Rolling Window and ema for crypto

Hi Dario!

My kind of questions!! (my previous life was formal methods)

Short answer:

Use this:

for symbol in self.symbols:
	self.fast_ema[symbol] = self.EMA(symbol,fast_moving_average,Resolution.Daily)
    self.ema_window[symbol] = RollingWindow[float](2)
    self.fast_ema[symbol].Updated += lambda sender,updated,symbol=symbol: self.ema_window[symbol].Add(self.fast_ema[symbol].Current.Value)

Long answer:

Your lambda is really a “closure”. It has an unbound variable, sumbol, which needs to be bound dynamically. In Python, closures are done syntactically, not dynamically. So, they depend on scopes, syntactical scopes. In your first code, the scope of the lambda is the for loop. Thus, each lambda created has the same scope and thus the same “capture of symbol” ("unbound variable capture" is the term used to say that an unbound variable is dynamically bound in the function and creates a closure). So……… all lambdas end up using the same version of the symbol variable, i.e., the last one (the first lambda first use the first symbol, then on the second iteration, both lambdas point to the second symbol, then, on the third iteration, all 3 point to the third one, etc).

One way to have dynamic capture is to add a param to the lambda and set it to the dynamic variable by default, like this:

for symbol in self.symbols:
	self.fast_ema[symbol] = self.EMA(symbol,fast_moving_average,Resolution.Daily)
    self.ema_window[symbol] = RollingWindow[float](2)
    self.fast_ema[symbol].Updated += lambda sender,updated,symbol=symbol: self.ema_window[symbol].Add(self.fast_ema[symbol].Current.Value)

The name of the param needs to be the same as the unbound variable. You are then basically cheating as the unbound variable is not unbound anymore so you do not have a closure anymore, there are no unbound variable to capture anymore.

The interface to the event handler is to have two params, not 3 (sender and updated). But the third one that you have has a default value set to the bound variable, so it is not needed, it is fixed dynamically has lambdas are created…

I have not tested this 😊. Just using neurons here - which can lead to surprises…

Fred