Skip to content

Latest commit

 

History

History
427 lines (321 loc) · 18 KB

README.md

File metadata and controls

427 lines (321 loc) · 18 KB

Disclaimer: pkg was created for use within containers and is not intended for use in serverless environments. For those using Vercel, this means that there is no requirement to use pkg in your projects as the benefits it provides are not applicable to the platform.

Build Status

This command line interface enables you to package your Node.js project into an executable that can be run even on devices without Node.js installed.

Use Cases

  • Make a commercial version of your application without sources
  • Make a demo/evaluation/trial version of your app without sources
  • Instantly make executables for other platforms (cross-compilation)
  • Make some kind of self-extracting archive or installer
  • No need to install Node.js and npm to run the packaged application
  • No need to download hundreds of files via npm install to deploy your application. Deploy it as a single file
  • Put your assets inside the executable to make it even more portable
  • Test your app against new Node.js version without installing it

Usage

npm install -g pkg

After installing it, run pkg --help without arguments to see list of options:

pkg [options] <input>

  Options:

    -h, --help           output usage information
    -v, --version        output pkg version
    -t, --targets        comma-separated list of targets (see examples)
    -c, --config         package.json or any json file with top-level config
    --options            bake v8 options into executable to run with them on
    -o, --output         output file name or template for several files
    --out-path           path to save output one or more executables
    -d, --debug          show more information during packaging process [off]
    -b, --build          don't download prebuilt base binaries, build them
    --public             speed up and disclose the sources of top-level project
    --public-packages    force specified packages to be considered public
    --no-bytecode        skip bytecode generation and include source files as plain js
    --no-native-build    skip native addons build
    --no-signature       skip signature of the final executable on macos
    --no-dict            comma-separated list of packages names to ignore dictionaries. Use --no-dict * to disable all dictionaries
    -C, --compress       [default=None] compression algorithm = Brotli or GZip

  Examples:

  – Makes executables for Linux, macOS and Windows
    $ pkg index.js
  – Takes package.json from cwd and follows 'bin' entry
    $ pkg .
  – Makes executable for particular target machine
    $ pkg -t node16-win-arm64 index.js
  – Makes executables for target machines of your choice
    $ pkg -t node16-linux,node18-linux,node16-win index.js
  – Bakes '--expose-gc' and '--max-heap-size=34' into executable
    $ pkg --options "expose-gc,max-heap-size=34" index.js
  – Consider packageA and packageB to be public
    $ pkg --public-packages "packageA,packageB" index.js
  – Consider all packages to be public
    $ pkg --public-packages "*" index.js
  – Bakes '--expose-gc' into executable
    $ pkg --options expose-gc index.js
  – reduce size of the data packed inside the executable with GZip
    $ pkg --compress GZip index.js

The entrypoint of your project is a mandatory CLI argument. It may be:

  • Path to entry file. Suppose it is /path/app.js, then packaged app will work the same way as node /path/app.js
  • Path to package.json. Pkg will follow bin property of the specified package.json and use it as entry file.
  • Path to directory. Pkg will look for package.json in the specified directory. See above.

Targets

pkg can generate executables for several target machines at a time. You can specify a comma-separated list of targets via --targets option. A canonical target consists of 3 elements, separated by dashes, for example node18-macos-x64 or node14-linux-arm64:

  • nodeRange (node8), node10, node12, node14, node16 or latest
  • platform alpine, linux, linuxstatic, win, macos, (freebsd)
  • arch x64, arm64, (armv6, armv7)

(element) is unsupported, but you may try to compile yourself.

You may omit any element (and specify just node14 for example). The omitted elements will be taken from current platform or system-wide Node.js installation (its version and arch). There is also an alias host, that means that all 3 elements are taken from current platform/Node.js. By default targets are linux,macos,win for current Node.js version and arch.

If you want to generate executable for different architectures, note that by default pkg has to run the executable of the target arch to generate bytecodes:

  • Linux: configure binfmt with QEMU.
  • macOS: possible to build x64 on arm64 with Rosetta 2 but not opposite.
  • Windows: possible to build x64 on arm64 with x64 emulation but not opposite.
  • or, disable bytecode generation with --no-bytecode --public-packages "*" --public.

macos-arm64 is experimental. Be careful about the mandatory code signing requirement. The final executable has to be signed (ad-hoc signature is sufficient) with codesign utility of macOS (or ldid utility on Linux). Otherwise, the executable will be killed by kernel and the end-user has no way to permit it to run at all. pkg tries to ad-hoc sign the final executable. If necessary, you can replace this signature with your own trusted Apple Developer ID.

To be able to generate executables for all supported architectures and platforms, run pkg on a Linux host with binfmt (QEMU emulation) configured and ldid installed.

Config

During packaging process pkg parses your sources, detects calls to require, traverses the dependencies of your project and includes them into executable. In most cases you don't need to specify anything manually.

However your code may have require(variable) calls (so called non-literal argument to require) or use non-javascript files (for example views, css, images etc).

require('./build/' + cmd + '.js');
path.join(__dirname, 'views/' + viewName);

Such cases are not handled by pkg. So you must specify the files - scripts and assets - manually in pkg property of your package.json file.

  "pkg": {
    "scripts": "build/**/*.js",
    "assets": "views/**/*",
    "targets": [ "node14-linux-arm64" ],
    "outputPath": "dist"
  }

The above example will include everything in assets/ and every .js file in build/, build only for node14-linux-arm64, and place the executable inside dist/.

You may also specify arrays of globs:

    "assets": [ "assets/**/*", "images/**/*" ]

Just be sure to call pkg package.json or pkg . to make use of package.json configuration.

Scripts

scripts is a glob or list of globs. Files specified as scripts will be compiled using v8::ScriptCompiler and placed into executable without sources. They must conform to the JS standards of those Node.js versions you target (see Targets), i.e. be already transpiled.

Assets

assets is a glob or list of globs. Files specified as assets will be packaged into executable as raw content without modifications. Javascript files may also be specified as assets. Their sources will not be stripped as it improves execution performance of the files and simplifies debugging.

See also Detecting assets in source code and Snapshot filesystem.

Options

Node.js application can be called with runtime options (belonging to Node.js or V8). To list them type node --help or node --v8-options.

You can "bake" these runtime options into packaged application. The app will always run with the options turned on. Just remove -- from option name.

You can specify multiple options by joining them in a single string, comma (,) separated:

pkg app.js --options expose-gc
pkg app.js --options max_old_space_size=4096
pkg app.js --options max-old-space-size=1024,tls-min-v1.0,expose-gc

Output

You may specify --output if you create only one executable or --out-path to place executables for multiple targets.

Debug

Pass --debug to pkg to get a log of packaging process. If you have issues with some particular file (seems not packaged into executable), it may be useful to look through the log.

Bytecode (reproducibility)

By default, your source code is precompiled to v8 bytecode before being written to the output file. To disable this feature, pass --no-bytecode to pkg.

Why would you want to do this?

If you need a reproducible build process where your executable hashes (e.g. md5, sha1, sha256, etc.) are the same value between builds. Because compiling bytecode is not deterministic (see here or here) it results in executables with differing hashed values. Disabling bytecode compilation allows a given input to always have the same output.

Why would you NOT want to do this?

While compiling to bytecode does not make your source code 100% secure, it does add a small layer of security/privacy/obscurity to your source code. Turning off bytecode compilation causes the raw source code to be written directly to the executable file. If you're on *nix machine and would like an example, run pkg with the --no-bytecode flag, and use the GNU strings tool on the output. You then should be able to grep your source code.

Other considerations

Specifying --no-bytecode will fail if there are any packages in your project that aren't explicitly marked as public by the license in their package.json. By default, pkg will check the license of each package and make sure that stuff that isn't meant for the public will only be included as bytecode.

If you do require building pkg binaries for other architectures and/or depend on a package with a broken license in its package.json, you can override this behaviour by either explicitly whitelisting packages to be public using --public-packages "packageA,packageB" or setting all packages to public using --public-packages "*"

Build

pkg has so called "base binaries" - they are actually same node executables but with some patches applied. They are used as a base for every executable pkg creates. pkg downloads precompiled base binaries before packaging your application. If you prefer to compile base binaries from source instead of downloading them, you may pass --build option to pkg. First ensure your computer meets the requirements to compile original Node.js: BUILDING.md

See pkg-fetch for more info.

Compression

Pass --compress Brotli or --compress GZip to pkg to compress further the content of the files store in the exectable.

This option can reduce the size of the embedded file system by up to 60%.

The startup time of the application might be reduced slightly.

-C can be used as a shortcut for --compress .

Environment

Var Description
PKG_CACHE_PATH Used to specify a custom path for node binaries cache folder. Default is ~/.pkg-cache
PKG_IGNORE_TAG Allows to ignore additional folder created on PKG_CACHE_PATH matching pkg-fetch version
MAKE_JOB_COUNT Allow configuring number of processes used for compiling

Examples

# 1 - Using export
export PKG_CACHE_PATH=/my/cache
pkg app.js

# 2 - Passing it before the script
PKG_CACHE_PATH=/my/cache pkg app.js

Usage of packaged app

Command line call to packaged app ./app a b is equivalent to node app.js a b

Snapshot filesystem

During packaging process pkg collects project files and places them into executable. It is called a snapshot. At run time the packaged application has access to snapshot filesystem where all that files reside.

Packaged files have /snapshot/ prefix in their paths (or C:\snapshot\ in Windows). If you used pkg /path/app.js command line, then __filename value will be likely /snapshot/path/app.js at run time. __dirname will be /snapshot/path as well. Here is the comparison table of path-related values:

value with node packaged comments
__filename /project/app.js /snapshot/project/app.js
__dirname /project /snapshot/project
process.cwd() /project /deploy suppose the app is called ...
process.execPath /usr/bin/nodejs /deploy/app-x64 app-x64 and run in /deploy
process.argv[0] /usr/bin/nodejs /deploy/app-x64
process.argv[1] /project/app.js /snapshot/project/app.js
process.pkg.entrypoint undefined /snapshot/project/app.js
process.pkg.defaultEntrypoint undefined /snapshot/project/app.js
require.main.filename /project/app.js /snapshot/project/app.js

Hence, in order to make use of a file collected at packaging time (require a javascript file or serve an asset) you should take __filename, __dirname, process.pkg.defaultEntrypoint or require.main.filename as a base for your path calculations. For javascript files you can just require or require.resolve because they use current __dirname by default. For assets use path.join(__dirname, '../path/to/asset'). Learn more about path.join in Detecting assets in source code.

On the other hand, in order to access real file system at run time (pick up a user's external javascript plugin, json configuration or even get a list of user's directory) you should take process.cwd() or path.dirname(process.execPath).

Detecting assets in source code

When pkg encounters path.join(__dirname, '../path/to/asset'), it automatically packages the file specified as an asset. See Assets. Pay attention that path.join must have two arguments and the last one must be a string literal.

This way you may even avoid creating pkg config for your project.

Native addons

Native addons (.node files) use is supported. When pkg encounters a .node file in a require call, it will package this like an asset. In some cases (like with the bindings package), the module path is generated dynamicaly and pkg won't be able to detect it. In this case, you should add the .node file directly in the assets field in package.json.

The way Node.js requires native addon is different from a classic JS file. It needs to have a file on disk to load it, but pkg only generates one file. To circumvent this, pkg will create a temporary file on the disk. These files will stay on the disk after the process has exited and will be used again on the next process launch.

When a package, that contains a native module, is being installed, the native module is compiled against current system-wide Node.js version. Then, when you compile your project with pkg, pay attention to --target option. You should specify the same Node.js version as your system-wide Node.js to make compiled executable compatible with .node files.

Note that fully static Node binaries are not capable of loading native bindings, so you may not use Node bindings with linuxstatic.

API

const { exec } = require('pkg')

exec(args) takes an array of command line arguments and returns a promise. For example:

await exec(['app.js', '--target', 'host', '--output', 'app.exe']);
// do something with app.exe, run, test, upload, deploy, etc

Troubleshooting

Error: ENOENT: no such file or directory, uv_chdir

This error can be caused by deleting the directory the application is run from. Or, generally, deleting process.cwd() directory when the application is running.

Error: ERR_INSPECTOR_NOT_AVAILABLE

This error can be caused by using NODE_OPTIONS variable to force to run node with the debug mode enabled. Debugging options are disallowed , as pkg executables are usually used for production environments. If you do need to use inspector, you can build a debuggable Node.js yourself.

Error: require(...).internalModuleStat is not a function

This error can be caused by using NODE_OPTIONS variable with some bootstrap or node options causing conflicts with pkg. Some IDEs, such as VS Code, may add this env variable automatically.

You could check on Unix systems (Linux/macOS) in bash:

$ printenv | grep NODE

Advanced

exploring virtual file system embedded in debug mode

When you are using the --debug flag when building your executable, pkg add the ability to display the content of the virtual file system and the symlink table on the console, when the application starts, providing that the environement variable DEBUG_PKG is set. This feature can be useful to inspect if symlinks are correctly handled, and check that all the required files for your application are properly incorporated to the final executable.

$ pkg --debug app.js -o output
$ DEBUG_PKG=1 output

or

C:\> pkg --debug app.js -o output.exe
C:\> set DEBUG_PKG=1
C:\> output.exe

Note: make sure not to use --debug flag in production.