如何设置gt::data_color的全局着色方向而非按行/列

问题描述 #

使用gt包的data_color()函数时，发现其direction参数仅支持"column"或"row"选项，但希望基于表格中所有数值字段的全局范围进行着色，而非仅按行或列计算。以下是可复现代码：

require(gt)
n <- 5
set.seed(23)
data.frame(
  product_name = sample(LETTERS,n),
  performance_1 = rnorm(n,-1,1),
  performance_2 = runif(n,-1,1),
  performance_3 = runif(n,-1,1),
  performance_4 = rnorm(n,0,1)
) |>
  gt::gt() |>
  data_color(
    columns = contains("performance"),
    direction = "row", # 希望替换为全表范围计算
    fn = scales::col_numeric(
      palette = c("tomato","gold","green4"),
      domain = NULL
    )
  )

也可使用以下dput数据复现：

structure(list(product_name = c("X", "S", "H", "K", "G"), performance_1 = c(-0.00339489316645392, 
0.107490487448087, -1.27808628285053, 0.0192054876250842, -0.954562821242861
), performance_2 = c(0.884923394303769, 0.256039271131158, 0.172795631922781, 
-0.450518076308072, -0.70468607544899), performance_3 = c(0.60282064601779, 
-0.227180351037532, 0.640901388134807, 0.36987454816699, 0.766778532415628
), performance_4 = c(-1.21637642702189, 0.30813689628972, -0.52017831497405, 
-0.442313800737427, -0.599312811975944)), class = "data.frame", row.names = c(NA, 
-5L))

解决方案 #

不需要修改direction参数，而是通过手动指定scales::col_numeric()的domain参数为全表数值的全局范围，即可实现基于全表的着色逻辑。具体步骤如下：

1. 先提取所有目标数值列的全局最小值和最大值；
2. 将该范围传入col_numeric()的domain参数，此时direction的设置会被忽略，颜色映射将基于全局范围计算。

修改后的代码如下：

require(gt)
set.seed(23)
n <- 5
df <- data.frame(
  product_name = sample(LETTERS,n),
  performance_1 = rnorm(n,-1,1),
  performance_2 = runif(n,-1,1),
  performance_3 = runif(n,-1,1),
  performance_4 = rnorm(n,0,1)
)

# 计算所有performance列的全局最小/最大值
global_range <- range(df[, grepl("performance", names(df))])

df |>
  gt::gt() |>
  data_color(
    columns = contains("performance"),
    # direction可保留任意有效值，不影响最终着色逻辑
    direction = "row",
    fn = scales::col_numeric(
      palette = c("tomato","gold","green4"),
      domain = global_range
    )
  )

原理说明 #

当scales::col_numeric()的domain参数被显式指定时，函数会优先使用该固定范围进行颜色映射，而不会根据direction参数去计算行或列的局部范围。因此，即使direction设为"row"或"column"，最终的着色依然会基于全表的数值范围生成。

内容的提问来源于stack exchange，提问作者Yann