并行化时含变量条件格式的flextable循环渲染RMarkdown失败

使用循环从带条件格式flextable的Rmd模板生成docx报告时，将串行for循环改为foreach() %dopar%并行运行后失败，报错“找不到对象'boolean'”。但用%do%串行运行完全正常，问题根源是flextable的条件格式公式中引用了主脚本定义的boolean变量。

复现示例 #

主脚本 #

library(dplyr)
library(foreach)
library(doParallel)

boolean <- TRUE

tables <- list()
for (n in 1:10) {
  tables[[n]] <- iris %>% slice_sample(n = 20)
}

number_of_cores <- parallel::detectCores() - 1
clusters <- parallel::makeCluster(number_of_cores)
doParallel::registerDoParallel(clusters)

foreach(
  x = 1:10,
  .export = c("boolean", "tables"),
  .packages = c("dplyr", "flextable", "officer"),
  .verbose = TRUE
) %dopar% {
  rmarkdown::render("Template.rmd", output_file = paste0("Iris Subset ", x))
}

报错的Rmd模板（变量在flextable公式内） #

---
title: "Iris Data"
output: word_document
date: ""
---

```{r table}
flextable(tables[[x]]) %>%
  bg(i = ~ (Species == "setosa" & boolean), bg = "yellow", part = "body") %>%
  bg(i = ~ (Species == "versicolor" & !boolean), bg = "skyblue", part = "body") %>%
  autofit()

正常运行的Rmd模板（变量在flextable外） #

---
title: "Iris Data"
output: word_document
date: ""
---

```{r table}
if (boolean) {
flextable(tables[[x]]) %>%
  bg(i = ~ (Species == "setosa"), bg = "yellow", part = "body") %>%
  autofit()
} else {
  flextable(tables[[x]]) %>%
  bg(i = ~ (Species == "versicolor"), bg = "skyblue", part = "body") %>%
  autofit()
}

原因分析 #

flextable的bg()函数中i参数的公式（~开头）会在公式自身的专属环境中求值，而非当前工作环境。并行运行时，子进程环境与主进程完全隔离，即便通过.export传递了boolean，公式仍无法在自身环境中找到该变量；而串行运行时，公式环境与工作环境一致，因此能正常识别变量。

解决办法 #

方法1：将变量注入数据框 #

把boolean作为列添加到目标表格中，让公式直接引用数据框内部的列，彻底避免依赖外部变量：

# 修改Rmd模板中的表格代码
flextable(tables[[x]] %>% mutate(boolean = !!boolean)) %>%
  bg(i = ~ (Species == "setosa" & boolean), bg = "yellow", part = "body") %>%
  bg(i = ~ (Species == "versicolor" & !boolean), bg = "skyblue", part = "body") %>%
  autofit()

方法2：用rlang::inject强制注入变量 #

使用rlang的注入功能，将外部变量直接插入公式，强制公式在当前环境中求值：

# 修改Rmd模板中的表格代码
library(rlang)

flextable(tables[[x]]) %>%
  bg(i = inject(~ (Species == "setosa" & !!boolean)), bg = "yellow", part = "body") %>%
  bg(i = inject(~ (Species == "versicolor" & !!(boolean))), bg = "skyblue", part = "body") %>%
  autofit()

方法3：提前计算逻辑向量，弃用公式 #

直接计算需要高亮的行索引，将索引值传给i参数，完全避开公式环境的问题：

# 修改Rmd模板中的表格代码
ft <- flextable(tables[[x]])
# 计算需要高亮的行位置
setosa_rows <- which(tables[[x]]$Species == "setosa" & boolean)
versicolor_rows <- which(tables[[x]]$Species == "versicolor" & !boolean)

ft %>%
  bg(i = setosa_rows, bg = "yellow", part = "body") %>%
  bg(i = versicolor_rows, bg = "skyblue", part = "body") %>%
  autofit()

额外优化：通过params传递变量（替代.export） #

可以不用.export，改用rmarkdown::render的params参数传递变量，逻辑更清晰，但仍需结合上述方法解决公式环境问题：

# 修改主脚本的foreach部分
foreach(
  x = 1:10,
  .packages = c("dplyr", "flextable", "officer", "rmarkdown"),
  .verbose = TRUE
) %dopar% {
  rmarkdown::render(
    "Template.rmd",
    output_file = paste0("Iris Subset ", x),
    params = list(boolean = boolean, table_data = tables[[x]])
  )
}

# 对应的Rmd模板开头添加params定义
---
title: "Iris Data"
output: word_document
date: ""
params:
  boolean: TRUE
  table_data: NULL
---

# 表格代码结合方法1（注入列）修改
```{r table}
flextable(params$table_data %>% mutate(boolean = params$boolean)) %>%
  bg(i = ~ (Species == "setosa" & boolean), bg = "yellow", part = "body") %>%
  bg(i = ~ (Species == "versicolor" & !boolean), bg = "skyblue", part = "body") %>%
  autofit()

内容的提问来源于stack exchange，提问作者Samuel Reichler