Error with two data feeds with different history length

aartajew

Hi guys,

I've found 2 problems when using 2 data feeds - one with history longer than the other:

from datetime import datetime
import backtrader as bt

class SmaCross(bt.SignalStrategy):
    def __init__(self):
        # ko
        crossover1 = bt.ind.CrossOver(self.data0, self.data1)
        # ok
        crossover2 = bt.ind.CrossOver(self.data1, self.data0)

        # ko
        self.data0 - self.data1
        # ok
        self.data1 - self.data0

cerebro = bt.Cerebro()
cerebro.addstrategy(SmaCross)

# data since 2004-11-18
data0 = bt.feeds.YahooFinanceCSVData(dataname='data/GLD.csv', 
                                     fromdate=datetime(2006, 1, 1),
                                     todate=datetime(2007, 1, 1))
cerebro.adddata(data0)

# data since 2006-04-28
data1 = bt.feeds.YahooFinanceCSVData(dataname='data/SLV.csv', 
                                     fromdate=datetime(2006, 1, 1),
                                     todate=datetime(2007, 1, 1))
cerebro.adddata(data1)

cerebro.run()
cerebro.plot()

The first error is in indicators/crossover.py:

$ python pair_bug.py
Traceback (most recent call last):
  File "pair_bug.py", line 31, in <module>
    cerebro.run()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/cerebro.py", line 1127, in run
    runstrat = self.runstrategies(iterstrat)
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/cerebro.py", line 1293, in runstrategies
    self._runonce(runstrats)
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/cerebro.py", line 1652, in _runonce
    strat._once()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/lineiterator.py", line 297, in _once
    indicator._once()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/lineiterator.py", line 297, in _once
    indicator._once()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/lineiterator.py", line 297, in _once
    indicator._once()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/lineiterator.py", line 318, in _once
    self.once(self._minperiod, self.buflen())
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/indicators/crossover.py", line 58, in once
    d = d0array[i] - d1array[i]
IndexError: array index out of range

The second is in linebuffer.py:

$ python pair_bug.py
Traceback (most recent call last):
  File "pair_bug.py", line 31, in <module>
    cerebro.run()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/cerebro.py", line 1127, in run
    runstrat = self.runstrategies(iterstrat)
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/cerebro.py", line 1293, in runstrategies
    self._runonce(runstrats)
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/cerebro.py", line 1652, in _runonce
    strat._once()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/lineiterator.py", line 297, in _once
    indicator._once()
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/linebuffer.py", line 631, in _once
    self.once(self._minperiod, self.buflen())
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/linebuffer.py", line 755, in once
    self._once_op(start, end)
  File "/Users/adam/.pyenv/versions/backtrader/lib/python3.5/site-packages/backtrader/linebuffer.py", line 772, in _once_op
    dst[i] = op(srca[i], srcb[i])
IndexError: array index out of range

As you can see in my code sample, it works when you flip the data feeds - if you put the shorter history data first (data1), it works fine, otherwise it throws error.

My env is macOS Catalina 10.15.3, Python 3.5.9 with packages:

$ pip freeze
backtrader==1.9.74.123
cycler==0.10.0
kiwisolver==1.1.0
matplotlib==3.0.3
numpy==1.18.3
pyparsing==2.4.7
python-dateutil==2.8.1
six==1.14.0

Any help would be appreciated :)

ab_trader

use

cerebro.run(runonce=False)

this will turn off vectorized mode.

aartajew

It worked, thanks!

But isn't vectorized mode much faster for optimization, etc.?

ab_trader

yes, faster. but in your case you have empty data bars, and looks like this doesn't work with vectorized approach.