Handle fine-grained show.legend properly

[yutannihilation]

Fix #3461

Currently, guide_geom.colorbar() doesn't follow the way how guide_geom.legend() handles the case when show.legend is a named logical vector. This PR fixes it.

Besides, this PR also fixes guide_geom.legend() to handle multiple logicals, which I should have fixed #3163.

library(dplyr, warn.conflicts = FALSE)

devtools::load_all("~/repo/R/ggplot2/")

tribble(
  ~x1, ~y1, ~x2, ~y2, ~id, ~id_int,
  1,   1,   2,   2, "A",       1,
  2,   2,   3,   3, "B",       2
) %>%
  ggplot(aes(xmin = x1, ymin = y1, xmax = x2, ymax = y2, fill = id_int)) +
  geom_rect(show.legend = c(size = FALSE, fill = TRUE))

[yutannihilation]

Travis fails on build R<=3.4, where isTRUE() was not tolerant for named TRUE...

# R 3.4.4 on RStudio Cloud

x <- c(a = TRUE, b = FALSE)
is.na(x["no_such_element"])
#> <NA> 
#> TRUE
isTRUE(is.na(x["no_such_element"]))
#> [1] FALSE

^{Created on 2019-08-18 by the reprex package (v0.2.1)}

[thomasp85]

Why does an empty show_legend lead to inclusion here?

[yutannihilation]

iiuc, guide_geom() shows a legend if

1. the layer's mapping contains any of the corresponding aes and
2. the user does not explicitly choose not to show the legend

So, the absence of the corresponding show.legend should be translated as inclusion.

Does this answer your question? Sorry, I might not get your point....

[thomasp85]

You are right - I'm probably just tired

[yutannihilation]

Thanks. Sorry for making you busy with reviewing pull requests :)

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/