draw() fails with parametric = TRUE if there aren't any parametric effects to actually plot

[gavinsimpson]

Consider this example from the GAM course:

# load data
rats_url <- "https://bit.ly/rat-hormone"
rats <- read_table(rats_url, col_types = "dddddddddddd-")
# ignore the warning - it"s due to trailing white space at the ends of each
#   row in the file

rats <- rats %>%
    mutate(treatment = fct_recode(factor(group, levels = c(1, 2, 3)),
                                  Low = "1",
                                  High = "2",
                                  Control = "3"),
           treatment = fct_relevel(treatment, c("Control", "Low", "High")),
           subject = factor(subject))

m1_gam <- gam(response ~ treatment:transf_time +
                  s(subject, bs = "re") +
                  s(subject, transf_time, bs = "re"),
              data = rats, method = "REML")

If we now plot this including the random effects:

> draw(m1_gam, parametric = TRUE)
Interaction terms are not currently supported.
Error in `unnest()`:
! Can't select columns that don't exist.
✖ Column `data` doesn't exist.
Run `rlang::last_trace()` to see where the error occurred.

The first warning is right - we don't allow plotting of interaction parametric effects, but as these are the only effects something goes wrong and we get an error.