先簡單的了解下日期和時(shí)間數(shù)據(jù)類型及工具

python標(biāo)準(zhǔn)庫包含于日期（date）和時(shí)間（time）數(shù)據(jù)的數(shù)據(jù)類型，datetime、time以及calendar模塊會被經(jīng)常用到。

datetime以毫秒形式存儲日期和時(shí)間，datetime.timedelta表示兩個(gè)datetime對象之間的時(shí)間差。

給datetime對象加上或減去一個(gè)或多個(gè)timedelta，會產(chǎn)生一個(gè)新的對象

from datetime import datetime

from datetime import timedelta

now = datetime.now()
now

    datetime.datetime(2017, 6, 27, 15, 56, 56, 167000)

datetime參數(shù)：datetime(year, month, day[, hour[, minute[, second[, microsecond[,tzinfo]]]]])

delta = now - datetime(2017,6,27,10,10,10,10)

delta

    datetime.timedelta(0, 20806, 166990)

delta.days

    0

delta.seconds

    20806

delta.microseconds

    166990

只有這三個(gè)參數(shù)了！

datetime模塊中的數(shù)據(jù)類型

字符串和datetime的相互轉(zhuǎn)換

1）python標(biāo)準(zhǔn)庫函數(shù)

日期轉(zhuǎn)換成字符串：利用str 或strftime

字符串轉(zhuǎn)換成日期：datetime.strptime

stamp = datetime(2017,6,27)

str(stamp)

    '2017-06-27 00:00:00'

stamp.strftime('%y-%m-%d')#%Y是4位年，%y是2位年

    '17-06-27'

#對多個(gè)時(shí)間進(jìn)行解析成字符串

date = ['2017-6-26','2017-6-27']

datetime2 = [datetime.strptime(x,'%Y-%m-%d') for x in date]

datetime2

    [datetime.datetime(2017, 6, 26, 0, 0), datetime.datetime(2017, 6, 27, 0, 0)]

2）第三方庫dateutil.parser的時(shí)間解析函數(shù)

from dateutil.parser import parse

parse('2017-6-27')

    datetime.datetime(2017, 6, 27, 0, 0)

parse('27/6/2017',dayfirst =True)

    datetime.datetime(2017, 6, 27, 0, 0)

3）pandas處理成組日期

pandas通常用于處理成組日期，不管這些日期是DataFrame的軸索引還是列，to_datetime方法可以解析多種不同的日期表示形式。

date

    ['2017-6-26', '2017-6-27']

import pandas as pd

pd.to_datetime(date)

    DatetimeIndex(['2017-06-26', '2017-06-27'], dtype='datetime64[ns]', freq=None)

datetime 格式定義

pandas時(shí)間序列基礎(chǔ)以及時(shí)間、日期處理

pandas最基本的時(shí)間序列類型就是以時(shí)間戳（時(shí)間點(diǎn)）（通常以python字符串或datetime對象表示）為索引的Series：

dates = ['2017-06-20','2017-06-21',         '2017-06-22','2017-06-23','2017-06-24','2017-06-25','2017-06-26','2017-06-27']

import numpy as np
ts = pd.Series(np.random.randn(8),index = pd.to_datetime(dates))

ts

    2017-06-20    0.788811
    2017-06-21    0.372555
    2017-06-22    0.009967
    2017-06-23   -1.024626
    2017-06-24    0.981214
    2017-06-25    0.314127
    2017-06-26   -0.127258
    2017-06-27    1.919773
    dtype: float64

ts.index

    DatetimeIndex(['2017-06-20', '2017-06-21', '2017-06-22', '2017-06-23',
                   '2017-06-24', '2017-06-25', '2017-06-26', '2017-06-27'],
                  dtype='datetime64[ns]', freq=None)

pandas不同索引的時(shí)間序列之間的算術(shù)運(yùn)算會自動(dòng)按日期對齊

ts[::2]#從前往后每隔兩個(gè)取數(shù)據(jù)

    2017-06-20    0.788811
    2017-06-22    0.009967
    2017-06-24    0.981214
    2017-06-26   -0.127258
    dtype: float64

ts[::-2]#從后往前逆序每隔兩個(gè)取數(shù)據(jù)

    2017-06-27    1.919773
    2017-06-25    0.314127
    2017-06-23   -1.024626
    2017-06-21    0.372555
    dtype: float64

ts + ts[::2]#自動(dòng)數(shù)據(jù)對齊

    2017-06-20    1.577621
    2017-06-21         NaN
    2017-06-22    0.019935
    2017-06-23         NaN
    2017-06-24    1.962429
    2017-06-25         NaN
    2017-06-26   -0.254516
    2017-06-27         NaN
    dtype: float64

索引為日期的Series和DataFrame數(shù)據(jù)的索引、選取以及子集構(gòu)造

方法：
1）.index[number_int]

2)[一個(gè)可以被解析為日期的字符串]

3）對于，較長的時(shí)間序列，只需傳入‘年’或‘年月’可返回對應(yīng)的數(shù)據(jù)切片

4）通過時(shí)間范圍進(jìn)行切片索引

ts

    2017-06-20    0.788811
    2017-06-21    0.372555
    2017-06-22    0.009967
    2017-06-23   -1.024626
    2017-06-24    0.981214
    2017-06-25    0.314127
    2017-06-26   -0.127258
    2017-06-27    1.919773
    dtype: float64

ts[ts.index[2]]

    0.0099673896063391908

ts['2017-06-21']#傳入可以被解析成日期的字符串

    0.37255538918121028

ts['21/06/2017']

    0.37255538918121028

ts['20170621']

    0.37255538918121028

ts['2017-06']#傳入年或年月

    2017-06-20    0.788811
    2017-06-21    0.372555
    2017-06-22    0.009967
    2017-06-23   -1.024626
    2017-06-24    0.981214
    2017-06-25    0.314127
    2017-06-26   -0.127258
    2017-06-27    1.919773
    dtype: float64

ts['2017-06-20':'2017-06-23']#時(shí)間范圍進(jìn)行切片

    2017-06-20    0.788811
    2017-06-21    0.372555
    2017-06-22    0.009967
    2017-06-23   -1.024626
    dtype: float64

帶有重復(fù)索引的時(shí)間序列

1）.index.is_unique檢查索引日期是否是唯一的

2）對非唯一時(shí)間戳的數(shù)據(jù)進(jìn)行聚合，通過groupby，并傳入level = 0（索引的唯一一層）

dates = pd.DatetimeIndex(['2017/06/01','2017/06/02','2017/06/02','2017/06/02','2017/06/03'])
dates

    DatetimeIndex(['2017-06-01', '2017-06-02', '2017-06-02', '2017-06-02',
                   '2017-06-03'],
                  dtype='datetime64[ns]', freq=None)

dup_ts = pd.Series(np.arange(5),index = dates)
dup_ts

    2017-06-01    0
    2017-06-02    1
    2017-06-02    2
    2017-06-02    3
    2017-06-03    4
    dtype: int32

dup_ts.index.is_unique

    False

dup_ts['2017-06-02']

    2017-06-02    1
    2017-06-02    2
    2017-06-02    3
    dtype: int32

grouped = dup_ts.groupby(level=0).mean()

grouped

    2017-06-01    0
    2017-06-02    2
    2017-06-03    4
    dtype: int32

dup_df = pd.DataFrame(np.arange(10).reshape((5,2)),index = dates )
dup_df

grouped_df = dup_df.groupby(level=0).mean()##針對DataFrame

grouped_df

總結(jié)

該篇博客主要內(nèi)容：

1）字符串、日期的轉(zhuǎn)換方法

2）日期和時(shí)間的主要python，datetime、timedelta、pandas.to_datetime等

3）以時(shí)間為索引的Series和DataFrame的索引、切片

4）帶有重復(fù)時(shí)間索引時(shí)的索引，.groupby(level=0)應(yīng)用