dplyr::pull、purrr::pluck与magrittr::extract2的区别及适用场景

Great question! Let's break down these three functions—they all let you pull a single vector from a data frame (and beyond, in some cases) but have distinct strengths and ideal use cases, especially when working within the tidyverse ecosystem.

Core Differences, Advantages & Use Cases #

1. dplyr::pull(): The Tidyverse Data Frame Workhorse #

This is the go-to if you're already working in a dplyr pipeline. It's purpose-built for extracting columns from data frames (and tibbles) with tidyverse-friendly features that the other two don't offer:

• Tidy selection support: You can use helpers like contains(), starts_with(), or ends_with() to target columns without typing the exact name.
• Vector naming: Use the name argument to set the names of your output vector using another column from the data frame.
• Explicit error handling: If you reference a column that doesn't exist, pull() throws a clear error instead of returning NULL (like the other two).

Best for: When you're wrapping up a dplyr workflow (filtering, mutating, grouping) and need to extract a column as a vector. It plays nicely with other tidyverse functions and makes your code more readable if you're already using dplyr syntax.

Example code:

# Extract by exact column name
mtcars %>% mutate(wt_to_hp = wt/hp) %>% pull(wt_to_hp)

# Extract by position (1 = first column, mpg)
mtcars %>% pull(1)

# Use tidy selector to grab a column matching a pattern
mtcars %>% pull(contains("t")) # Returns the `wt` column

# Name the output vector using values from another column
mtcars %>% pull(wt, name = hp) # Vector of wt values, named with hp values

2. magrittr::extract2(): The Minimal [[ Pipe Wrapper #

Think of this as a pipe-friendly version of base R's [[ operator. It does one thing and does it simply: extracts a single element (column, in data frame terms) using a name or position, just like df[[col]].

Advantages:

• Lightweight and consistent with other magrittr extract functions (like extract() which is the pipe version of [).
• No extra bells and whistles—perfect if you want a direct, base-R-equivalent operation in a pipe.

Best for: When you're using magrittr pipes but don't need the extra features of pull() or pluck(). It's a clean, minimal choice for straightforward column extraction.

Example code:

# Exact equivalent to mtcars[["wt"]] in pipe form
mtcars %>% extract2("wt")

# Extract by position
mtcars %>% extract2(3) # Returns the `disp` column

3. purrr::pluck(): The Universal Nested Structure Extractor #

This is the most flexible of the three—it's not limited to data frames. pluck() is designed to extract elements from any nested structure: lists, nested data frames, lists of lists, etc.

Key strengths:

• Multi-level extraction: You can pass multiple indices to dig into nested structures in one step (e.g., extract a column from a nested data frame inside a list).
• Works with all R objects: Not just data frames—use it to pull values from lists, S3 objects, or even nested tibbles.
• Graceful failure: Like extract2() and [[, it returns NULL if the element doesn't exist (instead of throwing an error).

Best for: When you're dealing with nested data (common in tasks like JSON parsing, grouped data with nest(), or list columns) and need to extract elements across multiple levels. It's overkill for simple data frame column pulls, but indispensable for complex nested structures.

Example code:

# Basic data frame column extraction (same as extract2() here)
mtcars %>% mutate(wt_to_hp = wt/hp) %>% pluck("wt_to_hp")

# Extract from a nested list
nested_data <- list(
  metadata = list(year = 2024, source = "mtcars"),
  data = mtcars
)
nested_data %>% pluck("metadata", "year") # Returns 2024

# Extract from a nested data frame
mtcars %>% group_by(cyl) %>% nest() %>% pluck("data", 2, "hp") # Pulls hp column from the 6-cylinder group

Quick Cheat Sheet #

内容的提问来源于stack exchange，提问作者crazybilly