如何修正R中ggalluvial/ggsankey桑基图的id变量排序

现有如下R数据集，使用ggsankey预处理后用ggalluvial绘制桑基图时，id节点的排序出现异常，显示为A1、A10、A2...的顺序，而非预期的A1到A10依次排列。

原始数据集代码 #

id <- c("A1", "A2", "A3", "A4", "A5", "A6", "A7", "A8", "A9", "A10")
variation <- c("aaa1", "aaa1", "bbb1", "aaa2", "b1","a3", "a1", "b1", "a1", "b1" )
result <- c("Way1", "Way1", "Way2", "Way2", "Way3","Way1", "Way2", "Way3", "Way4", "Way1" )

data <- data.frame(id, variation, result)
head(data)

#     id variation result
# 1   A1      aaa1   Way1
# 2   A2      aaa1   Way1
# 3   A3      bbb1   Way2
# 4   A4      aaa2   Way2
# 5   A5        b1   Way3
# 6   A6        a3   Way1
# 7   A7        a1   Way2
# 8   A8        b1   Way3
# 9   A9        a1   Way4
# 10 A10        b1   Way1

桑基图预处理与绘制代码 #

library(ggsankey)

data_df <- data %>%
  make_long(id, variation, result)
head(data_df)
# A tibble: 6 × 4
# x         node  next_x    next_node
# <fct>     <chr> <fct>     <chr>    
#   1 id        A1    variation aaa1     
# 2 variation aaa1  result    Way1     
# 3 result    Way1  NA        NA       
# 4 id        A2    variation aaa1     
# 5 variation aaa1  result    Way1     
# 6 result    Way1  NA        NA  

library(ggalluvial)

ggplot(data_df, aes(x = x, next_x = next_x, node = node, next_node = next_node, fill = factor(node), label = node)) +
  geom_alluvial(flow.alpha = .6) +
  geom_alluvial_text(size = 3, color = "black") +
  scale_fill_viridis_d() +
  theme_alluvial(base_size = 18) +
  labs(x = NULL) +
  theme(legend.position = "none",
        plot.title = element_text(hjust = .5)) +
  ggtitle("")

问题根源在于id是字符型变量，R默认按字典序排序（"A10"的首字符"A"后是"1"，比"A2"的"2"小，所以排在前面）。要解决这个问题，需要将id转换为有序因子，并指定正确的排序顺序，具体步骤如下：

• 在原始数据中设置id的有序因子属性
  在创建data数据框前，先把id转换成有序因子，顺序按原始向量的顺序：

  id <- factor(c("A1", "A2", "A3", "A4", "A5", "A6", "A7", "A8", "A9", "A10"), 
               levels = c("A1", "A2", "A3", "A4", "A5", "A6", "A7", "A8", "A9", "A10"),
               ordered = TRUE)
  

  如果已经创建了data数据框，也可以直接修改列属性：

  data$id <- factor(data$id, levels = data$id, ordered = TRUE)
  

• 重新生成桑基图数据集并绘图
  重新运行make_long预处理代码，再绘制桑基图，此时id节点会按A1到A10的顺序排列。

如果不想修改原始数据，也可以直接在预处理后的data_df中调整node列的因子顺序（需要列出所有节点的水平）：

data_df$node <- factor(data_df$node, 
                       levels = c("A1", "A2", "A3", "A4", "A5", "A6", "A7", "A8", "A9", "A10", 
                                  "aaa1", "bbb1", "aaa2", "b1", "a3", "a1", 
                                  "Way1", "Way2", "Way3", "Way4"),
                       ordered = TRUE)

内容的提问来源于stack exchange，提问作者choij