Better way to use require fallback?

[vidartf]

See discussion in jupyter-widgets/pythreejs#171 .

[jasongrout]

@maartenbreddels - can you review this one, as you were involved in the issue over in pythreejs?

[jasongrout]

This can also go into a patch release of html manager, so we don't have to rush it for 7.2.

[maartenbreddels]

Let me check this one, I think the idea is solid.

[maartenbreddels]

There is one issue, but even when I fixed that and try
require_1.toUrl(moduleName) which gives good output, it does not lead to a network request.
Isn't a better solution for jupyter-threejs to provide an anonymous module?

[maartenbreddels]

this will add a key 'moduleName', I think you need to do sth like

            let config = {paths: {}}
            config.paths[moduleName] = moduleNameToCDNUrl(moduleName, moduleVersion)

You should be able to put [] around moduleName, but the resulting js was without that, so I wouldn't rely on that.

[maartenbreddels]

Ok, the issue is that require already 'knows' 'jupyter-threejs' failed, since it requested it before doing the CDN request (it queried for ./jupyter-threejs.js), and refuses to do so again. This can be fixed using require.undef('jupyter-threejs'), say in requireLoader, after the catch.

[maartenbreddels]

Didn't work, you need to take off the .js part, sth like this:

            var url = moduleNameToCDNUrl(moduleName, moduleVersion)
            conf.paths[moduleName] = url.substring(0, url.length-3);

[vidartf]

Since the function is not exported, I just removed the .js part from the return value.

[vidartf]

This works for me locally now.

[maartenbreddels]

yes works here too.