@ab_trader I don't agree your opinion,but I try my best to use english.The reason which I use chinese,maybe @Jingsi is a chinese,who can speak chinese more accurately. just as two persons who speaks english,they have a trip to Japan,when they communicate,you let them use Japanese?

@Jingsi 太恐怖了，你回测90万个期货合约？吓死宝宝的节奏。你是全球期货市场的合约吗？中国的其实没那么多。
看了一下chainer.py,还没有想到要怎么做能够加快速度。我现在5000+的数据都需要两分钟多了，能分享下你90万的回测速度怎么那么快吗？

you can have a look at pandafeed.py or chainer.py and implements your own datafeeds, it's very simple, especially for backtest, i think it's no more than load data in _load()

but if you want live trading, you have to deal with disconnection and reconnection and backfill the data properly

您的这段话，我没能完全明白是什么意思，如果您用中文，可以用中文分享下吗？

@vbs while optimize encountered different situation according different strategy, some very simple may run seccessfully,some complicated may not run,some unique may just walk and have no error reporting.I don't have so much time and experience to debug this.Now,I debug the analyzers,there are some errors.

@backtrader Yes,I use stdstats=False to prohibit the obversers. At first, I add all the analyzers when I optimize my strategy,but,I find,I just need several indicator to test my strategy and all analyzer consume too much time,so I reduce them.

However,I don't use the backtrader's original optimize function,because there is something wrong when I do it,I can't sovle it,so every time,my backtest must consume 13%(data._preload()),is there some ways to pcikle the preloaded data?

my conclusion is that :

1. reduce the data you load,the more data,the more time consumes(maybe linear,for example,5000+,10 mins,2500+,5mins+).
2. when you optimize,you'd better prohibit the obverser.
3. reduce the analyzer you add,may speed up your strategy.

@tianjixuetu hahaha,just as which function consume time,when I prohibit the self._next_observers() and reduce the analyzer,
my every backtest time drop from 280s to 120s.

@tianjixuetu

code                                             percent of total time
self._next_analyzers(minperstatus, once=True)            25.55%
self._next_observers(minperstatus, once=True)            25.42%
self.prenext()/self.next()                               11.24%
dts = [d.advance_peek() for d in datas]                   4.68%
datas[i].advance()                                        3.76%
self._brokernotify()                                      9.36%
data._start()                                             2.6%
data._preload()                                            13%                                             
total                                                    95.61%                                

ok,from the backtest time consume information,it's better to delete self._next_analyzers() and self._next_observers(),if possible,or at last,speed up self._next_analyzers() and self._next_observers() and data._preload()

@tianjixuetu

Line #      Hits         Time  Per Hit   % Time  Line Contents
==============================================================
   291                                               @profile
   292                                               def _oncepost(self, dt):
   293      2398      97209.0     40.5      0.0          for indicator in self._lineiterators[LineIterator.IndType]:
   294                                                       if len(indicator._clock) > len(indicator):
   295                                                           indicator.advance()
   296
   297      2398      27533.0     11.5      0.0          if self._oldsync:
   298                                                       # Strategy has not been reset, the line is there
   299                                                       self.advance()
   300                                                   else:
   301                                                       # strategy has been reset to beginning. advance step by step
   302      2398     337940.0    140.9      0.0              self.forward()
   303
   304      2398     116440.0     48.6      0.0          self.lines.datetime[0] = dt
   305      2398    4619271.0   1926.3      0.1          self._notify()
   306
   307      2398  105163575.0  43854.7      2.5          minperstatus = self._getminperstatus()
   308      2398      30777.0     12.8      0.0          if minperstatus < 0:
   309                                                       self.next()
   310      2398      10568.0      4.4      0.0          elif minperstatus == 0:
   311                                                       self.nextstart()  # only called for the 1st value
   312                                                   else:
   313      2398  735079500.0 306538.6     17.6              self.prenext()
   314
   315      2398 1672866860.0 697609.2     40.0          self._next_analyzers(minperstatus, once=True)
   316      2398 1666655645.0 695019.0     39.8          self._next_observers(minperstatus, once=True)
   317
   318      2398     190217.0     79.3      0.0          self.clear()

@tianjixuetu 在self._runonce(runstrats)中

      代码                    在该函数中占用的时间的比例         占用的总时间比例
dts = [d.advance_peek() for d in datas]         5.6%                 4.68%
datas[i].advance()                               4.5%                3.76%
self._brokernotify()                             11.2%               9.36%

@Saeed-Rahman Very good! I find that, MBATS can live trading on Oanda,is it using v20 API?

@Picture-Perfect
if you set data name in add data,you can use:

self.log("buy result: data:{},buy_price:{},buy_cost:{},commission:{}".format(order.p.data._name,
                            order.executed.price,order.executed.value,order.executed.comm))

order.p.data._name