ggplot中scale_color_identity()标签自动按颜色名重排问题

问题描述 #

希望通过label_col列指定的颜色控制字体颜色，但使用scale_color_identity()时，"black"和"white"会按字母顺序自动排序，导致标签位置与geom_col()的堆叠位置不匹配。统一设置label_col为单一颜色时位置正常，已知scale_color_manual()可解决，但想了解scale_color_identity()失效的原因及对应解决方案。

最小复现代码：

list_order <- c("a", "b", "c", "d", "e", "f")

test <- data.frame(indicator = rep(list_order, 2),
                   mean = c(1.1, 0.3, 6.3, 4, 7, 2.1, 1.2, 0.7, 6.9, 3.3, 5.4, 1.8),
                   group = c(rep("Group A", 6), rep("Group B", 6)))

test %>% 
  mutate(indicator = factor(indicator, levels = list_order),
         label = round(mean, 1),
         label_col = ifelse(indicator %in% c("a", "d", "e", "f"), "black", "white")) %>% 
  ggplot(aes(x = group, y = mean, fill = indicator)) +
  geom_col(width = 0.7, position = "stack") +
  geom_text(aes(label = label, color = label_col), size = 3, position = position_stack(vjust = 0.5)) +
  scale_color_identity() +
  scale_fill_viridis(discrete = T, option = "plasma")

失效原因 #

ggplot2中，当在aes()里指定color = label_col时，会自动将label_col视为离散分组变量，默认按字母顺序（black在前，white在后）对数据分组。而position_stack()是按分组的顺序计算堆叠位置的：

• geom_col()的堆叠顺序由indicator的因子水平（list_order）决定；
• geom_text()的堆叠顺序却被label_col的分组顺序覆盖，导致文本的堆叠逻辑与柱子不一致，最终位置错乱。
  scale_color_identity()仅负责将label_col的字符串直接映射为颜色，不会改变分组的排序逻辑，因此无法解决位置问题。

针对scale_color_identity()的解决方案 #

核心思路是让geom_text()的堆叠顺序与geom_col()保持一致，即遵循indicator的因子顺序，而非label_col的分组顺序。有两种实现方式：

方法1：在geom_text的aes中指定group=indicator #

强制geom_text按indicator分组堆叠，忽略color的分组影响：

test %>% 
  mutate(indicator = factor(indicator, levels = list_order),
         label = round(mean, 1),
         label_col = ifelse(indicator %in% c("a", "d", "e", "f"), "black", "white")) %>% 
  ggplot(aes(x = group, y = mean, fill = indicator)) +
  geom_col(width = 0.7, position = "stack") +
  # 添加group=indicator，强制按indicator的顺序堆叠文本
  geom_text(aes(label = label, color = label_col, group = indicator), 
            size = 3, position = position_stack(vjust = 0.5)) +
  scale_color_identity() +
  scale_fill_viridis(discrete = T, option = "plasma")

方法2：将label_col转换为与indicator绑定的因子 #

让label_col的因子水平顺序与indicator的顺序完全对应，确保分组顺序一致：

test %>% 
  mutate(indicator = factor(indicator, levels = list_order),
         label = round(mean, 1),
         label_col = ifelse(indicator %in% c("a", "d", "e", "f"), "black", "white"),
         # 按indicator的顺序生成label_col的因子水平，避免默认字母排序
         label_col = factor(label_col, levels = unique(label_col[match(list_order, indicator)]))) %>% 
  ggplot(aes(x = group, y = mean, fill = indicator)) +
  geom_col(width = 0.7, position = "stack") +
  geom_text(aes(label = label, color = label_col), 
            size = 3, position = position_stack(vjust = 0.5)) +
  scale_color_identity() +
  scale_fill_viridis(discrete = T, option = "plasma")

内容的提问来源于stack exchange，提问作者A. Piong