如何实现geom_bar按validated_2填充、按count分面并添加计数？

原始数据 #

structure(list(validated_1 = c("sombra", "sombra", "sombra", 
"sombra", "sombra", "sombra", "sombra", "sombra", "sombra", "sombra", 
"coscinodiscus", "sombra", "coscinodiscus", "coscinodiscus", 
"sombra", "coscinodiscus", "sombra", "coscinodiscus", "sombra", 
"coscinodiscus", "coscinodiscus", "detritos", "detritos", "coscinodiscus", 
"appendicularia", "detritos", "coscinodiscus", "coscinodiscus", 
"detritos", "coscinodiscus", "langanho", "detritos", "copepodo", 
"langanho", "copepodo", "langanho", "langanho", "coscinodiscus", 
"coscinodiscus", "coscinodiscus"), validated_2 = c("sombra", 
"sombra", "sombra", "sombra", "sombra", "sombra", "sombra", "sombra", 
"sombra", "sombra", "coscinodiscus", "sombra", "coscinodiscus", 
"coscinodiscus", "sombra", "coscinodiscus", "sombra", "coscinodiscus", 
"sombra", "coscinodiscus", "coscinodiscus", "detritos", "detritos", 
"coscinodiscus", "zooplâncton", "detritos", "coscinodiscus", 
"coscinodiscus", "detritos", "coscinodiscus", "langanho", "detritos", 
"zooplâncton", "langanho", "zooplâncton", "langanho", "langanho", 
"coscinodiscus", "coscinodiscus", "coscinodiscus")), class = c("tbl_df", 
"tbl", "data.frame"), row.names = c(NA, -40L))

现有处理方式的不足 #

已尝试两种方式，但均无法满足全部需求：

• 方式1：支持按计数分面并添加总计数标签，但无法用validated_2变量填充条形

df %>% 
  group_by(validated_1) %>% 
  summarise(count = n()) %>%
  arrange(desc(count)) %>% 
  mutate(groups = c(rep("high N", 2), rep("lower N", 4))) %>% 
  ggplot(aes(x = reorder(validated_1, -count), y = count)) +
  geom_bar(stat = 'identity') +
  facet_wrap(~ groups, nrow = 2, scales = "free") +
  geom_text(aes(label = count), vjust = -0.5, size = 3)

• 方式2：支持用validated_2填充条形，但无法按计数分面、添加计数标签，且千万行级大数据集下运行速度极慢

df %>%
  ggplot(aes(x = fct_infreq(validated_1), fill = validated_2)) +
  geom_bar()

最终解决方案 #

核心思路是先对数据进行多层聚合统计，压缩数据量以适配大数据集，再基于聚合后的数据集实现所有需求：

library(tidyverse)

# 1. 多层聚合统计：先按两级标签分组计数，再计算每个一级标签的总计数
agg_df <- df %>%
  group_by(validated_1, validated_2) %>%
  summarise(sub_count = n(), .groups = "drop_last") %>%
  mutate(total_count = sum(sub_count)) %>%
  ungroup() %>%
  # 2. 按总计数排序，分配分面组并转为有序因子
  arrange(desc(total_count)) %>%
  mutate(
    validated_1 = fct_inorder(validated_1),  # 保留总计数从高到低的顺序
    groups = case_when(
      row_number() %in% 1:2 ~ "high N",  # 前2个为高计数组
      TRUE ~ "lower N"                    # 其余为低计数组
    )
  )

# 3. 绘制堆叠条形图：包含填充、分面、双计数标签
agg_df %>%
  ggplot(aes(x = validated_1, y = sub_count, fill = validated_2)) +
  geom_col(position = "stack") +
  # 添加子类别计数标签（位于堆叠条形中间）
  geom_text(aes(label = sub_count), position = position_stack(vjust = 0.5), size = 3) +
  # 添加一级标签总计数标签（位于条形顶部）
  geom_text(
    data = . %>% distinct(validated_1, total_count, groups),
    aes(y = total_count, label = total_count, fill = NULL),
    vjust = -0.5, size = 3.5
  ) +
  facet_wrap(~ groups, nrow = 2, scales = "free") +
  labs(x = "Validated 1", y = "Count", fill = "Validated 2") +
  theme_bw()

方案细节说明 #

1. 大数据适配：通过提前聚合统计，将千万行原始数据压缩为validated_1与validated_2的组合级别数据，避免ggplot直接处理海量原始数据，大幅提升运行速度。
2. 条形填充：使用fill = validated_2配合geom_col(position = "stack")实现堆叠条形的分组填充效果。
3. 按计数分面：先按总计数排序，通过case_when为每个validated_1分配分面组，再用facet_wrap实现分面展示。
4. 计数标签：
   • 子类别计数：利用position_stack将标签精准放在每个堆叠条形的中间位置；
   • 总计数：提取每个validated_1的总计数数据，将标签放在条形顶部，避免重复绘制。

内容的提问来源于stack exchange，提问作者Wilson Souza