Why bother with purrr when I can just use a for loop?

I get this question a lot. Look, for loops aren't "wrong," but in R, they're often clunky. You have to pre-allocate a vector or a list to hold your results, manage an index variable, and write a lot of boilerplate code just to do something simple. It's a lot of surface area for bugs to hide in.

purrr gives us a consistent set of tools for "functional iteration." Instead of telling R how to loop (increment i, check length, assign to index), you tell it what to do to every element in a collection. It's cleaner, it's more readable, and it plays beautifully with the pipe operator. Once you get used to the map family of functions, going back to for loops feels like writing assembly language when you have a high-level language available.

What's with the tilde (~) and the .x syntax?

If you've looked at purrr code, you've probably seen things like map(data, ~ .x * 2). This is just "syntactic sugar" for creating an anonymous function. In base R, you'd have to write map(data, function(x) { x * 2 }). That's a lot of typing for a simple operation.

The tilde ~ tells purrr: "Everything following this is a formula that should be treated as a function." The .x is a placeholder for "the current item being processed." If you're mapping over two lists simultaneously using map2, you'll see .x and .y.

# The long way
results <- map(my_list, function(x) {
  mean(x, na.rm = TRUE)
})

# The purrr way (much faster to write)
results <- map(my_list, ~ mean(.x, na.rm = TRUE))

How do I avoid getting a list of lists back?

The standard map() always returns a list. That's great for flexibility, but it's a nightmare when you actually want a vector of numbers or a single data frame. This is where the "type-specific" map functions come in. I almost always use these instead of the generic map() because they enforce type stability.

If you want a numeric vector, use map_dbl(). If you want a character vector, map_chr(). But the real superpower is map_df(). It takes the result of each iteration (which should be a data frame) and stacks them all on top of each other into one big tibble. This is how I handle the "read 50 CSV files and combine them" problem in about two lines of code:

library(purrr)
library(readr)

files <- list.files(path = "data/monthly_reports", full.names = TRUE)

# Instead of a loop and rbind(), just do this:
all_data <- map_df(files, read_csv)

Is there a way to iterate just to "do something" without returning data?

Sometimes you don't want a result. Maybe you're saving a plot to a folder for every city in a dataset, or sending an email to a list of clients. If you use map() for this, R will dutifully return a massive list of NULL values or empty outputs, which just clutters your console.

For these "side effects," use walk(). It's identical to map() in terms of how it works, but it returns the original input invisibly. It's a signal to anyone reading your code that "I'm doing this for the side effect, not to transform the data."

# Saving a plot for each category
categories <- c("Electronics", "Clothing", "Home")

walk(categories, ~ {
  p <- ggplot(df, aes(x, y)) + filter(category == .x)
  ggsave(filename = paste0("plot_", .x, ".png"), plot = p)
})
# No giant list of NULLs printed to the screen!

📋 Practical Task

Exercise: Processing Regional Sales Logs

You have been given a list of file paths to three CSV files, each containing sales data for a different region. Your goal is to read these files and calculate the total revenue for each region, returning the results as a named numeric vector.

Setup: Run this code first to create the dummy files in your environment:

dir.create("sales_data", showWarnings = FALSE)
write.csv(data.frame(item="A", price=10, qty=2), "sales_data/north.csv", row.names=FALSE)
write.csv(data.frame(item="B", price=20, qty=1), "sales_data/south.csv", row.names=FALSE)
write.csv(data.frame(item="C", price=15, qty=3), "sales_data/east.csv", row.names=FALSE)

file_paths <- list.files("sales_data", full.names = TRUE)
names(file_paths) <- c("North", "South", "East")

Your Task: Use purrr::map_dbl() to iterate over file_paths. Inside the map function, read the CSV and calculate the total revenue (price * qty) for that file. The final output should be a numeric vector where the names are the regions and the values are the totals.