Added fable/feast support

[moutikabdessabour]

Implemented to_basic methods that handle the output of fabletools:autoplot.fbl_ts and Added tests

library(dplyr)
library(forecast)
library(tsibble)

data <- tourism %>%
    filter(Region == "Melbourne") %>%
    `[`(, c("Quarter", "Trips", "Region")) %>%
    distinct(Quarter, .keep_all = TRUE) %>%
    as_tsibble(key = Region) 
p <- data %>%
    model(
        ets = ETS(Trips ~ trend("A")),
    ) %>%
    forecast(h = "5 years") %>%
    autoplot(data)

p

ggplotly(p)

.

[cpsievert]

Hmm, this doesn't seem right, do you have an example where this becomes relevant?

[moutikabdessabour]

this is actually a trick because the autoplot returns only a data.frame and GeomHilo... this results in only drawing the predictions without the line.

[cpsievert]

Come to think of it, I think most of this as well as to_basic.GeomAlluvium() could be replaced by ribbon_dat()?

[moutikabdessabour]

It's the same code with minor changes.

[cpsievert]

@danton267 would you mind filing an issue to request this feature so that it should up in this list of issues? https://github.com/plotly/plotly.R/issues/created_by/danton267

[cpsievert]

@moutikabdessabour what package is the model() function from in your example?

[cpsievert]

@moutikabdessabour if we're going to add support for {fable}/{feasts}, we should try our best to cover anything that one could create via forecast::forecast() + autoplot(), which would at least include forecast::GeomForecast

[moutikabdessabour]

@cpsievert Ok I will work on it.

[moutikabdessabour]

@cpsievert The model function is exported from the {fabletools} package