如何將帶有批注（annotation_custom）（）的grob放置在繪圖區域的精確區域？

Question

我試圖用ggplot2重現以下[base R]圖

我已經完成了大部分工作，但目前困擾我的是線段的位置，這些線段以相應的標簽連接到位於圖右側的邊緣地毯圖。 標簽是通過anotation_custom()繪制的（在下面的第二張圖中），我使用了anotation_custom()的技巧，即關閉裁剪以允許在繪圖邊距中繪制。

盡管進行了許多嘗試，但我仍無法將segmentGrobs()放置在設備中的所需位置，以使它們加入正確的地毯滴答聲和標簽。

一個可重現的例子是

y <- data.frame(matrix(runif(30*10), ncol = 10))
names(y) <- paste0("spp", 1:10)
treat <- gl(3, 10)
time <- factor(rep(1:10, 3))

require(vegan); require(grid); require(reshape2); require(ggplot2)
mod <- prc(y, treat, time)

如果您沒有安裝素食主義者 ，則在Question的末尾附加要強化對象的dput和要使用ggplot()進行方便繪圖的fortify()方法（如果您希望運行示例和fortify() ggplot() 。 我還包括一個有點冗長的函數myPlt() ，它說明了我到目前為止的工作，如果您已加載軟件包並可以創建mod ，則可以在示例數據集上使用mod 。

我已經嘗試了很多選擇，但是在正確放置線段時，我似乎陷入了困境。

我不是在尋找為示例數據集繪制標簽/段的特定問題的解決方案，而是可以通過編程方式放置段和標簽的通用解決方案，因為這將構成autoplot()方法的基礎class(mod)對象。 我已經解決了標簽問題，但不是線段。 因此，對於以下問題：

1. 當我要放置包含從數據坐標x0 ， y0到x1 ， y1的行的段grob時，如何使用xmin ， xmax ， ymin ， ymax參數？
2. 也許以不同的方式問到，如何使用annotation_custom()在已知數據坐標x0 ， y0到x1 ， y1之間的繪圖區域之外繪制段？

我很高興收到答案，該答案只是在繪圖區域中有任何舊繪圖，但展示了如何在繪圖邊緣的已知坐標之間添加線段。

我不願意使用annotation_custom()所以如果有更好的解決方案，我也會考慮。 我確實想避免在繪圖區域中放置標簽； 我想我可以通過使用annotate()並通過expand參數擴展x軸限制來實現這一點。

fortify()方法

fortify.prc <- function(model, data, scaling = 3, axis = 1,
                        ...) {
    s <- summary(model, scaling = scaling, axis = axis)
    b <- t(coef(s))
    rs <- rownames(b)
    cs <- colnames(b)
    res <- melt(b)
    names(res) <- c("Time", "Treatment", "Response")
    n <- length(s$sp)
    sampLab <- paste(res$Treatment, res$Time, sep = "-")
    res <- rbind(res, cbind(Time = rep(NA, n),
                            Treatment = rep(NA, n),
                            Response = s$sp))
    res$Score <- factor(c(rep("Sample", prod(dim(b))),
                          rep("Species", n)))
    res$Label <- c(sampLab, names(s$sp))
    res
}

dput()

這是fortify.prc(mod)的輸出：

structure(list(Time = c(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 
3, 4, 5, 6, 7, 8, 9, 10, NA, NA, NA, NA, NA, NA, NA, NA, NA, 
NA), Treatment = c(2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 
3, 3, 3, 3, 3, 3, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA), Response = c(0.775222658013234, 
-0.0374860102875694, 0.100620532505619, 0.17475403767196, -0.736181209242918, 
1.18581913245908, -0.235457236665258, -0.494834646295896, -0.22096700738071, 
-0.00852429328460645, 0.102286976108412, -0.116035743892094, 
0.01054849999509, 0.429857364190398, -0.29619258318138, 0.394303081010858, 
-0.456401545475929, 0.391960511587087, -0.218177702859661, -0.174814586471715, 
0.424769871360028, -0.0771395073436865, 0.698662414019584, 0.695676522106077, 
-0.31659375422071, -0.584947748238806, -0.523065304477453, -0.19259357510277, 
-0.0786143714402391, -0.313283220381509), Score = structure(c(1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L), .Label = c("Sample", 
"Species"), class = "factor"), Label = c("2-1", "2-2", "2-3", 
"2-4", "2-5", "2-6", "2-7", "2-8", "2-9", "2-10", "3-1", "3-2", 
"3-3", "3-4", "3-5", "3-6", "3-7", "3-8", "3-9", "3-10", "spp1", 
"spp2", "spp3", "spp4", "spp5", "spp6", "spp7", "spp8", "spp9", 
"spp10")), .Names = c("Time", "Treatment", "Response", "Score", 
"Label"), row.names = c("1", "2", "3", "4", "5", "6", "7", "8", 
"9", "10", "11", "12", "13", "14", "15", "16", "17", "18", "19", 
"20", "spp1", "spp2", "spp3", "spp4", "spp5", "spp6", "spp7", 
"spp8", "spp9", "spp10"), class = "data.frame")

我嘗試過的

myPlt <- function(x, air = 1.1) {
    ## fortify PRC model
    fx <- fortify(x)
    ## samples and species scores
    sampScr <- fx[fx$Score == "Sample", ]
    sppScr <- fx[fx$Score != "Sample", ]
    ord <- order(sppScr$Response)
    sppScr <- sppScr[ord, ]
    ## base plot
    plt <- ggplot(data = sampScr,
                  aes(x = Time, y = Response,
                      colour = Treatment, group = Treatment),
                  subset = Score == "Sample")
    plt <- plt + geom_line() + # add lines
            geom_rug(sides = "r", data = sppScr) ## add rug
    ## species labels
    sppLab <- sppScr[, "Label"]
    ## label grobs
    tg <- lapply(sppLab, textGrob, just = "left")
    ## label grob widths
    wd <- sapply(tg, function(x) convertWidth(grobWidth(x), "cm",
                                              valueOnly = TRUE))
    mwd <- max(wd) ## largest label

    ## add some space to the margin, move legend etc
    plt <- plt +
        theme(plot.margin = unit(c(0, mwd + 1, 0, 0), "cm"),
              legend.position = "top",
              legend.direction = "horizontal",
              legend.key.width = unit(0.1, "npc"))
    ## annotate locations
    ##  - Xloc = new x coord for label
    ##  - Xloc2 = location at edge of plot region where rug ticks met plot box
    Xloc <- max(fx$Time, na.rm = TRUE) +
        (2 * (0.04 * diff(range(fx$Time, na.rm = TRUE))))
    Xloc2 <- max(fx$Time, na.rm = TRUE) +
        (0.04 * diff(range(fx$Time, na.rm = TRUE)))
    ## Yloc - where to position the labels in y coordinates
    yran <- max(sampScr$Response, na.rm = TRUE) -
        min(sampScr$Response, na.rm = TRUE)
    ## This is taken from vegan:::linestack
    ## attempting to space the labels out in the y-axis direction
    ht <- 2 * (air * (sapply(sppLab,
                        function(x) convertHeight(stringHeight(x),
                                                  "npc", valueOnly = TRUE)) *
                 yran))
    n <- length(sppLab)
    pos <- numeric(n)
    mid <- (n + 1) %/% 2
    pos[mid] <- sppScr$Response[mid]
    if (n > 1) {
        for (i in (mid + 1):n) {
            pos[i] <- max(sppScr$Response[i], pos[i - 1] + ht[i])
        }
    }
    if (n > 2) {
        for (i in (mid - 1):1) {
            pos[i] <- min(sppScr$Response[i], pos[i + 1] - ht[i])
        }
    }
    ## pos now contains the y-axis locations for the labels, spread out

    ## Loop over label and add textGrob and segmentsGrob for each one
    for (i in seq_along(wd)) {
        plt <- plt + annotation_custom(tg[[i]],
                                       xmin = Xloc,
                                       xmax = Xloc,
                                       ymin = pos[i],
                                       ymax = pos[i])
        seg <- segmentsGrob(Xloc2, pos[i], Xloc, pos[i])

        ## here is problem - what to use for ymin, ymax, xmin, xmax??
        plt <- plt + annotation_custom(seg,
                                       ## xmin = Xloc2,
                                       ## xmax = Xloc,
                                       ## ymin = pos[i],
                                       ## ymax = pos[i])
                                       xmin = Xloc2,
                                       xmax = Xloc,
                                       ymin = min(pos[i], sppScr$Response[i]),
                                       ymax = max(pos[i], sppScr$Response[i]))
    }
    ## Build the plot
    p2 <- ggplot_gtable(ggplot_build(plt))
    ## turn off clipping
    p2$layout$clip[p2$layout$name=="panel"] <- "off"
    ## draw plot
    grid.draw(p2)
}

基於我在myPlt()嘗試過的圖

這是我從上面用myPlt() 。 請注意通過標簽繪制的水平小刻度線-這些應該是上方第一張圖中的斜線段。

Answer 1

這就是我的處理方法，

library(gtable)
library(ggplot2)
library(plyr)

set.seed(1)
d <- data.frame(x=rep(1:10, 5),
                y=rnorm(50),
                g = gl(5,10))

# example plot
p <- ggplot(d, aes(x,y,colour=g)) +
  geom_line() +
  scale_x_continuous(expand=c(0,0))+
  theme(legend.position="top",
        plot.margin=unit(c(1,0,0,0),"line"))

# dummy data for the legend plot
# built with the same y axis (same limits, same expand factor)
d2 <- ddply(d, "g", summarise, x=0, y=y[length(y)])
d2$lab <- paste0("line #", seq_len(nrow(d2)))

plegend <- ggplot(d, aes(x,y, colour=g)) +
  geom_blank() +
  geom_segment(data=d2, aes(x=2, xend=0, y=y, yend=y), 
               arrow=arrow(length=unit(2,"mm"), type="closed")) +
  geom_text(data=d2, aes(x=2.5,label=lab), hjust=0) +
  scale_x_continuous(expand=c(0,0)) +
  guides(colour="none")+
  theme_minimal() + theme(line=element_blank(),
                          text=element_blank(),
                          panel.background=element_rect(fill="grey95", linetype=2))

# extract the panel only, we don't need the rest
gl <- gtable_filter(ggplotGrob(plegend), "panel")

# add a cell next to the main plot panel, and insert gl there
g <- ggplotGrob(p)
index <- subset(g$layout, name == "panel")
g <- gtable_add_cols(g, unit(1, "strwidth", "line # 1") + unit(1, "cm"))
g <- gtable_add_grob(g, gl, t = index$t, l=ncol(g), 
                     b=index$b, r=ncol(g))
grid.newpage()
grid.draw(g)

用特定的標簽和位置修改“傳奇”圖應該很簡單（留給感興趣的讀者練習）。

Answer 2

也許這可以說明annotation_custom ，

myGrob <- grobTree(rectGrob(gp=gpar(fill="red", alpha=0.5)),
                   segmentsGrob(x0=0, x1=1, y0=0, y1=1, default.units="npc"))

myGrob2 <- grobTree(rectGrob(gp=gpar(fill="blue", alpha=0.5)),
                   segmentsGrob(x0=0, x1=1, y0=0, y1=1, default.units="npc"))

p <- qplot(1:10, 1:10) + theme(plot.margin=unit(c(0, 3, 0, 0), "cm")) +
  annotation_custom(myGrob, xmin=5, xmax=6, ymin=3.5, ymax=5.5) +
  annotate("segment", x=5, xend=6, y=3, yend=5, colour="red") +
  annotation_custom(myGrob2, xmin=8, xmax=12, ymin=3.5, ymax=5.5) 

p

g <- ggplotGrob(p)
g$layout$clip[g$layout$name=="panel"] <- "off"
grid.draw(g)

顯然有一個奇怪的錯誤，即如果我重用myGrob而不是myGrob2 ，它將第二次忽略放置坐標，並將其與第一層堆疊在一起。 此功能確實有問題。

Answer 3

這看起來像是使用ggrepel更好的解決方案： https ://stackoverflow.com/a/45631834/4927395