This is a shim HTTP server for directly running your CommonJS (or ES6) modules for development and testing (without a bundling step, and with the compilation/transformation integrated in the server shim).
Warnings:
-
The server basically exposes your whole filesystem over HTTP. It binds to
"localhost"
by default, but if you bind it to something else or proxy in some uncautious way, you are putting yourself at risk. -
This does everything syncronously, both on client and server.
-
This is a hack that I am using for my own development, not generally-useful, supported software. You might be able to make use of it, but if you run into a problem, you should try to debug it yourself, not ask me for help.
-
The code is likely to break if you do something that I'm not doing. Pull requests welcome!
You run the server for a given directory...
moduleserve demo/ --port 8080
It will start up an HTTP server on the given port, serving the content
of the demo
directory statically. In addition, it exposes a URL
/moduleserve/load.js
, which you use to load your main module:
<script src="/moduleserve/load.js" data-module="./mymodule"></script>
That will pull in the client-side scaffolding and look for
./mymodule
, resolved relative to the directory that the server is
running on. You can add a data-require
attribute to the tag to have
it set a global require
variable, which you can use to load modules
from the console.
That module is loaded as a
CommonJS module, and may
use require
with the regular node.js conventions (implicit .js
or
/index.js
, searching node_modules
, etc). The intention is that,
contrary to systems like JSPM, you don't have to set up your project
to please this tool, but you can just directly use it on an npm-style
codebase.
You can pass moduleserve
a --transform
option, which should point
at a node module that defines a transformer. Such a transformer is
called on a file before it is served. It should export a transform
function that takes (filename, text)
parameter and returns the
transformed text. If it also exports an init
function, that is
called once with the target directory of the server.
If you pass --transform babel
, you get a built-in transformer that
loads babel
in the context of the target directory (i.e. you have to
have it installed locally there, or globally) that you're running the
tool on, and uses it to transform any files whose path does not
contain node_modules/
.
If you pass --spa
, every file path where no file can be found automatically will fall back to serving the index.html
. This option allows you to implement routing for Single Page Applications.
The client is going to make a request for every single module file, so for a bigger project the initial load is bound to be slow, especially if you're using an expensive transform. The server then caches these and sends 302 responses whenever possible, so assuming localhost-level latency, subsequent loads should be faster.
This code is open-source under an MIT license. If you want to contribute, create pull requests on GitHub.