如何修改eval_data函数与logic列表实现R批量dplyr数据处理？

Let's walk through the issues with your current setup and fix them step by step to get your batch data processing working as intended:

1. Key Problems in the Original Code #

Your current eval_data function and logic list have a few critical issues that break some scenarios:

• Directly using eval(dplyr_logic) doesn't properly integrate dplyr expressions into the pipeline (you need quasiquotation tools from rlang here).
• The select({ select_vector }) syntax doesn't safely unquote your column selection vector.
• The I() in your logic list doesn't handle the "no transformation" case correctly for the pipeline.
• Chaining mutate calls with %>% inside an expr() causes evaluation errors.

2. Revised eval_data Function #

Here's the corrected function, with explanations of each change:

eval_data <- function(data, dplyr_logic, select_vector) {
  data %>%
    # Use !! to unquote and evaluate the dplyr expression in the pipeline
    { if (!rlang::is_empty(dplyr_logic)) !!dplyr_logic else . } %>%
    # Use all_of() to safely select columns from the character vector
    dplyr::select(dplyr::all_of(select_vector))
}

What Changed: #

• Quasiquotation with !!: The bang-bang operator (!!) tells dplyr to treat your dplyr_logic expression as part of the pipeline, not a literal object.
• Empty Logic Handling: The if/else check ensures that when there's no transformation needed, we just pass the original data through with ..
• Safe Column Selection: dplyr::all_of(select_vector) properly references column names from your vector, even for newly created columns like New_Column1.

3. Fixed logic List #

We need to adjust how we define the expressions to play nicely with the revised function:

logic <- list(
  # Replace I() with an empty expression for the no-transformation case
  rlang::expr(),
  rlang::expr(mutate(New_Column1 = case_when(Sepal.Length > 7 ~ 'Big', Sepal.Length > 6 ~ 'Medium', TRUE ~ 'Small'))),
  # Combine multiple mutate steps into a single mutate() call instead of chaining %>%
  rlang::expr(mutate(
    New_Column2 = case_when(Sepal.Width > 3.5 ~ 'Big2', Sepal.Width > 3 ~ 'Medium2', TRUE ~ 'Small2'),
    New_Column3 = case_when(Petal.Width > 2 ~ 'Big3', Petal.Width > 1 ~ 'Medium3', TRUE ~ 'Small3')
  )),
  rlang::expr(filter(Sepal.Width > 3))
)

What Changed: #

• Empty Expression for No Transformation: rlang::expr() replaces I() because I() wraps the data object instead of acting as a no-op in the pipeline.
• Combined Mutate Calls: Instead of chaining mutate with %>% inside expr(), we put both transformations into one mutate() call. This ensures the expression evaluates correctly in the pipeline.

4. Run the Batch Processing #

Now you can use pmap exactly as you intended to process all your data frames in bulk:

library(purrr)
library(dplyr)
library(rlang)

# Your original data and select vector definitions (unchanged)
dd <- list()
dd$data <- list(mutate0 = iris, mutate1 = iris, mutate2= iris, filter1 = iris )
select_vec <- list( 
  c("Species", "Sepal.Length"), 
  c("Species", "New_Column1"), 
  c("Species", "New_Column2", "New_Column3"), 
  c("Species", "Sepal.Width") 
)

# Execute the batch processing
results <- pmap(list(dd$data, logic, select_vec), eval_data)

# Inspect the output
lapply(results, head)

This will return a list of data frames, each transformed and filtered according to your specifications, with only the columns you selected retained.

内容的提问来源于stack exchange，提问作者MayaGans