为何ggplot的scale_manual扩展图例项在ggplotly中无法正常显示？

问题场景 #

在ggplot中使用scale_color_manual的limits参数扩展图例，加入数据中不存在的取值（比如示例中的"10"），可以正常显示包含"ten"的完整图例，但用ggplotly()渲染后，新增的图例项会丢失。

原测试代码：

lims <- c("4", "6", "8", "10")
cols <- c("4" = "blue", "6" = "darkgreen", "8" = "red", "10" = "orange")
labs <- c("4" = "four", "6" = "six", "8" = "eight", "10" = "ten")

p <- ggplot(mtcars, aes(mpg, wt)) +
  geom_point(aes(color = factor(cyl)))

q <- p + 
  scale_color_manual(name = "Cylinder", values = cols, labels = labs, limits = lims)

# ggplot中正常显示完整图例
q

# ggplotly渲染后丢失"ten"图例项
ggplotly(q)

解决方法 #

方法1：添加geom_blank层保留缺失水平 #

ggplotly会过滤掉无对应数据点的图例类别，通过添加geom_blank（空数据点）让它识别到该类别的存在，从而保留图例项。修改后的代码如下：

lims <- c("4", "6", "8", "10")
cols <- c("4" = "blue", "6" = "darkgreen", "8" = "red", "10" = "orange")
labs <- c("4" = "four", "6" = "six", "8" = "eight", "10" = "ten")

# 将cyl转换为包含所有目标水平的因子
mtcars$cyl_factor <- factor(mtcars$cyl, levels = lims)

p <- ggplot(mtcars, aes(mpg, wt)) +
  geom_point(aes(color = cyl_factor)) +
  # 添加空数据点对应缺失的"10"水平
  geom_blank(aes(color = "10")) +
  scale_color_manual(name = "Cylinder", values = cols, labels = labs, limits = lims)

# ggplot和ggplotly均能显示完整图例
p
ggplotly(p)

方法2：手动修改plotly对象 #

如果不想改动ggplot代码，可以在生成ggplotly对象后，手动插入缺失的图例项：

lims <- c("4", "6", "8", "10")
cols <- c("4" = "blue", "6" = "darkgreen", "8" = "red", "10" = "orange")
labs <- c("4" = "four", "6" = "six", "8" = "eight", "10" = "ten")

p <- ggplot(mtcars, aes(mpg, wt)) +
  geom_point(aes(color = factor(cyl)))

q <- p + 
  scale_color_manual(name = "Cylinder", values = cols, labels = labs, limits = lims)

# 生成ggplotly对象
ply <- ggplotly(q)

# 手动添加"ten"图例项
ply$x$data <- c(ply$x$data, list(
  list(
    type = "scatter",
    mode = "markers",
    marker = list(color = cols["10"]),
    name = labs["10"],
    showlegend = TRUE,
    x = numeric(0),
    y = numeric(0)
  )
))

# 渲染修改后的plotly对象
ply

原理说明 #

ggplotly默认会自动过滤掉没有对应数据点的图例类别，通过geom_blank添加空数据点，或者手动插入空的散点图层，可以让plotly识别到该类别的存在，从而保留对应的图例项。

内容的提问来源于stack exchange，提问作者BTT