如何在R的grid包中获取不规则polygon grob的准确外尺寸？

我尝试使用R语言grid包的grobX()和grobY()函数，为polygonGrob()绘制的多边形添加边界框。对于矩形这类规则形状，能正确获取外尺寸，但不规则多边形的结果存在偏差。

原始代码与问题复现 #

首先定义多边形数据和grob对象：

x <- c(0.65, 0.614906666467847, 0.42604722665004, 0.425, 0.359046106882114,
       0.259046106882114, 0.425, 0.526047226650039, 0.614906666467847)
y <- c(0.5, 0.692836282905962, 0.895442325903662, 0.759807621135332,
       0.602606042997701, 0.397393957002299, 0.240192378864668, 0.204557674096338,
       0.307163717094038)

library(grid)

pg <- polygonGrob(x=x, y=y,
                  default.units = "native",
                  gp=gpar(fill="gray"))
pntg <- pointsGrob(x=x, y=y,
                   pch = 20,
                   size = grid::unit(3, "mm"),
                   default.units = "native",
                   gp=gpar(col="red"))

rcg <- rectGrob(x=0.33, height=0.7, width=0.2,
                gp=gpar(fill="gray"))

不规则多边形的错误边界框

使用grobX()/grobY()绘制边界框时，结果无法准确覆盖多边形：

pushViewport(plotViewport(c(5, 4, 2, 2)))
pushViewport(dataViewport(x, y))
grid.rect()
grid.xaxis()
grid.yaxis()

grid.draw(pg)
grid.draw(pntg)

x1 <- grid::convertWidth(grid::grobX(pg, "west"), "npc", TRUE)
x2 <- grid::convertWidth(grid::grobX(pg, "east"), "npc", TRUE)
y1 <- grid::convertHeight(grid::grobY(pg, "south"), "npc", TRUE)
y2 <- grid::convertHeight(grid::grobY(pg, "north"), "npc", TRUE)

grid::grid.polyline(x = c(x1, x1, x2, x2, x1),
                    y = c(y1, y2, y2, y1, y1),
                    default.units = "npc",
                    gp = gpar(col = "blue"))

矩形的正确边界框

同样方法对矩形rectGrob()有效：

pushViewport(plotViewport(c(5, 4, 2, 2)))
pushViewport(dataViewport(x, y))
grid.rect()
grid.xaxis()
grid.yaxis()

grid.draw(rcg)

x1 <- grid::convertWidth(grid::grobX(rcg, "west"), "npc", TRUE)
x2 <- grid::convertWidth(grid::grobX(rcg, "east"), "npc", TRUE)
y1 <- grid::convertHeight(grid::grobY(rcg, "south"), "npc", TRUE)
y2 <- grid::convertHeight(grid::grobY(rcg, "north"), "npc", TRUE)

grid::grid.polyline(x = c(x1, x1, x2, x2, x1),
                    y = c(y1, y2, y2, y1, y1),
                    gp = gpar(col = "red"))

解决方法 #

grobX()/grobY()对polygonGrob()使用"west"/"east"/"north"/"south"参数时，并非返回多边形顶点的实际极值，而是基于grob的对齐参考点计算，因此会出现偏差。正确做法是直接提取多边形顶点坐标，计算极值后转换单位绘制边界框：

pushViewport(plotViewport(c(5, 4, 2, 2)))
pushViewport(dataViewport(x, y))
grid.rect()
grid.xaxis()
grid.yaxis()

grid.draw(pg)
grid.draw(pntg)

# 提取多边形顶点的native单位数值
pg_x <- convertUnit(pg$x, "native", valueOnly = TRUE)
pg_y <- convertUnit(pg$y, "native", valueOnly = TRUE)

# 计算坐标极值
x_min <- min(pg_x)
x_max <- max(pg_x)
y_min <- min(pg_y)
y_max <- max(pg_y)

# 将极值转换为npc单位
x1 <- convertX(unit(x_min, "native"), "npc", valueOnly = TRUE)
x2 <- convertX(unit(x_max, "native"), "npc", valueOnly = TRUE)
y1 <- convertY(unit(y_min, "native"), "npc", valueOnly = TRUE)
y2 <- convertY(unit(y_max, "native"), "npc", valueOnly = TRUE)

# 绘制准确边界框
grid.polyline(x = c(x1, x1, x2, x2, x1),
              y = c(y1, y2, y2, y1, y1),
              default.units = "npc",
              gp = gpar(col = "blue"))

这种方法直接基于多边形原始顶点计算边界框，能准确覆盖所有顶点，解决grobX()/grobY()的偏差问题。

内容的提问来源于stack exchange，提问作者Crops