timer problem

[Li Mike]

Dear Backtrader, I am testing timer. It seems the notify_timer is triggered on wrong. In the follow code, I expect it triggers on first day on each month.
But the output is as follows, you can see that : timer 2013-02-28, 2013-03-28, 2013-04-30 is wrong. Am I missing something. The problem seems relate to data. you can downloan the code and data at this link.
link：https://pan.baidu.com/s/1mgH4nuPEoiRrSdiGLAXM6w
password：3n45

output:
ticker AAPL
ticker XOM
timer 2012-02-28
timer 2012-03-01
timer 2012-04-02
timer 2012-05-01
timer 2012-06-01
timer 2012-07-02
timer 2012-08-01
timer 2012-09-04
timer 2012-10-01
timer 2012-11-01
timer 2012-12-03
timer 2013-01-02
timer 2013-02-01
timer 2013-02-28
timer 2013-03-28
timer 2013-04-30
timer 2013-05-31
timer 2013-06-28
timer 2013-07-31
timer 2013-08-30
timer 2013-09-30
timer 2013-10-31
...

from datetime import datetime
import pandas as pd
import backtrader as bt
import os.path 
import sys  



modpath = os.path.dirname(os.path.abspath(sys.argv[0]))

datapath = os.path.join(modpath, './survivorship-free/tickers.csv')
tickers = pd.read_csv(datapath, header=None)[1].tolist()

maxtickersNum = 2
tickers = tickers[0:maxtickersNum] 

####################################


class Strategy(bt.Strategy):


    def __init__(self): 
        

        self.add_timer(
            when=bt.Timer.SESSION_START,
            monthdays=[1], 
            monthcarry=True, 
            
        )

       

    def notify_timer(self, timer, when, *args, **kwargs):        
        print('timer',self.data0.datetime.date(0))


    


##########################
# main
#########################

cerebro = bt.Cerebro(stdstats=False)


spy = bt.feeds.YahooFinanceCSVData(
    dataname=os.path.join(modpath, './survivorship-free/SPY.csv'),
    fromdate=datetime(2012, 2, 28),
    todate=datetime(2018, 2, 28),
    plot=False)
cerebro.adddata(spy)  # S&P 500

for ticker in tickers:
    print('ticker', ticker)
    df = pd.read_csv(
        f"survivorship-free/{ticker}.csv", parse_dates=True, index_col=0)
    
    data = bt.feeds.PandasData(
        dataname=df,
        datetime=None,  
        open=0, 
        high=1,  
        low=2,  
        close=3,  
        volume=4,  
        openinterest=-1,  
        plot=False

    )
    
    if len(df) > 100: 
        cerebro.adddata(data)

cerebro.addstrategy(Strategy)
results = cerebro.run()

[vladisld]

Could it be the same problem as described in the following post ? : https://community.backtrader.com/topic/3093/timer-monthday-reports-trading-on-days-prior-to-target

[Li Mike]

Seems not the same with that post. I use self.data0.datetime.date(0), but the result is still wrong.

[Li Mike]

@vladisld if output when, the when is correct, see below. Any idea?
```
def notify_timer(self, timer, when, *args, **kwargs):
print('timer',self.data0.datetime.date(0), 'when', when)

ticker AAPL
ticker XOM
timer 2012-02-28 when 2012-02-28 00:00:00
timer 2012-03-01 when 2012-03-01 00:00:00
timer 2012-04-02 when 2012-04-02 00:00:00
timer 2012-05-01 when 2012-05-01 00:00:00
timer 2012-06-01 when 2012-06-01 00:00:00
timer 2012-07-02 when 2012-07-02 00:00:00
timer 2012-08-01 when 2012-08-01 00:00:00
timer 2012-09-04 when 2012-09-04 00:00:00
timer 2012-10-01 when 2012-10-01 00:00:00
timer 2012-11-01 when 2012-11-01 00:00:00
timer 2012-12-03 when 2012-12-03 00:00:00
timer 2013-01-02 when 2013-01-02 00:00:00
timer 2013-02-01 when 2013-02-01 00:00:00
timer 2013-02-28 when 2013-03-01 00:00:00
timer 2013-03-28 when 2013-04-01 00:00:00
timer 2013-04-30 when 2013-05-01 00:00:00
timer 2013-05-31 when 2013-06-03 00:00:00
timer 2013-06-28 when 2013-07-01 00:00:00
timer 2013-07-31 when 2013-08-01 00:00:00
timer 2013-08-30 when 2013-09-03 00:00:00
timer 2013-09-30 when 2013-10-01 00:00:00

[vladisld]

I still think it could be related to the time zone difference as described in the referenced post above. Could you please re check? (See the last response in the referenced post)

[Li Mike]

@vladisld When I use

def notify_timer(self, timer, when, *args, **kwargs):        
       print('timer',self.data0.datetime.date(0), self.data0.datetime.datetime(0), 'when', when)

Then output is as follows. See the last row. I use daily bars. I found that if we use data = bt.feeds.PandasData(...), then the datetime of the data bar is for example 2013-03-28 00:00:00, not 2013-03-28 23:59:59. This seems cause the problem. Any idea?

timer 2012-02-28 2012-02-28 23:59:59.999989 when 2012-02-28 00:00:00
timer 2012-03-01 2012-03-01 23:59:59.999989 when 2012-03-01 00:00:00
timer 2012-04-02 2012-04-02 23:59:59.999989 when 2012-04-02 00:00:00
timer 2012-05-01 2012-05-01 23:59:59.999989 when 2012-05-01 00:00:00
timer 2012-06-01 2012-06-01 23:59:59.999989 when 2012-06-01 00:00:00
timer 2012-07-02 2012-07-02 23:59:59.999989 when 2012-07-02 00:00:00
timer 2012-08-01 2012-08-01 23:59:59.999989 when 2012-08-01 00:00:00
timer 2012-09-04 2012-09-04 23:59:59.999989 when 2012-09-04 00:00:00
timer 2012-10-01 2012-10-01 23:59:59.999989 when 2012-10-01 00:00:00
timer 2012-11-01 2012-11-01 23:59:59.999989 when 2012-11-01 00:00:00
timer 2012-12-03 2012-12-03 23:59:59.999989 when 2012-12-03 00:00:00
timer 2013-01-02 2013-01-02 23:59:59.999989 when 2013-01-02 00:00:00
timer 2013-02-01 2013-02-01 23:59:59.999989 when 2013-02-01 00:00:00
timer 2013-02-28 2013-02-28 23:59:59.999989 when 2013-03-01 00:00:00
timer 2013-03-28 2013-03-28 23:59:59.999989 when 2013-04-01 00:00:00

[vladisld]

good point - you may be right. I'll try to debug it further - will update.