legendrank (bar chart example)

提供された R reprex を実行し、plotly_json(px) を調べて、bar trace に対する legendrank のマッピングがどのように表現されているかに注目します。生成された凡例の順序を要求された rank と比較します。完了とは、trace が rank 順に表示されるか、issue が説明され、適切な regression test でカバーされていることを意味します。

This bug report concerns R plotly 4.10.1.
I was going to file a report regarding a more complex composite chart; involving bars and lines; as I noticed when there was an NA(or NaN) value to plot against the Y axis , the legend orders got scrambled. I was able to directly address this shortcoming in my own code by na.omit() on my input dataset. However , in the course of trying to simplify my code down to be more minimal for this reprex, I encountered a more basic issue of the legendrank seemingly not being respected

library(tidyverse)
library(plotly)
(tot_df <- structure(list(
  rank = structure(c(
    1L, 1L, 1L, 1L, 2L, 2L, 2L,
    2L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 4L
  ), levels = c(
    "1", "2", "3",
    "4"
  ), class = "factor"), value = c(
    1, 0.316227766016838, 2, 0.447213595499958,
    3, 0.547722557505166, 4, 0.632455532033676, 5, 0.707106781186548,
    3, 0.547722557505166, 2, 0.447213595499958, 1, 0.316227766016838
  ), name = c(
    "stat1", "stat_2", "stat1", "stat_2", "stat1", "stat_2",
    "stat1", "stat_2", "stat1", "stat_2", "stat1", "stat_2", "stat1",
    "stat_2", "stat1", "stat_2"
  ), `Display name` = structure(c(
    1L,
    1L, 2L, 2L, 1L, 1L, 2L, 2L, 1L, 1L, 2L, 2L, 1L, 1L, 2L, 2L
  ), levels = c(
    "Reference",
    "Current"
  ), class = "factor"), series_name_text = structure(c(
    1L,
    2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L
  ), levels = c(
    "Reference stat1",
    "Reference stat_2", "Current stat1", "Current stat_2"
  ), class = "factor"),
  my_legend_rank = c(
    1L, 2L, 3L, 4L, 1L, 2L, 3L, 4L, 1L, 2L,
    3L, 4L, 1L, 2L, 3L, 4L
  )
), row.names = c(NA, -16L), class = c(
  "tbl_df",
  "tbl", "data.frame"
))
)


px <- plot_ly(
  data = tot_df,
  width = 800,
  height = 600,
  x = ~rank,
  y = ~value,
  name = ~series_name_text,
  legendrank = ~my_legend_rank,
  color = ~`Display name`,
  colors = c("green", "blue"),
  textposition = "none",
  type = "bar",
  opacity = .5,
  yaxis = "y"
) |> layout(legend = list(
  orientation = "h",
  xanchor = "center",
  yanchor = "bottom",
  y = -0.3,
  x = 0.5
))

# table(tot_df$series_name_text,tot_df$my_legend_rank
#                  1 2 3 4
# Reference stat1  4 0 0 0
# Reference stat_2 0 4 0 0
# Current stat1    0 0 4 0
# Current stat_2   0 0 0 4

plotly_json(px)

It seems to me that based on legendrank 'Reference stat1 should be the first i.e. leftmost trace name observed within the legend box, and this isnt the case, the first seen is Current stat1, that should be in 3rd position...)
Is there a gotcha here ? some incorrect setup ? something that invalidates the use of legendrank ?