DonchianChannels and input_sampling

[powellnorma]

I think currently the input_sampling works by just removing the last output value and recalculating it with the current input.
However, for DonchianChannels, I think it should keep the lowest low and highest high of any intermediary candle

[nardew]

Hi, yes, what sampling does is it simply uses the last value from the window. It means that indicator values after one 5min candle are potentially different than after five 1min candles in the same period.

What you are describing e.g. for Donchian Channels makes sense but that's not how it is designed at moment. I may check what it takes to adapt the behaviour as you describe it but I fear it will not be trivial.

[powellnorma]

I may check what it takes to adapt the behaviour as you describe it but I fear it will not be trivial.

One way could be to handle the sampling logic by creating separate "streams" of candles. That way, each indicator can look at the past n candles of its declared timeframe. That would be similar to how https://github.com/MerlinR/Hexital does it.

It could then also make sense to adopt the concept of an "indicator group", which manages subgroups of indicators by a specific timeframe. So the e.g. 4h subgroup then just computes the 4h candles once, and passes them down to the indicators. That way, this re-sampling calculation is only done once (for each timeframe).

Some pseudo code:

ema_4h = talipp.EMA(timeframe='4h')
ema_1h = talipp.EMA(timeframe='1h')
ema_1m = talipp.EMA(timeframe='1m')
group = talipp.IndicatorGroup(indicators=[ema_4h, ema_1h, ema_1m], input_timeframe='1m')

# __init__ has now created group.subgroups: dict[timedelta, list[Indicator]]
# It might also set some attribute Indicator.skip_resampling to let the Indicator know that resampling is already handled for it

# User just does:
group.add(candle_1min)
# And the "right things" will happen in the background