hybridModel behavior when component models perfectly fit data

[dashaub]

Consider the case the individual component models perfectly fit the time series.

hm <- hybridModel(ts(1:100, f = 4), weights = "insample.errors")
hm #weights of 0/NaN
accuracy(hm)#NA/NaN
fitted(hm)#NA/NaN
forecast(hm)# NaN point forecasts

This is definitely a corner case and not likely to occur with real data, however, it's worth considering what the correct behavior should be (e.g. throw error, revert to weights = "equal", etc)

[dashaub]

Closed here: 5f0ddeaafbdbf04fda13de368b1b8ac9dac952e7