跳到主要内容

Frequently Asked Questions

Why does my node_modules folder use disk space if packages are stored in a global store?

pnpm creates hard links from the global store to the project's node_modules folders. Hard links point to the same place on the disk where the original files are. So, for example, if you have foo in your project as a dependency and it occupies 1MB of space, then it will look like it occupies 1MB of space in the project's node_modules folder and the same amount of space in the global store. However, that 1MB is the same space on the disk addressed from two different locations. So in total foo occupies 1MB, not 2MB.

For more on this subject:

Does it work on Windows?

Short answer: Yes. Long answer: Using symbolic linking on Windows is problematic to say the least, however, pnpm has a workaround. For Windows, we use junctions instead.

But the nested node_modules approach is incompatible with Windows?

Early versions of npm had issues because of nesting all node_modules (see this issue). However, pnpm does not create deep folders, it stores all packages flatly and uses symbolic links to create the dependency tree structure.

Although pnpm uses linking to put dependencies into node_modules folders, circular symlinks are avoided because parent packages are placed into the same node_modules folder in which their dependencies are. So foo's dependencies are not in foo/node_modules, but foo is in node_modules together with its own dependencies.

One package can have different sets of dependencies on one machine.

In project A foo@1.0.0 can have a dependency resolved to bar@1.0.0, but in project B the same dependency of foo might resolve to bar@1.1.0; so, pnpm hard links foo@1.0.0 to every project where it is used, in order to create different sets of dependencies for it.

Direct symlinking to the global store would work with Node's --preserve-symlinks flag, however, that approach comes with a plethora of its own issues, so we decided to stick with hard links. For more details about why this decision was made, see this issue.

Does pnpm work across different subvolumes in one Btrfs partition?

While Btrfs does not allow cross-device hardlinks between different subvolumes in a single partition, it does permit reflinks. As a result, pnpm utilizes reflinks to share data between these subvolumes.

Does pnpm work across multiple drives or filesystems?

The package store should be on the same drive and filesystem as installations, otherwise packages will be copied, not linked. This is due to a limitation in how hard linking works, in that a file on one filesystem cannot address a location in another. See Issue #712 for more details.

pnpm functions differently in the 2 cases below:

Store path is specified

If the store path is specified via the store config, then copying occurs between the store and any projects that are on a different disk.

If you run pnpm install on disk A, then the pnpm store must be on disk A. If the pnpm store is located on disk B, then all required packages will be directly copied to the project location instead of being linked. This severely inhibits the storage and performance benefits of pnpm.

Store path is NOT specified

If the store path is not set, then multiple stores are created (one per drive or filesystem).

If installation is run on disk A, the store will be created on A .pnpm-store under the filesystem root. If later the installation is run on disk B, an independent store will be created on B at .pnpm-store. The projects would still maintain the benefits of pnpm, but each drive may have redundant packages.

What does pnpm stand for?

pnpm stands for performant npm. @rstacruz came up with the name.

pnpm does not work with <YOUR-PROJECT-HERE>?

In most cases it means that one of the dependencies require packages not declared in package.json. It is a common mistake caused by flat node_modules. If this happens, this is an error in the dependency and the dependency should be fixed. That might take time though, so pnpm supports workarounds to make the buggy packages work.

Solution 1

In case there are issues, you can use the node-linker=hoisted setting. This creates a flat node_modules structure similar to the one created by npm.

Solution 2

In the following example, a dependency does not have the iterall module in its own list of deps.

The easiest solution to resolve missing dependencies of the buggy packages is to add iterall as a dependency to our project's package.json.

You can do so, by installing it via pnpm add iterall, and will be automatically added to your project's package.json.

  "dependencies": {
...
"iterall": "^1.2.2",
...
}

Solution 3

One of the solutions is to use hooks for adding the missing dependencies to the package's package.json.

An example was Webpack Dashboard which wasn't working with pnpm. It has since been resolved such that it works with pnpm now.

It used to throw an error:

Error: Cannot find module 'babel-traverse'
at /node_modules/inspectpack@2.2.3/node_modules/inspectpack/lib/actions/parse

The problem was that babel-traverse was used in inspectpack which was used by webpack-dashboard, but babel-traverse wasn't specified in inspectpack's package.json. It still worked with npm and yarn because they create flat node_modules.

The solution was to create a .pnpmfile.cjs with the following contents:

module.exports = {
hooks: {
readPackage: (pkg) => {
if (pkg.name === "inspectpack") {
pkg.dependencies['babel-traverse'] = '^6.26.0';
}
return pkg;
}
}
};

After creating a .pnpmfile.cjs, delete pnpm-lock.yaml only - there is no need to delete node_modules, as pnpm hooks only affect module resolution. Then, rebuild the dependencies & it should be working.