查看“Pandas dataframe”的源代码

===pandas dataframe conversion===
===dict ===
==== dict -> dataframe====
 d1 = {"columns":["Apple","Pear"],"data":[[12,0], [8,7], [1, 9]]}
 df2 = pd.DataFrame(d1['data'])
    Apple  Pear
 0     12     0
 1      8     7
 2      1     9
 
 df2.columns=d1['columns']
 Index(['Apple', 'Pear'], dtype='object')
====dataframe -> dict====
Syntax: DataFrame.to_dict(orient=’dict’, into=)

Parameters:
* orient: String value, (‘dict’, ‘list’, ‘series’, ‘split’, ‘records’, ‘index’) Defines which dtype to convert Columns(series into). For example, ‘list’ would return a dictionary of lists with Key=Column name and Value=List (Converted series).
* into: class, can pass an actual class or instance. For example in case of defaultdict instance of class can be passed. Default value of this parameter is dict.
==== Example====
 ...
 df2.to_dict('split')
 {'index': [0, 1, 2], 'columns': ['Apple', 'Pear'], 'data': [[12, 0], [8, 7], [1, 9]]}
 
 df2.to_dict('records')
 [{'Apple': 12, 'Pear': 0}, {'Apple': 8, 'Pear': 7}, {'Apple': 1, 'Pear': 9}]
===list===
====list -> dataframe====
　　See also: dict -> dataframe
====dataframe -> list====
 a1 = df1.values    # values方法将dataframe转为numpy.ndarray
 l1 = a1.tolist()
 l1[0]              # get frist value
 usys.utime(l1[0][7].value/10**9) * P.S. 日期型的字段转换后格式：Timestamp('2017-04-13 13:48:32')，pandas._libs.tslibs.timestamps.Timestamp. 可以使用 usys.utime(l1[0][7].value/10**9) 转换。