Bundling support for Require.js #86
Comments
|
Its already been three months for this post, but in case it still matters to you:
If the config option
|
|
Hello Mattias,
thank you for your efforts to the project.
I’m currently set down the priority for bundling the javascript in my project, but this task will continue in the next 2 or 3 month. So i’m looking forward to test your new implementation. I invested much time in the past to study the code and find out, that there was no way to configure the bundling with requirejs.net correctly, even if require.js supports it. I hope your changes will unleash this autobundling-feature even in my case with a changed base path.
The other, in my point of view, design-error of requirejs.net is, that it writes always the whole configuration to each page. My Ecommerce-Application is a Multipage-Application, and it’s really ugly to see this extra-bytes on each page. Do u think there’s a chance to „exclude“ the configuration to a javascript-file and reference it, so the configuration can be loaded from the client cache? Maybe as part of the bundling-process?
Kind Regards
Holger
Von: Mattias Åslund [mailto:notifications@github.com]
Gesendet: Mittwoch, 22. Februar 2017 19:53
An: vtfuture/RequireJSDotNet <RequireJSDotNet@noreply.github.com>
Cc: Holger Rasch <privat@holgerrasch.de>; Author <author@noreply.github.com>
Betreff: Re: [vtfuture/RequireJSDotNet] Bundling support for Require.js (#86)
Its already been three months for this post, but in case it still matters to you:
1. I believe the sample project RequireJSNet.Examples illustrates this situation in url /home/entrypoint. The required entrypoint is relocated to ~/App/. It is also bundled by the autobundler to /bundles/auto/entrypoint.js (which may be a bit hard to realize).
2. When the compressor nuget package is installed it adds a build target in the project file which will bundle all dependencies according to RequireJS.json. It also creates RequireJS.overrides.json which contains all the information requirejs needs to use the bundles.
If the config option LoadOverrides = true is set in the Html.RenderRequireJsSetup` configuration, RequireJSDotNet will automatically include this information when requirejs loads. This makes it easy to debug without bundling and deploy to production with bundles.
3. I'm interested in this myself. If you reaffirm your need for this I will see if I can make it happen anytime soon.
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub <#86 (comment)> , or mute the thread <https://github.com/notifications/unsubscribe-auth/AWN9GmHxQ6x9h3ZdWvgmSRD79Qpnl4Rmks5rfIQPgaJpZM4K7ZRa> .
|
|
I too work professionally with a multi-page application. It is pretty huge and we just recently started moving to requirejs, which is why I’ve taken up this as a pet project.
My requirejs config is currently ~7KB and counting, so I’m experimenting a bit on moving the config to an external script file.
Right now I’m dabbling with an HttpModule that returns the script file from memory, much like how SignalR/hubs work, but I havn’t settled on an approach. Input is always appreciated.
My main reason for not moving this into the compression step is because the current code uses the current entrypoint for selecting the correct cached overrides. How this exactly works is actually a bit of a mystery to me right now. I’m just trying to let alone what I don’t fully grasp at the moment.
|
|
I have created a working alpha regarding offlining the configuration into its own "file". The code is found under aquamoth/requirejsdotnet, branch httpmodule. There is no documentation at the moment except the examples have been updated to use the new code. There is a call in global.asax to register a new httproutehandler and the supported configurations. Then the setup call in the shared layout views are simplified a lot. ´/home/index´ illustrates how the new call is used while still inlining the config. ´/home/complex´ offlines the config using the httphandler while ´/home/entrypoint´ hasnt changed at all and still uses the original way of declaring the config in the view layout. I'd really like any feedback on this before moving it into the main repo. Especially since there is one unfortunate breaking change; the call to determine locale now must get an HttpContext instead of an HtmlHelper... Stefan; I want to talk to you about how we move this branch forward too.. Gitter-users anybody?? |
|
Ok, I reviewed your implementation a little bit and want you give some feedback. Maybe we are loosing some focus on the sollution with this changes. We should remember that requirejs.net is a asp.net mvc wrapper around require.js to bring some very nice features:
what we don't like on the implementation:
Maybe I currently don't get the importance for the overrides, but if I would kick out Requirejs.Net and use require.js in it's native js-implementation, i would create two different configs, one including the bundles and the other without bundles for debugging. The compression-task would be done with r. Require.js won't support localizable resources, passing the options or the detection, if a JS-File exists for my action. On the other hand, the config file is a fixed ressource. No need for a distinct HttpHandler to resolve the configuration. This is the reason, why I don't agree with this solution. Yes, it's easy to render the output of the configuration just to a httpHandler instead of the Page. But there are dependencies from the rendered page to this rendered content, like the requireJsConfig-options or ressources. Because of this, each page would have it's own handler-call, and the config cannot be read from the client-cache. I think the configuration should be splitted to a clean require.config (and a second for overrides) and any page-dependent configuration should be rendered to the page (like the options, ressources..) or through a httphandler. |
|
I appreciate your input but maybe I dont fully grasp your critique; Is an httphandler intrinsically wrong in your opinion, or is it foremost that it currently creates different urls for different pages? I dont lightly break existing code, but what if we add a property "ignore-endpoint" that returns the same config-url for all pages with the same config? Just to be clear; the httphandler supports caching with both last-modified header and etag so the config gets cached on the clients. After all that was the main goal... I do agree the solution with an extra httphander adds a bit of complexity. Maybe we can compress to a file at compile time instead, if we get rid of the entrypoint requirement? |
|
In my opinion a HttpHandler should be used if non-static content that don't belong to the current (page) request should be transmitted to the client. There are some exceptions: The content has a different file-format than the main-request, uses a different infrastructure and is accessable in a restful kind of request, The same request should return the same response. This in my mind, i wouldn't use an http handler for content that can be served in a static way. I think the "general" configuration is such a piece of static content. It can be saved in some javascript file like require.js does it natively. So the question should not be "why not using an httphandler?", but "what are the benefits using an http-handler?". Ok, our goal is to exclude the configuration part from the page. But what is about the data, that is page dependent? Is it really false to include it to the page? It's harder to pass some information, that belongs to one request to a second one, e.g. if u want to move the pageOptions to the httpHandler-Request. I think a clean solution would introduce something like a proxy-call (to your http-handler) just for the configuration-settings. These configuration-settings should include the paths, shims, bundles, ... it has to return the bootstrapping-configuration for require.js. This makes sense, because we don't need a client-side part to read the configuration from json e.g. and bootstrap require.js. All other content should stay in the page (options, locale, require for the endpoint) because it is page dependent content. Making an distinct request for page-dependent content is for me something like an anti-pattern. |
Hello,
we are currently developing an ecommerce application where requireJS.Net is included. No I try to bundle my solution, but I'm struggle with 3 problems:
-> app
--> assets
---> controllers
---> js
---> libs
The reason for this structure is because some included libs like bootstrap contain javascript, css, fonts, aso. I don't wanna split this external libs. But it's not possible to override the baseUrl for the autoBundle-functionality.
The second and more import point is, that no bundle-config is passed to require.js. The autoBundle-functionality would build up some bundles, but require.js need the information what bundle to load if a module is required. If I understand it correctly, this is normally configured through the bundle-section in require.js. RequireJS.Net do not support this section.
My last point is that it would be great to have an option to build an js-file with the config, because if it's getting complex, it could be cached on client-side and is not included in each page.
The text was updated successfully, but these errors were encountered: