Good afternoon, I just upgraded to 4.3.0 and it seems to have caused my existing hierarchical time series forecast code to break. I confirmed that the error only shows up in 4.3.0 by running the code fine on a machine that uses 4.2.3. All packages were up to date with same versions. The error is:

Error in method == "comb" && fmethod == "rw" :

'length = 3' in coercion to 'logical(1)'

3.

forecast.gts(y, h = forecast_horizon, FUN = function(x) do.call(my_method,

list(x, lambda = "auto")))

2.

forecast(y, h = forecast_horizon, FUN = function(x) do.call(my_method,

list(x, lambda = "auto")))

1.

forecast_school(school_name, school_time_series, start_year,

freq, current_year, current_cycle_month, current_month_name,

nodes, funnel_stages)

This appears to be coming from this section of the code:

Blockquote

if (my_method == "nnetar") {

allf <- forecast(y, h = forecast_horizon, FUN = function(x) do.call(my_method, list(x, repeats = 200, lambda = "auto", P = 1, maxit = 1000, decay = 1, scale.inputs = FALSE, size = 2)))

} else if (my_method == "auto.arima") {

allf <- forecast(y, h = forecast_horizon, FUN = function(x) do.call(my_method, list(x, max.d = 1, max.D = 1)))

} else {

allf <- forecast(y, h = forecast_horizon, FUN = function(x) do.call(my_method, list(x, lambda = "auto")))

}

Blockquote

It appears that the `forecast()`

function is calling `forecast.gts()`

on the `hts`

object. It is possible that some changes in R or the forecast package resulted in the `forecast()`

function behaving differently. I then switched to explicitly calling the hts explicitly with forecast.hts.

Blockquote

if (my_method == "nnetar") {

allf <- forecast.hts(y, h = forecast_horizon, FUN = function(x) do.call(my_method, list(x, repeats = 200, lambda = "auto", P = 1, maxit = 1000, decay = 1, scale.inputs = FALSE, size = 2)))

} else if (my_method == "auto.arima") {

allf <- forecast.hts(y, h = forecast_horizon, FUN = function(x) do.call(my_method, list(x, max.d = 1, max.D = 1)))

} else {

allf <- forecast.hts(y, h = forecast_horizon, FUN = function(x) do.call(my_method, list(x, lambda = "auto")))

Blockquote

But then I got this error:

Error in forecast.hts(y, h = forecast_horizon, FUN = function(x) do.call(my_method, :

could not find function "forecast.hts"

At this point I just ran the simple code:

forecast.hts(hts(), h = 1)

And got the same error. I reinstalled hts but the problem has persisted. Any advice on how to proceed is appreciated. I can downgrade to an older version of R but this is not a long term solution.