This package provides a way to package npm
packages that depend on the wasi
imports to be built using ComponentizeJS
. This repository includes a library and two node executable (knitwit
and knitwit-postinstall
).
This configuration file plays a central role in defining how the knitwit
tool interacts with your project. The file specifies the packages, their corresponding wit paths, and the world names necessary for building with wasi
imports. This file is edited by the postinstall executable and read by the library. The schema for this file is defined using yup for validation and looks like this:
interface PackageConfig {
witPath: string;
world: string;
}
interface ProjectConfig {
witPaths?: string[];
worlds?: string[];
}
interface KnitWitConfig {
version: number;
project: ProjectConfig;
packages: Record<string, PackageConfig>;
}
-
version
- It is the version which describes the format of the config file -
project
- This contains the project specificwitPaths
andworlds
-
packages
- This field contains information about the different npm packages installed.
Here's an example of what a knitwit.json
file might look like:
{
"version": 1,
"project": {},
"packages": {
"example-package-1": {
"witPath": "path/to/example1.wit",
"world": "exampleWorld1"
},
"example-package-2": {
"witPath": "path/to/example2.wit",
"world": "exampleWorld2"
}
}
}
It can be found in bin/knitwit.mjs
. This is used to parse the knitwit.json
to get the combined wit
output. This needs to be run after a npm install
which may install packages that may have dependencies on wasi
. It also needs to be run after any changes in the project
field. This can be added to the postinstall
script in package.json
to automate updating based on package installs.
It can be found in bin/knitwit-postinstall.js
, this can be used in the postinstall scripts of packages that use wasi
imports to update the knitwit.json
. The executable uses configuration provided in the package.json
to update knitwit.json
.
The following field needs to be added to the config
key of the package.json
:
-
knitwit.witPath
- The relative path to thewit
file/folder from themain
entrypoint of the package. -
knitwit.world
- The name of the world that contains all the names required for the package.
An example of how the configuration will look is as below:
{
...
"config": {
"knitwit": {
"witPath": "<relative path to wit from entrypoint>",
"world": "<world with all imports>"
}
}
...
}
The executable needs to be added to the postinstall
script after adding the @fermyon/knitwit
package to the dependency list. The post install script takes in 2 arguments for --wit-path
and --world
optionally which it uses to update the configuration. If the values are not provided, it defaults to attempting to parse it of the package.json
{
...
"scripts": {
"postinstall": "knitwit-postinstall --wit-path <path_to_wit> --world <default_world>"
}
...
}
Note: The postinstall script should not include npx
as that will alter the environment variables which the executable relies on to run in the context of the consumer.
The @fermyon/knitwit
package provides a function knitWit to parse knitwit.json
, merging packages and their required worlds into a unified output package.
The library exposes one function knitWit
which has the following signature:
interface knitWitOptions {
witPaths?: string[];
worlds?: string[];
outputWorld?: string;
outputPackage?: string;
outDir?: string;
}
export declare function knitWit(opts?: knitWitOptions, ignoreConfigFile?: boolean): Promise<void>;
-
witPaths
andworlds
in opts are concatenated with those parsed fromknitwit.json
. - Additional options in opts include:
-
outputWorld
: Name of the world in the generated wit package (default: "combined"). -
outputPackage
: Package specifier in the generated wit package (default: "local:combined"). -
outDir
: Directory where the generated wit package is saved (default: "combined-wit").
-
To ignore parsing the config file, set ignoreConfigFile to true.
import { knitWit } from '@fermyon/knitwit';
knitWit({ worlds: ["additionalWorld1", "additionalWorld2"] });
The output generated by the above can then be used as input to ComponentizeJS
.