reg
is a package manager for native ES Modules. It's
built to enable dependency management for Universal JavaScript
(JavaScript that can run in the Browser and in Node.js w/o a compiler).
This library is highly experimental and still likely to break without notice. DO NOT USE THIS IN PRODUCTION.
Supporting Universal JavaScript is quite difficult as the Browser's
module system has a very unique set of constraints. In order to build
effective dependency management on-par with the features you'd expect
from npm
the approach reg
takes is radically different.
reg
statically links the dependency tree by hash reference, re-writing
the import statements of the output files (this is the only alteration to
your source file reg
does, it is not a compiler). A package registry
is then just a namespace that maps users, package names, and versions to
specific package hashes. You can then directly import the resulting package
references in a Browser or in Node.js with a special loader.
Note: since this is built for Universal JavaScript, you cannot use any
of the Node.js standard library or any packages from npm
since none
of this is available in Browsers without a compiler.
This command takes an input file, statically links the dependency tree, stages it in a local cache and executes the input file.
The same command can also take a package reference from the registry.
Publish an input file to a package to the registry.
Positionals:
filename Filename of script to run. Example `reg input.js`
semver Package version number.
Options:
--help Show help
--version Show version number
--token GitHub personal access token
Defaults to process.env.GHTOKEN || process.env.GITHUB_TOKEN
Note that this command requires a GitHub Token that only needs enough permissions to validate the user, no write or read access to any of your repositories are required.
All package names must be proceeded by the user's GitHub username. There are currently no top level packages.
There are several more commands that have been useful while developin this software. They may eventually be removed from the public API.
Run a local script file in reg
Commands:
cli.js stage <filename> Run the linker and stage the tree in
local cache
cli.js linker <filename> Run the static linker
cli.js info <name> Get info for named alias
cli.js cat <name> Print the file data for the named
alias
cli.js pkg-info <cid|name> Get package information
reg
implements a data model for its package data that is similar to
git
in many ways. The highlights of this data structure are:
- Optimized for offline, sync, and decentralization (just like git!)
- A module is only a single file (the Browser requires this) with the dependency tree attached.
- Every module gets a unique hash (like a git commit) which means cache de-duplication works across differing module names and version.
- Every file is chunked with an algorithm called Rabin which creates good block boundaries for diffing (this is what rsync uses). This gives us sub-file de-duplication in cache which is especially useful for de-duplicating file parts between versions.
This data structure also enables some important HTTP/2 features we need in order to be competitive with bundle performance.
- HTTP/2 Push for all of the dependencies required by a single import.
- If an old e-tag is presented for a module,
reg
can diff the two dependency trees and use HTTP/2 Push of only the assets that have changed.
For a more detailed look at the data structure you can read the schema.
Some visual examples follow: