')Out[48]: p1    1p2    3p3    5Name: id, dtype: int64
In [53]: df.get('id').get('p2')Out[53]: 3In [54]: type(df.get('id').get('p2'))Out[54]: numpy.int64
In [56]: df.get('id').get('p22') # 如果key不存在，返回默认值NoneIn [62]: df.get('id').get('p22')==None    # 验证Out[62]: True

In [52]: df.get('id').get('p2').item()        # item返回Python原生对象Out[52]: 3In [55]: type(df.get('id').get('p2').item())  # 验证Out[55]: int

In [63]: df                                                                     Out[63]:     id nickname    register genderp1   1       gz  2019-04-13   Malep2   3       lg  2019-04-14   Nonep3   5       jh  2019-04-15   None
In [64]: df.at['p3','register']   # at 通过行、列标签访问Out[64]: datetime.date(2019, 4, 15)
In [65]: df['nickname'].at['p2']Out[65]: 'lg'
# 如果访问的key不存在，抛出KeyError错误
In [68]: df.iat[1,3]          # 通过整数索引访问，像matlab那样访问数据                                                 
In [69]: df.iat[0,1]Out[69]: 'gz'
In [70]: df.iat[2,2]Out[70]: datetime.date(2019, 4, 15)

In [71]: df                                                                     Out[71]:     id nickname    register genderp1   1       gz  2019-04-13   Malep2   3       lg  2019-04-14   Nonep3   5       jh  2019-04-15   None
In [72]: df.loc['p3','nickname']   # 此种访问建议走at访问模式Out[72]: 'jh'
In [73]: df.loc['p2']  # 行标签访问整行Out[73]: id                   3nickname            lgregister    2019-04-14gender            NoneName: p2, dtype: object
In [74]: df.loc['nickname']  # 报错！
In [75]: df.loc[:,'nickname']Out[75]: p1    gzp2    lgp3    jhName: nickname, dtype: object
In [76]: df.iloc[2]   # 访问行整数索引为2的行Out[76]: id                   5nickname            jhregister    2019-04-15gender            NoneName: p3, dtype: object
In [77]: df.iloc[:,3]     # 访问列整数索引为3的列Out[77]: p1    Malep2    Nonep3    NoneName: gender, dtype: object03