Middleware for Koa servers that resolves Node package specifiers in standard JS modules to relative paths for use on the web.
The following import uses a bare module specifier, which won't currently load natively in browsers (until import maps are available):
import{foo}from"stuff";
koa-node-resolve solves this problem by resolving stuff using the same rules as Node require(), and transforming the import specifier to a path that can be loaded natively by any browser that supports standard JS modules:
import{foo}from"./node_modules/stuff/index.js";
Because this is middleware, you can use it in a simple static file server as well as a proxy server sitting in front of a test server such as the one karma starts up. (See karma testing setup below.)
Note: HTML and JavaScript are parsed on every request for those content-types, it is intended for use in development context to facilitate build-free testing/iteration as opposed to in a high volume production web server.
Installation
$ npm install --save koa-node-resolve
Usage
Create your own mini-development server in file ./dev-server.js. This one depends on koa and koa-static, so you'll need to npm install --save-dev koa koa-static for your project to use it.
Now you can serve up your web assets and Node package specifiers will be transformed on request.
Configuration
nodeResolve(options={})
Options
root the on-disk directory that maps to the served root URL, used to resolve module specifiers on the filesystem. In most cases this should match the root directory configured in your downstream static file server middleware.
logger an alternative logger to use (console is the default). The logger will receive error() to record exceptions during parsing/transforming of JavaScript modules, warn() when a specifier is unresolvable, info() to report URLs with transformed content, debug() to report all Node module specifier resolutions. All log messages are prefixed with [koa-node-resolve]. To disable all logging, provide false.
logLevel (defaults to warn) sets the minimum level of severity for an event to be logged. Options in order of severity: debug, info, warn, error.
htmlParser function to convert HTML source to a Parse5.DefaultTreeNode. The default implementation is equivalent to:
The most common reason to provide your own JavaScript parser function is to enable the middleware to process content which has syntax requiring babel syntax plugins that your project is making use of, such as decorators, dynamic imports or import meta etc.
jsSerializer function to generate output string from AST of Babel.Node. The default implementation is equivalent to:
The @babel/generator documentation shows many options available for tweaking output. Maybe you want to generate source maps and embed them in the response; you could do that here.
Karma Testing Setup
In a karma setup, your karma.conf.js file could create the Koa server before exporting the config. The Koa server uses the koa-proxy package (therefore npm install --save-dev koa-proxy) in between the browser and the Karma server, transforming all the Node package specifiers encountered in documents located under the base/ URL namespace, which is a special Karma behavior for partitioning the package resources under test from Karma support resources.
In this setup, the Koa proxy server that runs the Node resolution middleware will be on port 9877 and the Karma server will be on port 9876, so be sure to open up http://127.0.0.1:9877 in your browser rather than http://127.0.0.1:9876. The upstreamProxy configuration block tells Karma, when it launches browsers, to points them to the Koa app instead of directly to the Karma server.
Note also that in this configuration its important to tell Karma that the test files are modules and to serve those up, but to list the other files, like the ones in node_modules as available but not "included" (i.e. Karma can serve them by request, but shouldn't add inline dependencies on them when generating its "context" HTML).
请发表评论