我一直在绞尽脑汁地思考一个问题，那就是在geom_col图上向外对齐geom_text标签.我做了一个重现我问题的MWE.问题是，我在正负两个区域都有列，and一些负值比正值大得多，把我轴的"0"标记移到了一边.

以下是常规设置(请注意，我使用的是Windows 11，使用的是R版本4.2.2)和数据:

library(ggplot2)
library(dplyr)
library(tidyr)
ggplot2::theme_set(theme_minimal())

df <- tibble(period = c("one_y", "three_y", "five_y", "seven_y", "ten_y", "whole_period"),
             nominal_gross = c(-21.13, NA, NA, 3.04, 4.1, 4),
             nominal_net = c(-21.64, -3.48, 0.6, 2.58, 3.69, 3.7),
             real_net = c(-25.65, -8.97, -3.75, -0.89, 1.04, 1.25)) %>%
  pivot_longer(!period, names_to = "return_type", values_to = "return_values")

我曾try 使用hjust="outward"，但据我所知，该命令不以轴上的"0"为参考，而是以绘图区的中点(在我的例子中约为-10)为参考.下面是代码，我插入下面的结果图:

ggplot(data = df, 
       mapping = aes(
         x = factor(period, 
                  levels = rev(c("one_y", "three_y", "five_y", "seven_y", "ten_y", "whole_period")),
                  labels = rev(c("1 year", "3 years", "5 years", "7 years", "10 years", "Whole period"))),
         y = return_values, 
         label = scales::label_percent(scale = 1)(return_values),
         fill = return_type)) +
  geom_col(position = position_dodge2(reverse = T), na.rm = T) +
  geom_text(position = position_dodge2(reverse = T, width = 0.9),
            na.rm = T,
            hjust = "outward", 
            vjust = 0.4,
            size = 3,
            check_overlap = T) +
  scale_fill_manual(values = c("#006FB4", "darkgrey", "#01416B"), 
                    labels = c("Nominal gross", "Nominal net", "Real net"),
                    name = NULL) +   
  scale_y_continuous(n.breaks = 10,
                     expand = expansion(add = ifelse(max(abs(df$return_values), na.rm = T)> 10, 3.5, 2)),
                     labels = scales::label_percent(scale = 1)) +
  labs(x = NULL, y = "Annualised returns") +
    coord_flip()

结果没那么差，but我的负值在0到-10之间的标签在条里面…… 然后我试着用ifelse把hjust变成return_values的条件.同样，以下是代码和生成的图表:

ggplot(data = df, 
       mapping = aes(
         x = factor(period, 
                    levels = rev(c("one_y", "three_y", "five_y", "seven_y", "ten_y", "whole_period")),
                    labels = rev(c("1 year", "3 years", "5 years", "7 years", "10 years", "Whole period"))),
         y = return_values, 
         label = scales::label_percent(scale = 1)(return_values),
         fill = return_type)) +
  geom_col(position = position_dodge2(reverse = T)) +
  geom_text(position = position_dodge2(reverse = T, width = 0.9),
            hjust = ifelse(test = df$return_values < 0, yes = 1.25, no = -0.25),
            vjust = 0.4,
            size = 3,
            check_overlap = T) +
  scale_fill_manual(values = c("#006FB4", "darkgrey", "#01416B"), 
                    labels = c("Nominal gross", "Nominal net", "Real net"),
                    name = NULL) +   
  scale_y_continuous(n.breaks = 10,
                     expand = expansion(add = ifelse(max(abs(df$return_values), na.rm = T)> 10, 3.5, 2)),
                     labels = scales::label_percent(scale = 1)) +
  labs(x = NULL, y = "Annualised returns") +
  coord_flip()

这实际上更糟糕:我的大部分标签现在都在wine 吧里，但有些放在了正确的位置.然后，我得出结论，我的条件语句存在问题.我曾try 删除Nas，以用零替换Nas，但我找不到问题的根源.如果有人能帮我找到解决方案，我将不胜感激.

请注意，如果有一种方法可以通过"打断"轴并删除-10到-20之间的区域来对绘图进行"裁剪"，我也希望如此(下图中两条黑线之间的区域).我try 过将限制设置在scale_y_continuous以内，并使用了oob参数，但在我try 过的所有解决方案中，最长的条形图似乎就停在了限制值上，并且我失go 了三者之间的渐变.

EDIT

实际上，我刚刚找到了答案:条件语句必须传递给geom_text，并带有hust作为美学:

geom_text(aes(hjust = ifelse(return_values < 0, 1.25, -0.25)),
            position = position_dodge2(reverse = T, width = 0.9),
            vjust = 0.4,
            size = 3,
            check_overlap = T) 

那么它就运行得很好.