fix baseline plotting

[RaphaelS1]

Adds two sided CD line for baseline plots and changes lty type for baseline (see images)

[jemus42]

Gave it a quick go for the style = 1, there seems to be a bug I'll try sorting out

Error in `$<-.data.frame`:
! replacement has 0 rows, data has 17
Backtrace:
 1. global plot_results(...)
 3. mlr3benchmark:::autoplot.BenchmarkAggr(...)
 4. mlr3benchmark:::.plot_critdiff_1(...)
 6. base::`$<-.data.frame`(`*tmp*`, "baseline", value = `<lgl>`)

[jemus42]

Ah right I don't have access here -- anyway, at the top of .plot_critdiff_X added the following workaround:

  if (!is.null(cd$data$baseline)) {
    cd$data$baseline = as.logical(cd$data$baseline)
  } else {
    cd$data$baseline = FALSE
  }

Because obj$.__enclos_env__$private$.crit_differences (is it odd to call a private method like that btw?) does not necessarily return cd$data with a baseline variable, and as.logical then fails on NULL.

The proper workaround would probably be to have obj$.__enclos_env__$private$.crit_differences always include a baseline variable that is all 0 (or FALSE, which seems to be needed anyway?), such that further plot functions can reliably test for it.

[jemus42]

Can this be merged?

[jemus42]

Ping?

[RaphaelS1]

if you're pinging me, i have no idea? from our conversation it looks like probably but IDK?

[jemus42]

I guess I should be pinging @sebffischer to merge this 😅