Add additional checks for forecast data time periods

[mitchelloharawild]

https://stackoverflow.com/questions/60264409/forecast-initialization-with-the-fable-package

Error should be raised if new_data doesn't continue from the training data.

[Timood]

Hi Mitchell and others,

You might also want to adapt the check on the extra variables required to produce forecasts.

The code example below now produces an error, saying that extra variables might be required: (fable 0.3.1)

fit <- insurance %>%
  model(ARIMA(Quotes ~ pdq(d = 0) +
                lag(TVadverts)))

insurance_future <- new_data(insurance, 1)

fit %>%
  forecast(insurance_future) -> fc

However, there are no extra variables required to produce this forecast. The model contains a lagged predictor and the forecast is demanded for just one step ahead.

Edit: there is actually an easy way to prevent this error by using new_data(insurance, 1, keep_all = TRUE).