How to Use map2() and pmap() in R

purrr

purrr map2()

Learn to iterate over multiple lists in parallel with purrr’s map2() and pmap(). Apply functions with 2 or more arguments across corresponding elements.

Published

April 3, 2026

Introduction

While map() iterates over a single list, map2() iterates over two lists in parallel, and pmap() iterates over any number of lists. These functions are essential when your operation needs inputs from multiple sources.

When to use: - map2() - when you have exactly two parallel inputs - pmap() - when you have three or more parallel inputs, or a data frame of parameters

Getting Started

library(tidyverse)

map2(): Iterate Over Two Lists

Basic syntax

map2() takes two lists and applies a function using corresponding elements:

xs <- list(1, 2, 3)
ys <- list(10, 20, 30)

# Add corresponding elements
map2(xs, ys, \(x, y) x + y)

Practical example: weighted means

values <- list(
  a = c(1, 2, 3),
  b = c(10, 20, 30),
  c = c(100, 200, 300)
)

weights <- list(
  a = c(0.5, 0.3, 0.2),
  b = c(0.1, 0.2, 0.7),
  c = c(0.33, 0.33, 0.34)
)

# Calculate weighted mean for each pair
map2_dbl(values, weights, weighted.mean)

String operations with two inputs

first_names <- c("John", "Jane", "Bob")
last_names <- c("Doe", "Smith", "Johnson")

# Combine names
map2_chr(first_names, last_names, \(f, l) paste(f, l))

# Or using paste directly
map2_chr(first_names, last_names, paste)

Creating file paths

folders <- c("data", "data", "output")
files <- c("input.csv", "config.json", "results.csv")

map2_chr(folders, files, file.path)

map2() Variants

Just like map(), map2() has typed variants:

xs <- 1:3
ys <- 4:6

map2_dbl(xs, ys, `+`)    # numeric: 5, 7, 9
map2_int(xs, ys, `+`)    # integer: 5L, 7L, 9L
map2_chr(xs, ys, paste)  # character: "1 4", "2 5", "3 6"
map2_lgl(xs, ys, `<`)    # logical: TRUE, TRUE, TRUE

pmap(): Iterate Over Multiple Lists

Basic syntax

pmap() takes a list of lists (or a data frame) and applies a function:

params <- list(
  n = c(10, 20, 30),
  mean = c(0, 5, 10),
  sd = c(1, 2, 3)
)

# Generate random samples with different parameters
set.seed(42)
pmap(params, rnorm)

Using a data frame as input

Data frames are naturally suited for pmap():

# Create parameter grid
param_grid <- expand_grid(
  n = c(100, 1000),
  mean = c(0, 10),
  sd = c(1, 5)
)

param_grid

# Generate samples for each row
set.seed(42)
samples <- pmap(param_grid, rnorm)
map_dbl(samples, mean)  # Check the means

Named arguments matter

pmap() matches list names to function arguments:

# Names must match function argument names
params <- list(
  x = c(1, 2, 3),
  y = c(4, 5, 6),
  z = c(7, 8, 9)
)

# Custom function
my_sum <- function(x, y, z) x + y + z

pmap_dbl(params, my_sum)

Practical Example: Model Comparison

Run multiple models with different specifications

# Define model specifications
model_specs <- tibble(
  formula = c(
    "mpg ~ wt",
    "mpg ~ wt + hp",
    "mpg ~ wt + hp + am"
  ),
  data = list(mtcars, mtcars, mtcars)
)

# Fit all models
models <- pmap(model_specs, \(formula, data) {
  lm(as.formula(formula), data = data)
})

# Extract R-squared values
map_dbl(models, \(m) summary(m)$r.squared)

Create multiple plots

# Parameters for plots
plot_params <- tibble(
  x_var = c("wt", "hp", "disp"),
  y_var = rep("mpg", 3),
  title = c("Weight vs MPG", "Horsepower vs MPG", "Displacement vs MPG")
)

# Generate plots (returns list of ggplots)
plots <- pmap(plot_params, \(x_var, y_var, title) {
  ggplot(mtcars, aes(x = .data[[x_var]], y = .data[[y_var]])) +
    geom_point() +
    geom_smooth(method = "lm") +
    labs(title = title) +
    theme_minimal()
})

# Display first plot
plots[[1]]

Scatter plot of car weight versus miles per gallon created with pmap and ggplot2 in R showing a linear regression trend line

imap(): Iterate with Index or Name

When you need both the element AND its index/name, use imap():

x <- c(a = 10, b = 20, c = 30)

# imap passes element and name/index
imap_chr(x, \(value, name) paste(name, "=", value))
# "a = 10", "b = 20", "c = 30"

With unnamed vectors, you get the index

letters[1:3] |>
  imap_chr(\(letter, idx) paste0(idx, ": ", letter))
# "1: a", "2: b", "3: c"

Practical use: progress messages

files <- c("data1.csv", "data2.csv", "data3.csv")

iwalk(files, \(file, i) {
  message(sprintf("[%d/%d] Processing %s", i, length(files), file))
  # read_csv(file) ...
})

Add row numbers to nested data

penguins |>
  drop_na() |>
  split(~species) |>
  imap(\(df, species_name) {
    df |> mutate(species_label = species_name)
  }) |>
  list_rbind()

Performance: pmap vs Nested Loops

pmap() is cleaner and often faster than nested loops:

# Parameter grid
params <- expand_grid(a = 1:10, b = 1:10, c = 1:10)

# Instead of nested loops:
# for (a in 1:10) {
#   for (b in 1:10) {
#     for (c in 1:10) { ... }
#   }
# }

# Use pmap:
results <- pmap_dbl(params, \(a, b, c) a * b + c)

This is more readable and avoids growing vectors in loops.

walk2() and pwalk(): For Side Effects

When you want side effects (like saving files) instead of return values:

# Save multiple datasets
datasets <- list(
  iris = iris,
  mtcars = mtcars
)
filenames <- c("iris.csv", "mtcars.csv")

# This would write files (commented out)
# walk2(datasets, filenames, \(data, file) {
#   write_csv(data, file)
# })

Common Mistakes

1. Lists of different lengths

# Shorter list is NOT recycled (unlike base R)
xs <- 1:3
ys <- 1:2

# This errors:
# map2(xs, ys, `+`)

# Make lengths match explicitly
map2(xs, c(ys, NA), `+`)

2. Forgetting to name pmap arguments

# This might not work as expected if names don't match
params <- list(c(10, 20), c(0, 5), c(1, 2))  # unnamed

# Better: use names that match function arguments
params <- list(n = c(10, 20), mean = c(0, 5), sd = c(1, 2))
pmap(params, rnorm)

3. Using map2 when you need pmap

# For more than 2 inputs, use pmap
# Wrong approach: nested map2
# Right approach: pmap with all inputs

Summary

Function	Inputs	Returns	Use Case
`map2()`	2 lists	list	Combine two parallel sequences
`map2_dbl()`	2 lists	numeric	Two inputs, numeric output
`pmap()`	n lists	list	Multiple parameters, complex operations
`pmap_dfr()`	n lists	data frame	Parameter grid → combined results
`walk2()`	2 lists	invisible	Two inputs, side effects only
`pwalk()`	n lists	invisible	Multiple inputs, side effects

Use map2() for exactly two parallel inputs
Use pmap() for three or more inputs, or when working with data frames
Name your list elements to match function argument names
Use walk2()/pwalk() for side effects like saving files

--- title: "How to Use map2() and pmap() in R" description: "Learn to iterate over multiple lists in parallel with purrr's map2() and pmap(). Apply functions with 2 or more arguments across corresponding elements." date: 2026-04-03 categories: ['purrr', 'purrr map2()'] image: /images/purrr/map2-pmap-in-r-weight-vs-mpg-ggplot.png format: html: code-fold: false code-tools: true --- ## Introduction While `map()` iterates over a single list, `map2()` iterates over two lists in parallel, and `pmap()` iterates over any number of lists. These functions are essential when your operation needs inputs from multiple sources. **When to use:** - `map2()` - when you have exactly two parallel inputs - `pmap()` - when you have three or more parallel inputs, or a data frame of parameters ## Getting Started ```r library(tidyverse) ``` ## map2(): Iterate Over Two Lists ### Basic syntax `map2()` takes two lists and applies a function using corresponding elements: ```r xs <- list(1, 2, 3) ys <- list(10, 20, 30) # Add corresponding elements map2(xs, ys, \(x, y) x + y) ``` ### Practical example: weighted means ```r values <- list( a = c(1, 2, 3), b = c(10, 20, 30), c = c(100, 200, 300) ) weights <- list( a = c(0.5, 0.3, 0.2), b = c(0.1, 0.2, 0.7), c = c(0.33, 0.33, 0.34) ) # Calculate weighted mean for each pair map2_dbl(values, weights, weighted.mean) ``` ### String operations with two inputs ```r first_names <- c("John", "Jane", "Bob") last_names <- c("Doe", "Smith", "Johnson") # Combine names map2_chr(first_names, last_names, \(f, l) paste(f, l)) # Or using paste directly map2_chr(first_names, last_names, paste) ``` ### Creating file paths ```r folders <- c("data", "data", "output") files <- c("input.csv", "config.json", "results.csv") map2_chr(folders, files, file.path) ``` ## map2() Variants Just like `map()`, `map2()` has typed variants: ```r xs <- 1:3 ys <- 4:6 map2_dbl(xs, ys, `+`) # numeric: 5, 7, 9 map2_int(xs, ys, `+`) # integer: 5L, 7L, 9L map2_chr(xs, ys, paste) # character: "1 4", "2 5", "3 6" map2_lgl(xs, ys, `<`) # logical: TRUE, TRUE, TRUE ``` ## pmap(): Iterate Over Multiple Lists ### Basic syntax `pmap()` takes a list of lists (or a data frame) and applies a function: ```r params <- list( n = c(10, 20, 30), mean = c(0, 5, 10), sd = c(1, 2, 3) ) # Generate random samples with different parameters set.seed(42) pmap(params, rnorm) ``` ### Using a data frame as input Data frames are naturally suited for `pmap()`: ```r # Create parameter grid param_grid <- expand_grid( n = c(100, 1000), mean = c(0, 10), sd = c(1, 5) ) param_grid # Generate samples for each row set.seed(42) samples <- pmap(param_grid, rnorm) map_dbl(samples, mean) # Check the means ``` ### Named arguments matter `pmap()` matches list names to function arguments: ```r # Names must match function argument names params <- list( x = c(1, 2, 3), y = c(4, 5, 6), z = c(7, 8, 9) ) # Custom function my_sum <- function(x, y, z) x + y + z pmap_dbl(params, my_sum) ``` ## Practical Example: Model Comparison ### Run multiple models with different specifications ```r # Define model specifications model_specs <- tibble( formula = c( "mpg ~ wt", "mpg ~ wt + hp", "mpg ~ wt + hp + am" ), data = list(mtcars, mtcars, mtcars) ) # Fit all models models <- pmap(model_specs, \(formula, data) { lm(as.formula(formula), data = data) }) # Extract R-squared values map_dbl(models, \(m) summary(m)$r.squared) ``` ### Create multiple plots ```r # Parameters for plots plot_params <- tibble( x_var = c("wt", "hp", "disp"), y_var = rep("mpg", 3), title = c("Weight vs MPG", "Horsepower vs MPG", "Displacement vs MPG") ) # Generate plots (returns list of ggplots) plots <- pmap(plot_params, \(x_var, y_var, title) { ggplot(mtcars, aes(x = .data[[x_var]], y = .data[[y_var]])) + geom_point() + geom_smooth(method = "lm") + labs(title = title) + theme_minimal() }) # Display first plot plots[[1]] ``` ![Scatter plot of car weight versus miles per gallon created with pmap and ggplot2 in R showing a linear regression trend line](/images/purrr/map2-pmap-in-r-weight-vs-mpg-ggplot.png) ## imap(): Iterate with Index or Name When you need both the element AND its index/name, use `imap()`: ```r x <- c(a = 10, b = 20, c = 30) # imap passes element and name/index imap_chr(x, \(value, name) paste(name, "=", value)) # "a = 10", "b = 20", "c = 30" ``` ### With unnamed vectors, you get the index ```r letters[1:3] |> imap_chr(\(letter, idx) paste0(idx, ": ", letter)) # "1: a", "2: b", "3: c" ``` ### Practical use: progress messages ```r files <- c("data1.csv", "data2.csv", "data3.csv") iwalk(files, \(file, i) { message(sprintf("[%d/%d] Processing %s", i, length(files), file)) # read_csv(file) ... }) ``` ### Add row numbers to nested data ```r penguins |> drop_na() |> split(~species) |> imap(\(df, species_name) { df |> mutate(species_label = species_name) }) |> list_rbind() ``` ## Performance: pmap vs Nested Loops `pmap()` is cleaner and often faster than nested loops: ```r # Parameter grid params <- expand_grid(a = 1:10, b = 1:10, c = 1:10) # Instead of nested loops: # for (a in 1:10) { # for (b in 1:10) { # for (c in 1:10) { ... } # } # } # Use pmap: results <- pmap_dbl(params, \(a, b, c) a * b + c) ``` This is more readable and avoids growing vectors in loops. ## walk2() and pwalk(): For Side Effects When you want side effects (like saving files) instead of return values: ```r # Save multiple datasets datasets <- list( iris = iris, mtcars = mtcars ) filenames <- c("iris.csv", "mtcars.csv") # This would write files (commented out) # walk2(datasets, filenames, \(data, file) { # write_csv(data, file) # }) ``` ## Common Mistakes **1. Lists of different lengths** ```r # Shorter list is NOT recycled (unlike base R) xs <- 1:3 ys <- 1:2 # This errors: # map2(xs, ys, `+`) # Make lengths match explicitly map2(xs, c(ys, NA), `+`) ``` **2. Forgetting to name pmap arguments** ```r # This might not work as expected if names don't match params <- list(c(10, 20), c(0, 5), c(1, 2)) # unnamed # Better: use names that match function arguments params <- list(n = c(10, 20), mean = c(0, 5), sd = c(1, 2)) pmap(params, rnorm) ``` **3. Using map2 when you need pmap** ```r # For more than 2 inputs, use pmap # Wrong approach: nested map2 # Right approach: pmap with all inputs ``` ## Summary | Function | Inputs | Returns | Use Case | |----------|--------|---------|----------| | `map2()` | 2 lists | list | Combine two parallel sequences | | `map2_dbl()` | 2 lists | numeric | Two inputs, numeric output | | `pmap()` | n lists | list | Multiple parameters, complex operations | | `pmap_dfr()` | n lists | data frame | Parameter grid → combined results | | `walk2()` | 2 lists | invisible | Two inputs, side effects only | | `pwalk()` | n lists | invisible | Multiple inputs, side effects | - Use `map2()` for exactly two parallel inputs - Use `pmap()` for three or more inputs, or when working with data frames - Name your list elements to match function argument names - Use `walk2()`/`pwalk()` for side effects like saving files ## Related Posts - [How to Use map() in R](/purrr/how-to-use-map-in-r) - [How to Use map_dbl() and map_chr() in R](/purrr/how-to-use-map-variants-in-r) - [How to Use reduce() in R](/purrr/how-to-use-reduce-in-r) - [How to Use expand_grid() in R](/tidyr/how-to-create-all-combinations-with-expandgrid-in-r) - [How to Use nest() in R](/tidyr/how-to-use-nest-in-r)