@oazmi/superbuild
v0.1.3
Published
A wrapper on top of esbuild to provide you with extended esbuild plugin-api capabilities (hence a super set of esbuild).
Maintainers
Readme
@oazmi/superbuild
A wrapper on top of esbuild to provide you with extended esbuild Plugin-API capabilities (hence a super set of esbuild).
[!note] This project is free of dependencies (only using my
jsr:@oazmi/kitchensinkfor utility functionality), and bundles down to just45kbwith minification. so, fear not of carrying extra baggage by using this library.
General Usage
import esbuild from "npm:esbuild"
import { SuperBuild } from "jsr:@oazmi/superbuild" // or use "npm:@oazmi/superbuild".
const spbuild = new SuperBuild(esbuild)
const result = await spbuild.build({
entryPoints: ["./mod.ts"],
// ... other build options.
plugins: [{
name: "my-plugin",
setup: my_plugin_setup, // see the examples that follow.
}],
})Extended API Features
onLoad
The super-build wrapper lets you use any arbitrary string for the returned loader field,
and then have that loaded resource get captured by the onTransform stage's filter.
example:
import type { OnLoadArgs, SuperPluginBuild, SuperPluginSetup } from "jsr:@oazmi/superbuild"
const my_plugin_setup: SuperPluginSetup = (build: SuperPluginBuild) => {
// ...
build.onLoad({ filter: /\.html$/ }, async (args: OnLoadArgs) => {
const contents = await (await fetch("...")).text()
return {
contents: contents,
loader: "html", // you may use any arbitrary string for the loader.
}
})
// ...
}You also don't have to go through an onLoad for your custom-loader resources if you're just loading it through the filesystem;
All you have to do is declare it in the loader option of the bundle build:
import esbuild from "npm:esbuild"
import { SuperBuild } from "jsr:@oazmi/superbuild"
const spbuild = new SuperBuild(esbuild)
const result = await spbuild.build({
entryPoints: ["./mod.ts", "./meow.html", "./neko/nyaa.html", "./styles.scss"],
loader: {
".html": "html",
".scss": "sass-loader",
".css": "lightningcss",
},
// ... other build options.
})onTransform
The newly added build.onTransform function lets you capture loaded content,
transform them in any arbitrary way, and then specify additional arbitrary imports that need to be bundled.
[!caution] your import entity's
keymust be json serializable!
example:
import type { ImportEntity, OnTransformArgs, SuperPluginBuild, SuperPluginSetup } from "@oazmi/superbuild"
const my_plugin_setup: SuperPluginSetup = (build: SuperPluginBuild) => {
// ...
build.onTransform({ filter: /.*/, loader: "html" }, async (args: OnTransformArgs) => {
const contents = typeof args.contents === "string" ? args.contents : new TextDecoder().decode(args.contents)
const html_doc = new DOMParser().parseFromString(contents, "text/html"),
js_imports: ImportEntity[] = scanJsImports(html_doc), // just a hypothetical function.
css_imports: ImportEntity[] = scanCssImports(html_doc), // just a hypothetical function.
my_additional_import: ImportEntity = {
key: "external-ref-0",
path: "http://example.com/favicon.ico",
external: true,
}
return {
contents: contents,
// below, you must specify an esbuild-compatible loader.
// for an html file , `copy` makes the most sense.
loader: "copy",
// all non-external files specified in the imports will also get bundled.
imports: [...js_imports, ...css_imports, my_additional_import],
// you may also attach arbitrary data that'll get re-captured during the onEmit stage.
emitData: { timeStamp: Date.now(), copywrites: ["Seto frikking Kaiba"] },
}
})
// ...
}onEmit
Finally, the new build.onEmit function lets you capture output file contents after they have been bundled,
transform them in any arbitrary manner (such as re-incorporating user-made imports from the transform stage),
and finally renaming them (in topological dependency order).
example:
import type { ImportedEntity, OnEmitArgs, SuperPluginBuild, SuperPluginSetup } from "@oazmi/superbuild"
const my_plugin_setup: SuperPluginSetup = (build: SuperPluginBuild) => {
// ...
build.onEmit({
filter: /.*/,
inputs: [{
// filter based on the resolved path of the resource that is bundled into the output file.
filter: /.*/,
namespace: undefined, // capture all namespaces.
loader: "html", // loader based filter.
transformLoader: "copy", // on-transform's loader based filter.
}],
}, async (args: OnEmitArgs) => {
const contents = typeof args.contents === "string" ? args.contents : new TextDecoder().decode(args.contents)
const html_doc = new DOMParser().parseFromString(contents, "text/html")
// just a hypothetical way of re-incorporating the bundled imports.
args.imports.forEach((imported_entity: ImportedEntity) => {
const { key, outputPath, write, external, kind } = imported_entity
if (external || !write) return
console.assert(kind === "user-import")
html_doc.querySelectorAll(`#${key}`).forEach((dom_element) => {
dom_element.setAttribute("src", outputPath)
})
})
const new_output_path = args.outputPath.replace(/\.html$/, ".xhtml")
return {
contents: html_doc.documentElement.outerHTML,
path: new_output_path,
write: true,
}
})
// ...
}Other Notable Features
- Dependency free, and under
45kbwhen bundled with minification. - Cross-runtime compatible (deno, node, bun, and the web (only when
writeisfalse)). - You can rename output files during the emission stage, and the updated paths will be reflected in the dependent output files'
args.imports. This is because the emitted output files are always updated in their natural topological dependency order (in parallel).
Documentation and Links
- Documentation page: https://oazmi.github.io/superbuild.
- JSR release page: https://jsr.io/@oazmi/superbuild.
- NPM release page: https://www.npmjs.com/package/@oazmi/superbuild.
- Github repository: https://github.com/oazmi/superbuild.
Prior Art
I was using esbuild-extra for a while for static-site generation, and I found it it be incredibly useful.
However, I needed additional freedom in manipulating resources, dynamically adding resources for bundling,
while supporting code-splitting and resource-sharing in the additional bundled resources (i.e. where sub-builds do not qualify).
So, I wrote this library, heavily inspired by the said library's vision.
However, do note that the esbuild-extra's onTransform is fundamentally different from this library's onTransform hook.
In essence, the onEmit hook of this library resembles closer to esbuild-extra's onTransform,
because it takes place after esbuild has successfully bundled the resources.
Whereas the onTransform hook of this library lets you manipulate the loaded resources before they get passed over to esbuild for bundling.
Another cool set of libraries that also manipulate esbuild's transformation and emission mechanism is @chialab/esbuild-rna.
Though, I could never get it to run with deno (during its early node-compatibility days), so I never gave it a try.
Notes
a few things that still need to be taken care of:
- I still need to figure out how to make
sourcemaps possible for user-specified imports. - Also,
<stdin>isn't captured by my plugin right now, since esbuild process it directly (skipping the plugin API layer).
esbuild Metafile
pathname in keyof Metafile.inputscorresponds to the template${namespace}:${resolve_path_string}, unless thenamespaceisfile, in which case it simply becomes${resolve_path_string}.- esbuild's own path resolution format reads like:
"node_modules/@oazmi/kitchensink/esm/alias.js". in other words, it strips away the leading"./", to indicate that it is in the current working directory (orabsWorkingDir). Metafile.inputs[string].imports[number].pathdictates the${namespace}:${resolve_path_string}name of the imported resource.pathname in keyof Metafile.outputscorresponds to the template${outdir}/${filepath}, without any leading"./".pathname in keyof Metafile.outputs[string].inputscorresponds to thepathname in keyof Metafile.inputs, using the same${namespace}:${resolve_path_string}template.- a new
entryPointis created for each dynamic importm using the same${namespace}:${resolve_path_string}template, however, not allMetafile.outputs[string]have anentryPointproperty, especially if they were an intermediate stplitted dependency. Metafile.outputs[string].inputscan be an empty record in case the file is a split-up chunk that only re-exports other imports.
