Added feature to separately color the middle line of crossbar and box…

[capebulbs]

…plot

To realize the minimalist style of boxplot like the one inside
http://biochemres.com/beautiful-minimalist-boxplots-with-r-and-ggplot2
Instead of adding extra geom_segment over geom_boxplot, now the color of middle bar can be colored separately by setting middle.colo(u)r. And this feature can be turned off by setting separate.middle = FALSE, the middle bar will be colored with the outline of boxplot/crossbar by the col parameter, which acts as what ggplot2 does currently

[capebulbs]

sample code:
# also supported in legends
ggplot(mpg, aes(displ, hwy, col = drv, fill = drv)) + geom_boxplot(middle.color = "white")

# turn off this feature by setting separate.middle = FALSE
ggplot(mpg, aes(displ, hwy, col = drv)) + geom_boxplot(middle.color = "white", separate.middle = FALSE)

[clauswilke]

Please take a look at this comment and the surrounding conversation:
#3485 (comment)

[capebulbs]

Please take a look at this comment and the surrounding conversation:
#3485 (comment)

I've read the #3485, and I think my pull request is not like it, I just want to make the plot look better, because in the past for making the boxplot/crossbar have the same fill and color which, in my point of view, looks more beautiful, the middle bar will also be the same color and cannot be seen anymore, I have to add an extra geom_segment to plot another middle bar, which is not an elegant approach, in my PR #3505, this feature has already been integrated into ggplot2 itself, and the three newly-added arguments are not mapped in aes, so it does note affect the mapping of data, just for a better look.

[yutannihilation]

I think a geom can have colour-related parameters outside of the hierarchy of aesthetics, as GeomBoxplot already has outlier.colour. But... I rather feel your middle.colour is not the case; it should be in the hierarchy, if it is to be implemented.

[clauswilke]

In my opinion, moving color parameters outside of the hierarchy of aesthetics is almost never warranted. outlier.colour was added a long time ago, before we fully understood how to implement hierarchical aesthetics. Now that we know how it works and that it works fine, it's unclear that keeping it as a parameter only is the right choice.

[yutannihilation]

Ah, agreed. So, do you mean this PR should consider moving outlier.colour into the hierarchy first, and then implement middle.colour in the hierarchy?

[clauswilke]

That should at least be under consideration, yes.

[capebulbs]

OK, I'll try to put it into the hierarchy of aesthetics, thank you for your suggestions!

[clauswilke]

Closing this PR.

@capebulbs: If you still plan to implement this feature along the lines outlined in my comments, please open a new PR.