Many implementations of filtering models (e.g. ARIMA and ETS) are implemented in ways that require the data (and forecast period) is contiguous. As a result, they have to implement their own checks, which are error-prone (tidyverts/fable#430).
A better place for these checks would be here, where a function that verifies contiguity of inputs can be inherited/used by models that need it. These checks could be easier to do using mixtime, rather than managing the different data classes of tsibble, and lack of operation support with intervals, etc.
Many implementations of filtering models (e.g. ARIMA and ETS) are implemented in ways that require the data (and forecast period) is contiguous. As a result, they have to implement their own checks, which are error-prone (tidyverts/fable#430).
A better place for these checks would be here, where a function that verifies contiguity of inputs can be inherited/used by models that need it. These checks could be easier to do using mixtime, rather than managing the different data classes of tsibble, and lack of operation support with intervals, etc.