We would like to know the feasibility of WebPack module federation build system in Ember JS App for invoking federated code from Ember App to connect with other front end codebase developed in React JS or other SPA.
Basically, we would to implement Micro-frontend using module federation approach where multiple frontends are already developed on Ember js and reactJS. In addition to that, our usecase depends on deeplinking, datasharing between these microfrontends.
Based on our understanding EmberJS by default uses broccoli and doesn't have support for webpack, Which is essential for Module Federation approach to accomplish Micro Frontend.
Looking forward for some pointers here
Have you seen this vite demo: https://github.com/lifeart/demo-ember-vite -- given that this is possible, however you "normally do module federation in webpack", is how you'd do it in ember.
you can do anything modern you can imagine (if you give up the older parts of the ecosystem) -- for example, any addon not in the v2 format likely won't work with the demo'd approach.
Broccoli isn't needed for EmberJS, it's just what happens to be the current default build system, and Embroider is 98% compatibility, with that 2% being the easy "just use webpack" parts. Embroider is pretty good for new apps if you don't use ember-data (newer than 4.7 at the time of writing), but that's the main purpose of embroider, bring the whole rest of the ecosystem to the modern era without folks needing to re-write their apps from scratch, and have to re-solve common problems.
Related
I have a 'core' Ember application that needs to be able to be extended by 'child' Ember apps. In Ember 0.10, this was achieved by heavily modding grunt tasks, but Ember 2 appears to have a possible workflow for this built in.
A super high level summary of my current (and target) setup:
core-application ('core')
contains shared business logic across All apps + templates and components
plugins
shared templates and logic that can be reused across apps (but not needed by all)
application
is composed of elements from core-application, plugins, + any app specific code. a note that routes should be able to be 'pulled in' from 'core'
In the current Ember 0.10 app structure, this has worked by modifying grunt tasks to build the apps in a quick, fairly fool-proof way.
Now, in Ember 2, it appears that this sort of pathway for app development is provided by using addons and blueprints. I suspect my 'core' app should become a 'blueprint' and plugins could be either an 'addon' OR 'blueprint' based on what is required by them. I'm writing proof of concept code now, but I have the following questions:
what does the --blueprint flag for the ember addon command do? I see that it essentially generates an app structure, but I don't see any realy documentation regarding where to go from there. This appears to to be what I want to use for my 'core' app, but the documentation is lacking here.
If the above --blueprint flag isn't what I want for this kind of set up, is there a better approach I should be considering?
Any other info regarding the above that folk with greater Ember 2 + ember-cli experience than I have can share on this would be hugely helpful.
Thanks in advance for any all feedback.
I found my answer by digging around existing Ember community addons.
The ember admin project seems to outline the structure AND consumption of an Ember addon that essentially creates an Ember app complete with routes and overridable/extendable elements.
The Host application then 'mounts' the admin addon by importing the admin addon's routes to the host application's routes and BOOM things work as expected. I've been able to write POC code to prove this concept works for my needs.
I have a single page app built with emberjs with an a rails backend. Is it a common pattern to build an admin interface on rails serverside on a subdomain. What is the right approach for this?
Your question is vague but I will try to answer it my best. I have done this with a Node and a Go backend combined with Ember.js.
No, there is technically nothing to prevent you from doing a single page application for an admin interface.
Rails is a good choice for this, and generally you should stick with the backend framework/language you and your team master the most.
As for what would be the right approach, there is no magic recipes. Document your code, write test and follow best practices for the tools you are using.
One key element though will be the communication between your frontend and your backend. Ember chose to follow the JSON API specification (http://jsonapi.org/) and comes out of the box with an adapter to talk with these kind of API. Using such adapter will help you save a lot of time.
Here are some implementation of JSON API for Ruby : http://jsonapi.org/implementations/#server-libraries-ruby
One more thing about your frontend code structure. You haven't say how big your app will be. If it gets big, you may want to take the pod approach in ember-cli (http://ember-cli.com/user-guide/#pod-structure). It basically changes the structure of the code so it's easier to maintain a lot of files.
I've downloaded Ember.js ver 1.13.13 for a test drive.
With other js frameworks, I am able to run from a file system. Does Ember require a server? I could not run directly from a file system. I did find some old tutorials that allows this. Is this a new thing?
You are using Ember-CLI which requires running ember serve in order to view your ember app. Ember-CLI uses conventions so that it knows where to locate the files that compose your ember app. As Ember-CLI locates your files, it knows how to combine them in a manner that ultimately results in the single JavaScript file that is executed in your browser. In theory you could use the globals style of development-which is the style reflected in the 'old tutorials' that you reference-and run the app directly without using any sort of "server." But, I don't recommend that. Learning Ember-CLI is useful as it is the preferred method of development moving forward. And, in my opinion, gives you a number of features that allow you to more quickly prototype apps. You can read more about that in the link I provided to the Ember-CLI website.
I am a bit confused regarding the ember websites "emberjs.com" and "ember-cli.com". Isn't ember-cli now a part of emberjs and documented at "emberjs.com"? If so, why have a different website for ember-cli? Also, why do the sites differ regarding the versions of the prerequisite JS frameworks? For example, the emberjs.com getting starting page says to use Node.js 0.12 or higher while the ember-cli.com site says to use the latest stable version of Node (version 4.0.x).
the emberjs.com getting starting page says to use Node.js 0.12 or
higher while the ember-cli.com site says to use the latest stable
version of Node (version 4.0.x).
Seems like emberjs.com is out of sync, you can create issue on emberjs.com website repository on GitHub or hope that it's already created and it will be resolved shortly. Also, there was as issue with bufferutil in Ember CLI + node.js 4.0, so they might wait until it's resolved, but for some reason it's recommended on ember-cli.com.
Isn't ember-cli now a part of emberjs and documented at "emberjs.com"?
It isn't part of Ember.js. It's part of Ember.js ecosystem. It's recommended tool to work with library and that's why examples in guides assume that you're using Ember CLI. But, you could also use Ember.js without Ember CLI by loading jQuery, Ember.js, Ember.js template compiler and working on globals.
Ember.js documents everything about library itself, how logic behind library works, core concepts. You have components, routing, controllers, models explained there, but you won't find information there how to add CoffeeScript to your project.
Ember CLI website documents everything about Ember CLI, which is tool that provides:
asset pipeline
a strong conventional project structure
powerful addon system for extension.
You can see in user-guide that it documents possible build configurations, how to make SASS or CoffeeScript working in your environment etc.
I am working on an Ember-Rails app. I have used Yeoman previously for building non-ember-rails apps and js plugins and I would love to be able to realize the benefits of Yeoman (especially Grunt's livereload) when working on my Ember-Rails projects. However, I am unsure as to whether Ember, Rails, and Yeoman are fully compatible and whether they overlap in their roles and responsibilities. For example:
Dependency Management
- I understand Bower is used for dependency management. Does Bower affect how assets are loaded through the rails asset pipeline? What are the advantages/disadvantages of loading dependencies through Bower instead of using Rails gems?
Livereload
- Does livereload function with a single-page app (like one built in Ember) in the same way it does with a multi-page app (for example, an html site that doesn't use a js framework)? Are individual models/views/controllers reloaded or does the whole app reload and/or recompile through the Rails asset pipeline?
Existing Project
- Are there pitfalls when integrating Yeoman into an existing Ember-Rails app? We're running Ember 1.3.0-beta and Ember Data 1.0.0-beta on our production. If you have experience with up-to-date Ember builds, are Ember-auth and Ember-data compatible with Yeoman?
If anyone has experience combining Ember, Rails, and Yeoman, or if you understand how the frontend and backend would compare with such a stack please share your thoughts! Would you recommend just integrating part of the Yeoman setup (e.g. Grunt) with an Ember-Rails app instead of the whole of Yeoman?
Thanks.
I'm not Rails&Ember guy, but here are my general thoughts about feasibility of using Yeoman:
Not using Yeoman: Feasibility
From my experience in combining Yeoman & Django, I must say that it starts to pay off only in medium-sized or bigger projects.
In smaller ones, particularly with tight deadlines and not much attention paid to the code quality & tech. solutions used (like Univ. projects), you'll be probably better off sticking with bare Rails (downloading JS libraries manually and committing them accordingly to the Rails project structure).
The reason is simple: It might be really time-consuming to fine tune both full-stack framework (Rails) with frontend-framework (Yeoman).
Particularly if Rails is driven by CoC principle.
It might seem to work after some setup, but as the project evolves, you will spot further obstacles and you'll have to tamper waaaaaaay more.
Cons & things I consider not worth this time investment:
Livereload
I like it very much, I was amazed with it at first, but after some time I see that I don't spend that much time editing HTML&CSS in IDE and watching static page on another screen automatically refresh. In majority of cases I still need to do Alt + tab and trigger some action, perform some click, so whether I add one Ctrl+R hit in-between doesn't do any difference.
In some cases you'll be better off playing around with Local Folders Mappings (Chrome Dev Tools) or web proxy (e.g.: Fiddler).
There are cases where Livereload does a brilliant job though, such as for instance not needing to perform full reload if you edited only CSS.
As for your questions:
For me if it detects changes in JS it reloads whole page. But maybe it's because I'm using JetBrains IDEs (filesystem cache'ing) and CoffeScript (compilation to JS).
Yeoman is best suited for SPAs. Would it be acceptable for you to make it a SPA, not a round trip app ?
Bower - Attempt to provide dependency management for github projects
What bower does for your app is basically downloading stuff from github. No rocket science here.
If the structure of downloaded thing is non-standard, Bower/RequireJS/Grunt-bower-install has no bloody idea what to do with it next, i.e. how to inject everything so that you won't get errors. For most popular libraries bower just works, for highly customized ones you will end up injecting downloaded stuff manually.
The ones to blame are people who don't package useful github projects properly.
Furthermore I heard of workflows where people commit bower_components due to problems with bower, never experienced personally though, perhaps the issues were fixed. If so possibility of not committing 3rd libraries to source code is definitely an advantage.
Note: As I pointed out above, it's how bower helps you developing your app, but bower has become somewhat a standard in the frontend community, for instance http://ngmodules.org/ is build on top of that, so it's a important tool.
Can Rails assembly pipeline fetch arbitrary github project ? Yes.
Can it inject library references to your html ? Duno, I'm not Rails guy.
Generators
Some could do a beautiful job, such as configuring whole heroku-related stuff for you. Too bad generator-heroku doesn't work as expected (tried it sth like 2 months ago). Same was true for travis generator. In this case fix was easy, but see the next point.
As for Angular generators (I'm Angular guy, not Ember guy, sorry :-) ) - it just adds 2 files and includes them in index.html, furthermore if you are using not so straight-forward syntax for creating JS framework related stuff (sample: Angular-related stuff in coffee) generator will most probably not know about it. Ok, you can submit a patch to generator, but then AngularJS team decides to change the syntax a little bit in the next release - you get the idea ? - again, see the next point.
What is more, if your project uses layout in which code is structured by feature/module (e.g. admin module, profile settings module, ...), not by type (directives, controllers, ...) framework-specific generators won't work.
Stability
Karma is evolving rapidly, so does Angular, Angular-UI and loads of fronted tools, frameworks.
It's really difficult for Yeoman to keep up with most recent changes, although they do a nice job here.
Employing Yeoman: Being cutting-edge
Yeoman provides some really cool stuff such as:
Linting
Compiling coffeescript, SASS/SCSS, etc. on the fly
CDNifying
Really useful when you want to have libraries downloaded locally in case of developing offline and still benefit from pros of CDN.
Without Grunt you would have ended up writing scripts parsing your HTMLs
Automated JS/CSS minification
Grunt does it for you. You only have to configure it properly.
Encouraging separation and low coupling of your frontend and backend
As for your question: I doubt that you'll be able to manage Rails part solely with Yeoman.
A good recipe for a web app in Yeoman & sth is: https://stackoverflow.com/a/19425461/1432478.
That's for Django, adapt it accordingly to Rails.
My opinion
In majority of small projects based purely on Spring MVC, Rails, Django, etc. you simply don't perform any of things done by Grunt (hence there's less time expenditure for setup).
There are cases when producing technologically advanced frontend is a must.
e.g.: Most recently I discovered that my bank account management system doesn't minify JSes. Even funny comments are left as they were. I didn't perform any rushed money withdrawal, but I hope they at least have server-side validation. :-)
Yeoman is a really good combination of solutions facilitating all those chores.
Using only Grunt? IMHO No. Yeoman is Grunt when it comes to app assembly. You get Livereload and other stuff for free.
If:
you're lucky and found sample config that seems to work for the web framework of your choice or managed to configure one on your own
don't have deadlines to hold
then you should probably give it a try.
If you will work on multiple projects, chore of configuring it once might pay off by copy-pasting config to future projects.
But keep in mind that frontend technologies are rapidly evolving, that's why tons of samples from the net simply don't work out of the box.
Further reading :)
http://blog.tfnico.com/2013/07/considerations-for-javascript-in-modern.html
Java world here, but I hope some concepts will be useful:
http://addyosmani.com/blog/making-maven-grunt/
Javascript web app and Java server, build all in Maven or use Grunt for web app?