Python pandas ヒストリカルデータのサマータイムを考慮していませんでした

df[df.index.dayofweek == 5].tail()

 open high low close volume
time
2017-12-02 112.107 112.282 112.099 112.176 2201.3
2017-12-09 113.489 113.525 113.454 113.461 3260.2
2017-12-16 112.581 112.636 112.562 112.585 3219.1
2017-12-23 113.268 113.343 113.254 113.260 3561.4
2017-12-30 112.680 112.714 112.647 112.658 5856.3

import calendar
import datetime
import functools
import pandas as pd

def read(filepath):
    dtype = { 'time': str, 'open': float, 'high': float, 'low': float, 'close': float, 'volume': float }
    names = ['time', 'open', 'high', 'low', 'close', 'volume']
    df = pd.read_csv(filepath, dtype=dtype, header=0, index_col='time', names=names, parse_dates=['time'])
    return df

def dstdt(year, month, week, hours):
    day = [i for i in calendar.Calendar().itermonthdays2(year, month) if i[0] != 0 and i[1] == calendar.SUNDAY][week][0]
    return datetime.datetime(year, month, day, 2, 0) + datetime.timedelta(hours=-hours)

def dst(year):
    dic = {
        1918: [(4, 0, -5), (10, -1, -4)],
        2007: [(3, 1, -5), (11, 0, -4)]
    }
    key = functools.reduce(lambda x, y: x if year < y else y, dic.keys(), 0)
    if key == 0:
        return []
    return [dstdt(year, i[0], i[1], i[2]) for i in dic[key]]

def offset(df):
    data = { 'open': df['open'].values, 'high': df['high'].values, 'low': df['low'].values, 'close': df['close'].values, 'volume': df['volume'].values }
    columns = ['open', 'high', 'low', 'close', 'volume']
    l = [dst(i) for i in range(2001, 2018)]
    l2 = [(i[0] <= df.index) & (df.index < i[1]) for i in l]
    is_dst = functools.reduce(lambda x, y: x | y, l2)
    eest = df.index[is_dst] + pd.DateOffset(hours=3) # サマータイムは UTC+3
    eet = df.index[~is_dst] + pd.DateOffset(hours=2) # サマータイム以外は UTC+2
    index = pd.concat([eest.to_series(), eet.to_series()]).sort_values()
    df = pd.DataFrame(data=data, columns=columns, index=index)
    return df

def resample(df, rule):
    df2 = df.resample(rule).ohlc()
    df3 = df.resample(rule).sum()
    data = { 'open': df2['open']['open'].values, 'high': df2['high']['high'].values, 'low': df2['low']['low'].values, 'close': df2['close']['close'].values, 'volume': df3['volume'].values }
    columns = ['open', 'high', 'low', 'close', 'volume']
    df4 = pd.DataFrame(data=data, columns=columns, index=df2.index).dropna()
    return df4

def main(filepath, rule, decimals, to):
    df = read(filepath)
    df2 = offset(df)
    df3 = resample(df2, rule)
    df4 = df3.round({ 'open': decimals[0], 'high': decimals[0], 'low': decimals[0], 'close': decimals[0], 'volume': decimals[1] })
    df4.to_csv(to)
    return to, df4['close'].count()

%time main(f'~/Documents/data/USDJPY_1 Min_Bid_2000.01.01_2018.01.01.csv', 'd', (3, 1), f'~/Documents/data/USDJPY_D.csv')

df = read(f'~/Documents/data/d/USDJPY_D.csv')
df[df.index.dayofweek == 5]