fs-extra vs rimraf vs del

File System Deletion Utilities in Node.js Projects

del, fs-extra, and rimraf are all npm packages designed to help developers delete files and directories in Node.js environments, but they differ significantly in scope, API design, and intended use cases. rimraf is a minimal utility that replicates the behavior of the Unix rm -rf command, offering a simple promise/callback-based interface for recursive deletion. del builds on top of globby to support glob patterns, enabling powerful batch deletions with fine-grained control over matching and exclusion rules. fs-extra is a comprehensive extension of Node.js’s built-in fs module that includes a .remove() method (which internally uses rimraf) alongside dozens of other file system utilities like copying, ensuring, and JSON operations.

Package	Downloads	Stars	Size	Issues	Publish	License

fs-extra	126,833,896	9,602	56.3 kB	15	2 months ago	MIT
rimraf	108,429,233	5,816	260 kB	9	a day ago	BlueOak-1.0.0
del	13,946,077	1,342	12.7 kB	17	2 months ago	MIT

del vs fs-extra vs rimraf: Choosing the Right File Deletion Tool

All three packages solve the problem of deleting files and directories in Node.js, but they do so with different philosophies, capabilities, and trade-offs. Understanding their technical distinctions helps you avoid unnecessary bloat or missing functionality in your toolchain.

🗑️ Core Purpose and Scope

rimraf is purpose-built for one thing: recursively deleting files and folders, just like rm -rf in the terminal. It doesn’t care about patterns — you give it a path, and it removes it. Under the hood, it handles cross-platform quirks like Windows file locking by retrying with exponential backoff.

// rimraf: Delete a single directory
import rimraf from 'rimraf';

// Callback style
rimraf('./node_modules', (err) => {
  if (err) console.error(err);
});

// Promise style (v3+)
await rimraf('./build');

del focuses on pattern-based deletion. It uses glob syntax to match multiple files or directories at once, making it ideal for cleaning up build artifacts or temporary files across complex directory structures.

// del: Delete multiple files using globs
import del from 'del';

// Remove all .log files except error.log
await del(['logs/*.log', '!logs/error.log']);

// Clean dist folder but keep license.txt
await del(['dist/**/*', '!dist/license.txt']);

fs-extra is not primarily a deletion library — it’s a full-featured extension of Node’s fs module. Its .remove() method wraps rimraf, so deletion is just one of many utilities it provides (e.g., .copy(), .ensureDir(), .readJson()).

// fs-extra: Remove using the same engine as rimraf
import { remove } from 'fs-extra';

await remove('./temp');

// But you might also use it for other tasks
await copy('./src', './dest');
const config = await readJson('./config.json');

🔍 Pattern Matching: Glob Support

This is where the biggest practical difference lies.

del: Full glob support via globby. Supports negation (!), brace expansion ({a,b}), and deep matching (**). Handles arrays of patterns and respects .gitignore by default (configurable).
rimraf: No glob support. You must resolve paths yourself before passing them in.
fs-extra: No built-in glob support in .remove(). If you need patterns, you’d have to combine it with another package like glob.

If your task involves "delete everything in out/ except out/manifest.json", only del can express that concisely.

⚙️ Error Handling and Edge Cases

All three handle non-existent paths differently:

del: Silently ignores paths that don’t exist. Returns an array of deleted paths, so you can verify what was actually removed.
rimraf: Also ignores non-existent paths — consistent with rm -rf behavior.
fs-extra.remove(): Same as rimraf since it delegates to it.

However, when it comes to permission errors or locked files:

rimraf (and by extension fs-extra) implements retry logic on Windows, which is critical for reliability in CI environments or when dealing with node_modules.
del does not retry — it fails fast if a file can’t be deleted on the first attempt. This can be a gotcha in Windows-heavy workflows.

📦 Dependency Footprint and Integration

rimraf has zero dependencies. It’s tiny and focused.
del depends on globby, which pulls in fast-glob and related utilities. This adds weight but delivers powerful matching.
fs-extra bundles many utilities, so if you’re not using its other features, including it just for .remove() adds unnecessary code.

In a minimal CLI tool, rimraf is often the pragmatic choice. In a build pipeline using Webpack or Vite, del pairs naturally with existing glob-based configurations.

🔄 API Style and Modern JavaScript

rimraf: Offers both callback and promise APIs. The promise version requires importing from rimraf/promises in older versions, but v3+ supports top-level await rimraf(...).
del: Promise-only. Designed for async/await from the ground up.
fs-extra: Fully promise-compatible with optional callback support, matching Node’s fs.promises style.

If your codebase is strictly modern async/await, del and newer rimraf feel more natural. If you mix callbacks and promises, fs-extra and rimraf offer flexibility.

🛠️ When to Use Which?

Use `rimraf` if:

You’re writing a script that deletes one known path (e.g., rimraf node_modules).
You need maximum reliability on Windows with retry logic.
You want the smallest possible dependency.

Use `del` if:

You’re cleaning build outputs with complex inclusion/exclusion rules.
Your workflow already uses glob patterns elsewhere (e.g., in ESLint or Prettier configs).
You prefer a clean, promise-only API and don’t mind the extra dependencies.

Use `fs-extra` if:

You’re already using it for other file operations (copying, JSON handling, etc.).
You want a unified I/O interface across your project.
You don’t need globbing and want rimraf’s behavior without adding another direct dependency.

💡 Pro Tip: Don’t Mix Them Unnecessarily

It’s common to see projects install both rimraf and del “just in case,” but this duplicates effort. Pick one based on whether you need globbing:

Need patterns? → del
Just deleting fixed paths? → rimraf (or fs-extra if already present)

Remember: fs-extra.remove() is rimraf under the hood, so there’s no functional advantage to choosing one over the other for simple deletion — only ecosystem fit matters.

✅ Final Recommendation

For most frontend build pipelines, del is the best fit because cleanup tasks almost always involve glob patterns (dist/**/*, coverage/, etc.).

For CLI tools, install scripts, or minimal utilities, rimraf is leaner and more predictable.

Only reach for fs-extra for deletion if you’re already leveraging its broader feature set — otherwise, it’s overkill.

fs-extra:
Choose fs-extra if your project already depends on it for other file system operations (like copying directories or reading JSON files) and you want consistency across your I/O codebase. Its .remove() method offers the same recursive deletion capability as rimraf but within a broader toolkit. Avoid adding it solely for deletion — its full feature set may be overkill if you only need to remove files.
rimraf:
Choose rimraf when you need a lightweight, battle-tested utility that mimics rm -rf behavior exactly. It’s perfect for simple recursive deletions without globbing needs, especially in CLI tools or minimal scripts where dependency footprint matters. Its callback and promise APIs are straightforward, and it reliably handles permissions and locked files on Windows.
del:
Choose del when you need to delete multiple files or directories using glob patterns (e.g., dist/**/*, !dist/keep.js). It’s ideal for build scripts or cleanup tasks where pattern-based selection is essential. Its promise-first API integrates cleanly with modern async/await workflows, and it handles edge cases like non-existent paths gracefully by resolving rather than throwing.

fs-extra

rimraf

del

Node.js: fs-extra

fs-extra adds file system methods that aren't included in the native fs module and adds promise support to the fs methods. It also uses graceful-fs to prevent EMFILE errors. It should be a drop in replacement for fs.

Why?

I got tired of including mkdirp, rimraf, and ncp in most of my projects.

Installation

npm install fs-extra

Usage

CommonJS

fs-extra is a drop in replacement for native fs. All methods in fs are attached to fs-extra. All fs methods return promises if the callback isn't passed.

You don't ever need to include the original fs module again:

const fs = require('fs') // this is no longer necessary

you can now do this:

const fs = require('fs-extra')

or if you prefer to make it clear that you're using fs-extra and not fs, you may want to name your fs variable fse like so:

const fse = require('fs-extra')

you can also keep both, but it's redundant:

const fs = require('fs')
const fse = require('fs-extra')

NOTE: The deprecated constants fs.F_OK, fs.R_OK, fs.W_OK, & fs.X_OK are not exported on Node.js v24.0.0+; please use their fs.constants equivalents.

ESM

There is also an fs-extra/esm import, that supports both default and named exports. However, note that fs methods are not included in fs-extra/esm; you still need to import fs and/or fs/promises seperately:

import { readFileSync } from 'fs'
import { readFile } from 'fs/promises'
import { outputFile, outputFileSync } from 'fs-extra/esm'

Default exports are supported:

import fs from 'fs'
import fse from 'fs-extra/esm'
// fse.readFileSync is not a function; must use fs.readFileSync

but you probably want to just use regular fs-extra instead of fs-extra/esm for default exports:

import fs from 'fs-extra'
// both fs and fs-extra methods are defined

Sync vs Async vs Async/Await

Most methods are async by default. All async methods will return a promise if the callback isn't passed.

Sync methods on the other hand will throw if an error occurs.

Also Async/Await will throw an error if one occurs.

Example:

const fs = require('fs-extra')

// Async with promises:
fs.copy('/tmp/myfile', '/tmp/mynewfile')
  .then(() => console.log('success!'))
  .catch(err => console.error(err))

// Async with callbacks:
fs.copy('/tmp/myfile', '/tmp/mynewfile', err => {
  if (err) return console.error(err)
  console.log('success!')
})

// Sync:
try {
  fs.copySync('/tmp/myfile', '/tmp/mynewfile')
  console.log('success!')
} catch (err) {
  console.error(err)
}

// Async/Await:
async function copyFiles () {
  try {
    await fs.copy('/tmp/myfile', '/tmp/mynewfile')
    console.log('success!')
  } catch (err) {
    console.error(err)
  }
}

copyFiles()

Methods

Async

Sync

NOTE: You can still use the native Node.js methods. They are promisified and copied over to fs-extra. See notes on fs.read(), fs.write(), & fs.writev()

What happened to `walk()` and `walkSync()`?

They were removed from fs-extra in v2.0.0. If you need the functionality, walk and walkSync are available as separate packages, klaw and klaw-sync.

Third Party

CLI

fse-cli allows you to run fs-extra from a console or from npm scripts.

TypeScript

If you like TypeScript, you can use fs-extra with it: https://github.com/DefinitelyTyped/DefinitelyTyped/tree/master/types/fs-extra

File / Directory Watching

If you want to watch for changes to files or directories, then you should use chokidar.

Obtain Filesystem (Devices, Partitions) Information

fs-filesystem allows you to read the state of the filesystem of the host on which it is run. It returns information about both the devices and the partitions (volumes) of the system.

Misc.

fs-extra-debug - Send your fs-extra calls to debug.
mfs - Monitor your fs-extra calls.

Hacking on fs-extra

Wanna hack on fs-extra? Great! Your help is needed! fs-extra is one of the most depended upon Node.js packages. This project uses JavaScript Standard Style - if the name or style choices bother you, you're gonna have to get over it :) If standard is good enough for npm, it's good enough for fs-extra.

What's needed?

First, take a look at existing issues. Those are probably going to be where the priority lies.
More tests for edge cases. Specifically on different platforms. There can never be enough tests.
Improve test coverage.

Note: If you make any big changes, you should definitely file an issue for discussion first.

Running the Test Suite

fs-extra contains hundreds of tests.

npm run lint: runs the linter (standard)
npm run unit: runs the unit tests
npm run unit-esm: runs tests for fs-extra/esm exports
npm test: runs the linter and all tests

When running unit tests, set the environment variable CROSS_DEVICE_PATH to the absolute path of an empty directory on another device (like a thumb drive) to enable cross-device move tests.

Windows

If you run the tests on the Windows and receive a lot of symbolic link EPERM permission errors, it's because on Windows you need elevated privilege to create symbolic links. You can add this to your Windows's account by following the instructions here: http://superuser.com/questions/104845/permission-to-make-symbolic-links-in-windows-7 However, I didn't have much luck doing this.

Since I develop on Mac OS X, I use VMWare Fusion for Windows testing. I create a shared folder that I map to a drive on Windows. I open the Node.js command prompt and run as Administrator. I then map the network drive running the following command:

net use z: "\\vmware-host\Shared Folders"

I can then navigate to my fs-extra directory and run the tests.

Naming

I put a lot of thought into the naming of these functions. Inspired by @coolaj86's request. So he deserves much of the credit for raising the issue. See discussion(s) here:

First, I believe that in as many cases as possible, the Node.js naming schemes should be chosen. However, there are problems with the Node.js own naming schemes.

For example, fs.readFile() and fs.readdir(): the F is capitalized in File and the d is not capitalized in dir. Perhaps a bit pedantic, but they should still be consistent. Also, Node.js has chosen a lot of POSIX naming schemes, which I believe is great. See: fs.mkdir(), fs.rmdir(), fs.chown(), etc.

We have a dilemma though. How do you consistently name methods that perform the following POSIX commands: cp, cp -r, mkdir -p, and rm -rf?

My perspective: when in doubt, err on the side of simplicity. A directory is just a hierarchical grouping of directories and files. Consider that for a moment. So when you want to copy it or remove it, in most cases you'll want to copy or remove all of its contents. When you want to create a directory, if the directory that it's suppose to be contained in does not exist, then in most cases you'll want to create that too.

So, if you want to remove a file or a directory regardless of whether it has contents, just call fs.remove(path). If you want to copy a file or a directory whether it has contents, just call fs.copy(source, destination). If you want to create a directory regardless of whether its parent directories exist, just call fs.mkdirs(path) or fs.mkdirp(path).

Credit

fs-extra wouldn't be possible without using the modules from the following authors:

License

Licensed under MIT

Copyright (c) 2011-2024 JP Richardson