I want to use a lot of style config files for a project based on Webpack. Now I'm using variable for this:
loaders: [
...
, { test: /\.styl$/, loader: "style!css!postcss!stylus?paths[]=node_modules&paths[]=app&import=apps/" + appName + "/config.styl" }
...
Everything going fine, but the problem now is that the auto-update does not work when I edit the config.styl. Maybe there is a better way? Or how can I fix auto update in this case?
A couple of ideas to try out (migrated from comments):
string-replace-webpack-plugin could be used to generate imports dynamically - Unfortunately it looks like HMR doesn't pick it up. You can try invoking it as a preLoader
Set up resolve.alias for configuration. This avoids having to use replace plugin and feels like a cleaner alternative.
Related
My reactjs app consists of a bunch of typescript files with a clean separation of GUI and model. The webworker needs most of model files, so maybe half of all files. I could imagine loading the webworker from exactly the same URL as the app as the model does reference neither the GUI nor React nor other unavailable stuff (At least, it shouldn't, and if so, it'd easy to clean it up).
There seem to be some problems:
finding the correct javascript files
injecting proper start up code into them
and probably others I haven't thought about yet.
The communication to the webworker is not a problem as all I need is a single async call passing and receiving some simple data.
There may be more issues like e.g., https://github.com/microsoft/TypeScript/issues/20595.
Before I learnt what I really need, I tried e.g., ttps://www.npmjs.com/package/#koale/useworker, which is nice, but seems to be able to deal with plain javascript dependencies only.
Finding the correct javascript files
What I can see in index.html is
<script src="/myapp/static/js/bundle.js"></script>
<script src="/myapp/static/js/0.chunk.js"></script>
<script src="/myapp/static/js/main.chunk.js"></script>
<script src="/myapp/main.4e45e2b4b645351b7733.hot-update.js"></script>
I guess, I could live without hot updates, however the names of the other three files change in production to something like "/myapp/static/js/2.28cf00cf.chunk.js".
Injecting proper start up code into them
When the worker loads, it executes some webpack code generated code which most probably crashes it. I'd need to avoid it somehow.
The questions
Is this doable at all?
Does it make sense or is there a better approach?
For a seamless integration of worker code with main-thread code, I recommend using comlink-loader. For example, if you have a main.ts file and a thingy.worker.ts file, you could seamlessly load it as a worker by using TS imports:
// main.ts
import { getThing } from "./thingy.worker.ts"; // make sure the file name ends with .worker.ts
async function test() {
console.log(`the thingy is: ${await getThing()}`);
}
// thingy.worker.ts
export async function getThing() {
return 3;
}
You'll need to add it to your webpack config like this:
// webpack.config.js
module.exports = {
module: {
rules: [
{
test: /\.worker\.(js|ts)$/i,
use: [{
loader: 'comlink-loader',
options: {
singleton: true
}
}]
}
]
}
}
The best part is that your editor's intellisense will work across modules and type safety will not be compromised. More documentation is available here.
You need to compile a separate js file for your WebWorker. There is a React lib called react-webworker that makes this easier for you. It in turn uses WebPack’s worker-plugin.
I'm using grunt-uncss with my AngularJs app and I'd like to know if it's possible to specify a nested class in the options to be ignored.
For example, my css output when using Angular animations would be something like this:
.button.ng-enter {...}
And in the options I would have to specify it like this:
options: {
ignore: ['.button.ng-enter']
},
It's ok if i have just a few.. But in a large application I would have to specify it explicit for all the use cases.
Is there a way to declare only the .ng-enter and process it for all the use cases, even when nesting with other classes?
I understand that requirejs and browserify can load my files dependent on its current context, and that it is amazing. I would really prefer to use the #section sections syntax that the razor engine uses. Was just wondering if there is a way to implement this into a typescript / angularjs application.
for example
index.html
#renderSection scripts;
// which could turn into something like
<script data-render="scripts"></scripts>
// the app.run() could declare all the scripts that will be needed on every
// page view
view.html
<script ng-section-repeat="injected in injection"></script>
// the ng-section-repeat is basically taking all the items in the
// typescript constructor and then finding out which ones are needed for
// that view.
I like the idea injecting application file dependencies in the view , without a configuration file and all the added extras that comes with the loaders.
I just want to easily define what files are needed in the actual view and get them loaded, with angular's dependency injection handling the dependency itself.
If you are handling all your dependencies with $inject then , as far as i can tell, dependency is technically already setup in the controllers, all one would need, is to load this as it is called. Which could even eliminate the need for the #section scripts completely
Update:
What i have done to sort of replicate the module loaders is to just use gulp-concat and define the file order in my gulp.config.js and then pass it to the gulp-src before running $.concat .this allows me to have the files in the gulp steam , in dependent order . They are however loaded on the first load. With gulp-uglify the files are tiny ( its now at 566Kb with 16 external libraries loading in 69ms . To put that into perspective it takes 209ms to load one google font ).
I dont know maybe i am not understanding browserify correctly but i honestly struggle to see the need for it, its seems extremely convoluted for something so simple
It is possible using external modules and an injector to do what you asked for:
I just want to easily define what files are needed in the actual view
import {UserFactory} from 'models/userFactory';
import {UserValidator} from 'models/userValidator';
import {Inject} from 'angular2/di';
and get them loaded, with angular's dependency injection handling the dependency itself.
Note: My example uses angular 2.x because I less familiar with angular 1.x and I'm sure you can do something really similar...
class SomeComponent {
userName: string;
userRating: number;
rating: number;
constructor(
#Inject(UserFactory) UserFactory
#Inject(UserValidator) UserValidator
)
{
this.UserFactory = UserFactory;
this.UserValidator = UserValidator;
}
}
Then you can use Browserify to create a bundle.js file that can be executed in a web browser.
I love it when multiple technologies come together to produce a doozy...
The following AngularJS template squawks an error in the IDE ("can't resolve file"). I find the inspection wildly convenient and I don't simply want to turn it off.
/my_project/www/templates/logo.html
...
<img src="img/logo.png"/> <<< file at /my_project/www/img/logo.png
...
Question:
How can we allow an IDE like IntelliJ IDEA or WebStorm to play nice with Ionic/AngularJS/Cordova in this situation?
NOTE: I cannot simply mark the www folder as a "resources root" and use absolute references because ionic needs relative refs...
Or does it? Is there a way to fix this on the cordova side of things to allow absolute refs? i.e., so it doesn't break when deploying to Android (because we need the prefix file://android_asset/www/)
Inspired by this answer, I ended up creating a recursive search/replace script in the build process. I made a cordova hook for "after_prepare" and used node with the replace package. Now I can use absolute refs and get the full benefit of the IDE... and then at build-time they get converted to relative.
Here is a sample hook file to get you started. Don't forget to add refs for css files or things like templateUrl in your app.js if you're using angular/ionic. And of course don't forget to modify the platform specifics to your needs.
#!/usr/bin/env node
var replace = require("replace");
var wwwDir = process.argv[2] + '\\platforms\\android\\assets\\www';
// convert all src and href tags to relative in index.html + components\*
replace({
regex: '(src|href)=([\'"])/+',
replacement: '$1=$2',
paths: [
wwwDir + '\\components',
wwwDir + '\\index.html'
],
recursive: true,
silent: false
});
I'm working in a Sencha application.
I've created a couple of Utilities classes as singleton components (helpers, services, etc).
I'm using alternateClassName to have a shorter name for those classes.
It works perfect, but stop working after compiling for production.
I don't know why, and need help to get this working!
Looks to the following example:
I've created a demo application using sencha cmd for simplicity. The application is "Demo".
The whole application is as default, but I've added a util folder inside app, with a single file Helper.js. This is the code:
Ext.define('Demo.util.Helper', {
singleton: true,
alternateClassName: 'Helper',
test: function () {
alert('It works !');
}
});
Then, I just need to update app.js to require this new file, and update the launch function to call test method after add the main view. So here is the code to use in app.js:
requires: [
'Ext.MessageBox',
'Demo.util.Helper'
],
The launch function:
launch: function () {
// Destroy the #appLoadingIndicator element
Ext.fly('appLoadingIndicator').destroy();
// Initialize the main view
Ext.Viewport.add(Ext.create('Demo.view.Main'));
Helper.test();
},
Now, if I try the example, after load the app, an alert msg is shown successfully.
But after compile it using sencha cmd
sencha app build production
I get this error:
I know the problem is with alternate class name, because if I use the full name (instead of alternate class name), it works anyway. But I want to use alternate class name, otherwise it doesn't make any sense.
Any idea on what's wrong with compiled version ?
TIA!
Milton
After some time, we realized that Sencha has a bug when compiles singleton classes for production (works on testing also).
The solution was to remove the singleton flag, and create application variable for all of the singleton classes, in the launch method.
For example:
Demo.Helper = Ext.create('Helper');
Hope this help!
UPDATE
Last version of Sencha Cmd is full of freaking bugs!
I found a lot of other issues after fixing this ones, and finally, I found this link http://www.sencha.com/forum/showthread.php?288972-MyAppName.app-not-working-on-build-production&p=1064635