如何在ggplot的expression函数中传入索引向量值生成图例标签？

操作步骤及问题 #

1. 构造示例数据：

# 示例数据
data.frame(out = c(seq(0,1,.1), seq(0,0.5,.05), seq(0,0.1,.01)),
           time = seq(0,50,5),
           group = factor(rep(letters[1:3],each=11))) -> d

2. 统计各组样本数量：

groupCounts <- table(d$group)
# 输出结果：
#  a  b  c 
# 11 11 11

3. 尝试设置图例标签时遇到问题：
   使用scale_colour_manual()传入expression()时，groupCounts的索引值未在图例中显示，仅手动输入数字时正常：

ggplot(data = d,
       mapping = aes(x = time, 
                     y = out,
                     colour = group)) +
  geom_line() +
  scale_colour_manual(labels = c(expression(paste("All (", italic("n"), "=", groupCounts[1], ")")),
                                 expression(paste("0 days (", italic("n"), "=", groupCounts[2], ")")),
                                 expression(paste("1-9 days (", italic("n"), "=", groupCounts[3], ")"))),
                      values = c("#440154",
                                 "#31688E",
                                 "#35B779"))

解决方案 #

原方法失效的原因是expression()属于延迟解析，内部的groupCounts[1]不会被求值，仅作为符号保留。以下两种方法可解决问题：

方法1：使用bquote()嵌入求值后的变量

bquote()允许通过.()语法将变量求值后嵌入表达式：

# 构造带样本量的图例标签
labels <- c(
  bquote(paste("All (", italic("n"), "=", .(groupCounts[1]), ")")),
  bquote(paste("0 days (", italic("n"), "=", .(groupCounts[2]), ")")),
  bquote(paste("1-9 days (", italic("n"), "=", .(groupCounts[3]), ")"))
)

ggplot(data = d,
       mapping = aes(x = time, 
                     y = out,
                     colour = group)) +
  geom_line() +
  scale_colour_manual(labels = labels,
                      values = c("#440154", "#31688E", "#35B779"))

方法2：先构造字符串再解析为表达式

先用sprintf()生成包含数值的字符串，再通过parse(text=)转换为ggplot可识别的表达式：

# 构造带数值的字符串模板，再解析为表达式
label_text <- c(
  sprintf("paste('All (', italic('n'), '=', %d, ')', sep='')", groupCounts[1]),
  sprintf("paste('0 days (', italic('n'), '=', %d, ')', sep='')", groupCounts[2]),
  sprintf("paste('1-9 days (', italic('n'), '=', %d, ')', sep='')", groupCounts[3])
)
labels <- parse(text = label_text)

ggplot(data = d,
       mapping = aes(x = time, 
                     y = out,
                     colour = group)) +
  geom_line() +
  scale_colour_manual(labels = labels,
                      values = c("#440154", "#31688E", "#35B779"))

内容的提问来源于stack exchange，提问作者llewmills