如何将geom_rect扩展至geom_tile的d1与d6边缘？

使用geom_tile绘制热图时，希望用红色矩形突出特定日期（2018/3/5）的整个深度区域，但当前绘制的矩形无法延伸至d6和d1的边缘，效果如下：

原代码：

df <- data.frame(
  sensor_date = rep(seq(from = as.Date("2018/3/4"), to = as.Date("2018/3/6"), "days"), each = 6),
  depth_class = as.factor(rep(c("d1", "d2", "d3", "d4", "d5", "d6"), 3)),
  freq = rnorm(18, 10, 4)
)


df_rect <- data.frame(
  sensor_date = as.Date(c(rep("2018/3/5", 6)), format = "%Y/%m/%d"),
  depth_class = c("d0")
)


ggplot(data = df) +
  geom_tile(aes(x = sensor_date, y = depth_class, fill = freq)) +
  geom_rect(
    data = df_rect, linewidth = 1, colour = "red", fill = NA, alpha = 0.2,
    aes(
      xmin = min(sensor_date - .5),
      xmax = max(sensor_date + .5),
      ymin = min("d6"),
      ymax = max("d1")
    )
  )

当前效果：

解决方案 #

问题核心是对因子型y轴的数值映射理解：ggplot会把depth_class这类因子变量转换为连续数值（比如d1对应1、d2对应2…d6对应6），每个geom_tile的y轴实际范围是[数值-0.5, 数值+0.5]。要让矩形延伸至边缘，需要把矩形的y轴范围扩展到所有深度因子对应的数值范围之外各加减0.5。

修改后的代码：

set.seed(123) # 设置随机种子保证结果可复现
df <- data.frame(
  sensor_date = rep(seq(from = as.Date("2018/3/4"), to = as.Date("2018/3/6"), "days"), each = 6),
  depth_class = as.factor(rep(c("d1", "d2", "d3", "d4", "d5", "d6"), 3)),
  freq = rnorm(18, 10, 4)
)

# 简化矩形数据，仅保留目标日期
df_rect <- data.frame(
  sensor_date = as.Date("2018/3/5", format = "%Y/%m/%d")
)

ggplot(data = df) +
  geom_tile(aes(x = sensor_date, y = depth_class, fill = freq)) +
  geom_rect(
    data = df_rect, linewidth = 1, colour = "red", fill = NA, alpha = 0.2,
    aes(
      xmin = sensor_date - 0.5,
      xmax = sensor_date + 0.5,
      # 将因子转换为数值，扩展范围至边缘
      ymin = min(as.numeric(df$depth_class)) - 0.5,
      ymax = max(as.numeric(df$depth_class)) + 0.5
    )
  )

关键改动说明： #

• 简化df_rect：仅需一行目标日期数据，无需重复多次
• 调整y轴范围：通过as.numeric()将因子转换为底层连续数值，再分别减/加0.5，让矩形覆盖从d6底部到d1顶部的完整边缘区域
• 简化x轴范围：单日期场景下，直接用sensor_date ± 0.5即可覆盖该日期的整个热图宽度

内容的提问来源于stack exchange，提问作者Dasr