pnpm vs npm

npm's flat tree#

npm maintains a flattened dependency tree. This leads to less disk space bloat, with a messy node_modules directory as a side effect.

On the other hand, pnpm manages node_modules by using hard linking and symbolic linking to a global on-disk content-addressable store. This lets you get the benefits of far less disk space usage, while also keeping your node_modules clean. There is documentation on the store layout if you wish to learn more.

The good thing about pnpm's proper node_modules structure is that it "helps to avoid silly bugs" by making it impossible to use modules that are not specified in the project's package.json.

Installation#

pnpm does not allow installation of packages without saving them to package.json. If no parameters are passed to pnpm add, packages are saved as regular dependencies. Like with npm, --save-dev and --save-optional can be used to install packages as dev or optional dependencies.

As a consequence of this limitation, projects won't have any extraneous packages when they use pnpm unless they remove a dependency and leave it orphaned. That's why pnpm's implementation of the prune command does not allow you to specify packages to prune - it ALWAYS removes all extraneous and orphaned packages.

Directory dependencies#

A directory dependency points to a directory in the filesystem. pnpm handles it according to its protocol:

• link: (also recorded by pnpm add ./dir) symlinks the directory into node_modules. pnpm does not install the directory's own dependencies.
• file: copies the directory into the virtual store as a package and installs its dependencies like any other package.

For example, if a package called foo (<root>/foo) has bar@link:../bar as a dependency, pnpm does not install the dependencies of <root>/bar when you run pnpm install on foo.

If you need to run installations in several packages at the same time, for instance in the case of a monorepo, you should look at the documentation for pnpm -r.