Allow sort with facets in barplots visualizer

[ColinConwell]

A nice addition to the radiant.data visualizer would be to add the option to sort barplots even with the addition of facets.

I've written a preliminary version of this that's working decently, using tidytext's reorder_within() function.

Beginning at line 510 in radiant.data's visualize.R:

if ("sort" %in% axes && facet_row == "." && facet_col == ".") {
  if ("flip" %in% axes) {
    tmp <- arrange_at(ungroup(tmp), .vars = j)
  } else {
    tmp <- arrange_at(ungroup(tmp), .vars = j, .funs = desc)
  }
  tmp[[i]] %<>% factor(., levels = unique(.))
}

#modified code beginning here...
plot_list[[itt]] <- ggplot(tmp, aes_string(x = i, y = j))

if ("sort" %in% axes && (facet_row != "." || facet_col != ".")) {
  if (facet_row != '.' && facet_col != '.') sorting_vars <- sprintf('list(%s, %s)', facet_row, facet_col)
  if (facet_row != '.' && facet_col == '.') sorting_vars <- sprintf('list(%s)', facet_row)
  if (facet_row == '.' && facet_col != '.') sorting_vars <- sprintf('list(%s)', facet_col)

  i_sorted <- sprintf('tidytext::reorder_within(%s, %s, %s)', i, j, sorting_vars)
  plot_list[[itt]] <- ggplot(tmp, aes_string(x = i_sorted, y = j)) +
    tidytext::scale_x_reordered() + xlab(i)
}

plot_list[[itt]] <- plot_list[[itt]] + {
  if (fill == "none") {
    geom_bar(stat = "identity", position = "dodge", alpha = alpha, fill = fillcol)
  } else {
    geom_bar(stat = "identity", position = "dodge", alpha = alpha)
  }
}

not sure what the easiest way to incorporate this might be, but hope it's possible!

this would also be complimented, I think, by the addition of a 'Scale-x' option in visualize_ui.R, since this frees the scale and makes the plot much easier to digest.

[vnijs]

Interesting suggestion @ColinConwell. Can you share an simple R example and the plot that you get out? I'm interested in this option but would prefer to avoid adding a dependency on tidytext

[ColinConwell]

Sure. Reprex below. (I use aes_string here only to maintain consistency with the code in your visualize function).

data <- get('mtcars') %>% rownames_to_column('model')

ggplot(data, aes_string('tidytext::reorder_within(model, mpg, list(cyl, vs))', 'mpg')) +
  facet_grid(as.formula('cyl~vs'), scales = 'free_y') +
  tidytext::scale_x_reordered() +
  geom_bar(stat = 'identity', position = 'identity') + coord_flip()

The plot that emerges looks like this:

There are of course other ways of doing this, but they tend to involve creating rankings and then swapping the ranks with the names of the discrete axis labels. If that sounds preferable, I could post an example of that method, as well?