bug with resampling

rsbowman

Hi,

I have some 5m data I'm trying to resample to hourly using cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes) but I get the following error on running cerebro.run()

Traceback (most recent call last):
  File "test.py", line 237, in <module>
    result = cerebro.run()
  File "/home/who/dev/py/trading/lib/python3.5/site-packages/backtrader/cerebro.py", line 1142, in run
    runstrat = self.runstrategies(iterstrat)
  File "/home/who/dev/py/trading/lib/python3.5/site-packages/backtrader/cerebro.py", line 1310, in runstrategies
    self._runnext(runstrats)
  File "/home/who/dev/py/trading/lib/python3.5/site-packages/backtrader/cerebro.py", line 1627, in _runnext
    self._check_timers(runstrats, dt0, cheat=True)
UnboundLocalError: local variable 'dt0' referenced before assignment

Indeed you can see at https://github.com/mementum/backtrader/blob/master/backtrader/cerebro.py#L1627 that the variable dt0 is not in scope, and again a few lines later. However, I'm not smart enough to know what the fix is -- when I changed dt0 to data0 I get a different error later....

Thanks so much for this great software!

ThatBlokeDave

Hey @rsbowman

Maybe you can also post your code? The cause might be something else other than resampling.

Also are you up to date?

backtrader

@rsbowman said in bug with resampling:

Indeed you can see at https://github.com/mementum/backtrader/blob/master/backtrader/cerebro.py#L1627 that the variable dt0 is not in scope

The error message only shows that you have triggered an error condition in which dt0 won't be defined. See below for a 1st good reason for it.

I have some 5m data I'm trying to resample to hourly
cerebro.resampledata(data, timeframe=bt.TimeFrame.Minutes)

That isolated (gives no real information about what's being done) already shows that:

5m data is being resampled to 1-minute data, which in itself is impossible.

In any case this diagnostic assumes that data has the right charateristics, which it may not due to the lack of context.

As pointed out by ThatBlokeDave, some code with the complete picture could really help.

rsbowman

Thank you both, and thank you for your patience. I think I have misunderstood the Cerebro.resampledata method. My example above was incorrect, here's the code I tried:

import datetime
# my backtrader __version__ = '1.9.58.122'
import backtrader as bt

if __name__ == '__main__':
    cerebro = bt.Cerebro()
    data = bt.feeds.GenericCSVData(
        dataname="some-5m-data.csv",
        fromdate=datetime.datetime(2015, 1, 1),
        todate=datetime.datetime(2015, 12, 31),
        openinterest=-1,
        timeframe=bt.TimeFrame.Ticks)

    cerebro.resampledata(d, timeframe=60*bt.TimeFrame.Minutes)
    #cerebro.resampledata(data, 
    #    timeframe=bt.TimeFrame.Ticks,
    #    compression=12)
    cerebro.run()

I'd like to turn my 5m data into hourly data. Now I think that multiplying 60 times TimeFrame.Minutes is not meaningful and instead I should do the thing that's commented out above... Is this correct? Compress 12 5m bars into a single 60m bar?

Thanks again!

backtrader

@rsbowman said in bug with resampling:

data = bt.feeds.GenericCSVData(
    dataname="some-5m-data.csv",
    fromdate=datetime.datetime(2015, 1, 1),
    todate=datetime.datetime(2015, 12, 31),
    openinterest=-1,
    timeframe=bt.TimeFrame.Ticks)

That is not being tagged as 5-minutes data. It is being tagged as Ticks

See here Docs - Data Resampling

The parameters timeframe and compression are explained for resampledata. They also apply to the data feed for the initial tagging.

See also the [See: Community - FAQ] for timeframe and compression

In any case:

@rsbowman said in bug with resampling:

Is this correct? Compress 12 5m bars into a single 60m bar?

You don't specify a transform, you specify your target

rsbowman

Ok, so in the case above the correct thing to do would be

# ...
data = bt.feeds.GenericCSVData(
    dataname="some-5m-data.csv",
    fromdate=datetime.datetime(2015, 1, 1),
    todate=datetime.datetime(2015, 12, 31),
    timeframe=bt.TimeFrame.Minutes,
    compression=5)
cerebro.resampledata(data, 
    timeframe=bt.TimeFrame.Minutes,
    compression=60)

This looks correct when I run it, thanks for your help!