好的，在艾伦的鼓励下，我自己画这些东西并没有那么糟糕，我决定也尝试解决这个问题。它正在做我试图避免这个问题的事情，但它可能对你们其他人有所帮助。
我采用了稍微不同的方法，主要区别在于 (1) 我们保留折线而不是转换为多边形，以及 (2) 我对三角函数不太满意，因此我使用 approxfun() 来插入线和(3) 我们将使用绝对单位而不是相对单位，因此在调整设备大小时不会很尴尬。
首先，当我打算在自定义 geom 函数中使用它时，我的目标是制作一个 grob 结构，以便在 geom 的绘制方法的末尾轻松粘贴。你可以给它一个 grob，或者一个 grob 的参数。它更改了 grob 的类（稍后将变得相关），删除线型参数并添加破折号和中断的信息。
library(grid)
library(scales)
linetypeGrob <- function(x, ..., dashes = 1, breaks = 1) {
if (!inherits(x, "polyline")) {
x <- polylineGrob(x, ...)
}
class(x)[[1]] <- "linetypeGrob"
x$gp$lty <- NULL
x$dashes <- dashes
x$breaks <- breaks
x
}
现在正如我上面提到的，我们将回到课堂。自定义 grob 类的巧妙之处在于，您可以在它们被绘制之前拦截它们，以便您可以在最后一刻进行更改。为此，我们为网格中的makeContext 函数编写了一个 S3 方法，以进行相关更改。我知道这是一个很长的函数，但我试图通过插入告诉我要做什么的 cmets 使其更容易理解。
makeContext.linetypeGrob <- function(x) {
# Sort out line IDs
id <- x$id
if (is.null(id)) {
if (is.null(x$id.lengths)) {
id <- rep(1L, length(x$x))
} else {
id <- rep(seq_along(x$id.lengths), x$id.lengths)
}
}
# Delete previous line IDs
x$id <- NULL
x$id.lengths <- NULL
# Take dashes and breaks parameters out of the old grob
dashes <- x$dashes
x$dashes <- NULL
breaks <- x$breaks
x$breaks <- NULL
# Convert to absolute units
newx <- convertX(x$x, "mm", TRUE)
newy <- convertY(x$y, "mm", TRUE)
# Express lines as points along a cumulative distances
dist <- sqrt(diff(newx)^2 + diff(newy)^2)
cumdist <- cumsum(c(0, dist))
# Take new lines as a sequence along the cumulative distance
starts <- seq(0, max(cumdist), by = (dashes + breaks))
ends <- seq(dashes, max(cumdist), by = (dashes + breaks))
if (length(ends) == length(starts) - 1) {
# Case when the end actually should have gone beyond `max(cumdist)`
ends <- c(ends, max(cumdist))
}
# Set index for graphical parameters
gp_i <- findInterval(starts, cumdist[cumsum(rle(id)$lengths)]) + 1
# Basically dealing with elbow pieces a bit
# Find mismatches between the original segments that starts and ends fall on
start_id <- findInterval(starts, cumdist)
end_id <- findInterval(ends, cumdist)
mismatch <- which(start_id != end_id)
# Insert elbow pieces
starts <- c(starts, cumdist[end_id[mismatch]])
starts <- starts[{o <- order(starts)}] # Need the order for later
ends <- sort(c(ends, cumdist[end_id[mismatch]]))
# Join elbow pieces
new_id <- seq_along(start_id)
if (length(mismatch)) {
i <- rep_len(1, length(new_id))
i[mismatch] <- 2
new_id <- rep(new_id, i)
}
# Seperate lines with different IDs
keepfun <- approxfun(cumdist, id)
keep <- (keepfun(starts) %% 1) == 0 & (keepfun(ends) %% 1) == 0
# Interpolate x
xfun <- approxfun(cumdist, newx)
x0 <- xfun(starts[keep])
x1 <- xfun(ends[keep])
# Interpolate y
yfun <- approxfun(cumdist, newy)
y0 <- yfun(starts[keep])
y1 <- yfun(ends[keep])
# Expand graphic parameters by new ID
x$gp[] <- lapply(x$gp, function(x){
if (length(x) == 1) {
return(x)
} else {
x[as.integer(gp_i)]
}
})
# Put everything back into the grob
x$x <- unit(as.vector(rbind(x0, x1)), "mm")
x$y <- unit(as.vector(rbind(y0, y1)), "mm")
x$id <- as.vector(rbind(new_id[keep], new_id[keep]))
class(x)[[1]] <- "polyline"
x
}
最后，为了证明它有效，我将使用这个新的 grob 绘制一些虚拟数据。您可以像绘制普通折线 grob 一样使用它。
set.seed(100)
x <- c(cumsum(rnorm(10)), cumsum(rnorm(10)))
y <- c(cumsum(rnorm(10)), cumsum(rnorm(10)))
id <- rep(c(1, 2), each = 10)
gp <- gpar(lwd = c(2, 10), lineend = "butt",
col = c("magenta", "blue"))
grob <- linetypeGrob(scales::rescale(x),
scales::rescale(y),
id = id, gp = gp, dashes = 5, breaks = 2)
grid.newpage(); grid.draw(grob)
如果我调整设备大小，您可以看到破折号和中断的长度保持不变：