Require.js best practise when loading different scripts depending on local conditions - requirejs

For the sake of argument, say I want to load Zepto by default, but use jQuery instead for IE (all versions).
What would be a sensible way to do this when using Require.js?

There are two approaches: "proper" but long and "sleight of hand" but short.
"Proper" but long:
require.config({
paths: {
jquery:'path/to/jquery'
, zepto: 'path/to/zepto'
}
})
var iNeed = []
if (!('__proto__' in {})) {
// This is IE
iNeed.push('jquery')
} else {
// Everything else
iNeed.push('zepto')
}
require(iNeed, callback)
"Sleight of hand" but short:
var AMDConfig = {
paths: {
jquery:'path/to/zepto'
}
}
if (!('__proto__' in {})) {
// This is IE
AMDConfig.paths.jquery = 'path/to/jquery'
}
require.config(AMDConfig)
require(['jquery'], callback)
The reason the "sleight of hand" is not "proper" is that you are masking the real nature of what stands behind "jquery" As you grow your app, some jQuery plugins may come and not work over zepto, but it will not be immediately clear what the issue is.
The "proper" solution is also a problem in one respect - if you define the requirements array dynamically, the build tool like r.js will not be able to find other dependencies you put there.
Your pick..

Related

Order of keys in package.json exports

I believe I understand the basic functioning of the exports key in package.json files:
// package.json
{
"exports": {
".": {
// used by typescript
"types": "./file_with_type_defs.d.ts",
// used by ESM resolution
"import": "./file_to_import.mjs",
// used by CJS resolution
"require": "./file_to_require.cjs",
// used by ...others?
"default": "./file_one_more.js"
}
}
}
Question: Does the order of the "types", "import", "require", and "default" keys matter? Normally I would think no way, since JSON object keys are unordered. From json.org:
An object is an unordered set of name/value pairs. An object begins with { ...
But Typescript documentation says that "types" must come first:
Entry-point for TypeScript resolution - must occur first!
"types": "./types/index.d.ts"
and the NodeJS documentation says "default" should come last
"default" - the generic fallback that always matches. Can be a CommonJS or ES module file. This condition should always come last.
So... Does the order of the export keys matter? If not, what do the NodeJS and Typescript documentation mean when they talk about "first" and "last"?
Having swapped the order of "types" with other keys, its order seems not to matter.
Webpack, which also uses the export key explains this field as follows:
Notes about ordering
In an object where each key is a condition, order of properties is significant. Conditions are handled in the order they are specified.
Rather than see this as a plain object, consider it being like this if-else case:
let file;
if (platform_supports('types')) {
file = "./file_with_type_defs.d.ts";
} else if (platform_supports('import')) {
file = "./file_to_import.mjs";
} else if (platform_supports('require')) {
file = "./file_to_require.cjs";
} else if (true) { // default
file = "./file_one_more.js";
}
If you were to swap the order, it might be like this:
let file;
if (true) { // default
file = "./file_one_more.js";
} else if (platform_supports('types')) {
file = "./file_with_type_defs.d.ts";
} else if (platform_supports('import')) {
file = "./file_to_import.mjs";
} else if (platform_supports('require')) {
file = "./file_to_require.cjs";
}
Even though Typescript understands .d.ts files it would use file_one_more.js since it matches first.
I have tried swapping the order of "types" with other keys. It seems not to matter.
It might be that Typescript prioritizes the types condition over others. However, I'd stick with what they advise—"must occur first" is not "ought to occur first", after all.
As a side-note, historically JavaScript object keys are unordered / the order is not guaranteed. In practice, browsers did preserve the key order and this behavior was standardized in ES2015: non-integer keys are preserved in insertion order.
The JSON standard doesn't make the same promise, as there are many implementations of JSON decoders in different languages and it the predates ES2015 standard.
JSON objects are "unordered set" as "key do not have a particular order(i.e. never have to be sorted)", not as "do not have an order"
Parsed JS object DOES have an order of keys matching the orger in JSON
Depending on if the code uses for (let k in json) or if (json.types) the behaviour may be different
So by whatever reason it's recommended to order keys in the way some parsers may expect
In your case I'd recommend you to try swapping them, seeing that nothing happens, and swapping them back

Changing anyMatch default for Filter.JS in ExtJS for MultiSelect search

I have a multiselect bound to a store in which I implemented use of anyMatch: true to allow for True to allow any match - no regex start/end line anchors will be added (as per the comment in Filter.js). My problem is that I need to implement this as per the answer to multiselect-search-whole-string, in particular the solution provided in this fiddle https://fiddle.sencha.com/#fiddle/jf5
What I want to do is just set anyMatch: true, regardless, so I set it in Filter.js, but this has no effect on use of it. I searched the entire codebase for other instances of anyMatch: false and the only other one is in ext-all-debug.js. Why isn't setting these values having any effect? I don't see where else this default value could be set?
EDIT 1
I tried a different override, and while it is not exhibiting the right behavior, it is actually doing something this time. I figured that since the chunk of code that does work when embedded in the search attribute within the MultiSelector control was pretty much what was found in the MultiSelectorSearch's search method, that this was what I needed to focus on for the override. Any suggestions on tweaking this would be most welcome:
Ext.define('Ext.overrides.view.MultiSelectorSearch', {
override: 'Ext.view.MultiSelectorSearch',
search: function (text, me) {
var filter = me.searchFilter,
filters = me.getSearchStore().getFilters();
if (text) {
filters.beginUpdate();
if (filter) {
filter.setValue(text);
} else {
me.searchFilter = filter = new Ext.util.Filter({
id: 'search',
property: me.field,
value: text,
anyMatch: true
});
}
filters.add(filter);
filters.endUpdate();
} else if (filter) {
filters.remove(filter);
}
}
});
EDIT 2
Got it! The key was that originally, since this code was embedded in a singleton, I could reference the method by passing me from the calling form.panel. This did not work globally as an override, and required me to define the method as
search: function (text) {
var me = this,
I hope this helps someone out there!
Changing in ext-all-debug.js is not safe, when you do a production build this file will not get included.
Best way is to override the Filter class, here is how you can do it.
Ext.define('Ext.overrides.util.Filter', {
override: 'Ext.util.Filter',
anyMatch: true
});
And import this class in Application.js
Ext.require([
'Ext.overrides.util.Filter'
]);

How to organize EmberJS nested resources?

I'm trying to create a small EmberJS application, but I'm struggling about how to architecture it correctly. I have a main view called "library" which displays on a sidebar a list of folders. User can click on each folder and display the content at the center (while the sidebar is still active).
I therefore have a library resource, and nested resources to display the folders in this specific context:
this.resource('library', function() {
this.resource('libraryFolders', {path: 'folders'}, function() {
this.resource('libraryFolder', {path: ':folder_id'};
}
};
To be able to access the folders in the parent root, I set up a dependency:
App.LibraryController = Ember.Controller.extend({
needs: ["libraryFolders"],
folders: null,
foldersBinding: "controllers.libraryFolders"
});
App.LibraryRoute = Ember.Route.extend({
setupController: function(controller) {
controller.set('controllers.libraryFolders.model', App.Folder.find());
}
});
First question: is this a good way? I feel it a bit strange that a parent controller have a dependency to its children.
Now, another problem arises: what if I want to reuse folders in another context? All the methods I would write in LibraryFoldersController would be specific to this one, not really DRY. What I came up is adding a root "folders" resource, and add the dependency to this one instead:
this.resources('folders');
App.LibraryController = Ember.Controller.extend({
needs: ["Folders"],
folders: null,
foldersBinding: "controllers.folders"
});
App.LibraryRoute = Ember.Route.extend({
setupController: function(controller) {
controller.set('controllers.folders.model', App.Folder.find());
}
});
What do you think? Am I doing it wrong?
IMO it looks good so far. You are using the needs API which is the correct (ember) way to setup dependencies between controllers.
Maybe if you find yourself writing repeating code you could consider creating a Mixin for a more general controller an put there your logic, that should be agnostic to the use cases it handles.
For example defined a mixin:
App.ControllerMixin = Ember.Mixin.create({
// "use case" agnostic logic here
});
You mix mixins into classes by passing them as the first arguments to .extend.
App.LibraryController = Ember.ObjectController.extend(App.ControllerMixin, {
// now you can use here the logic defined in your mixin
// and add custom code as you please
});
Another possibility is to write a super class and then extend from it to inherit common logic:
Snippet taken from the docs:
App.Person = Ember.Object.extend({
helloWorld: function() {
alert("Hi, my name is " + this.get('name'));
}
});
var tom = App.Person.create({
name: 'Tom Dale'
});
tom.helloWorld(); // alerts "Hi, my name is Tom Dale".
One thing worth mentioning (though I think it's simply a typo) is: needs: ["Folders"] should be needs: ["folders"],
Hope it helps.

Extending the YUI Panel

I have a requirement to extend the YUI Panel with some custom functionality that will be in a new file and shared across multiple views.
I am at a bit of a loss as to how best to go about this, can anyone give me any pointers please?
Let's say you want to extend a Panel to create one that has a list in its body. I usually use Y.Base.create for this. It's a more declarative way of extending YUI classes than using a constructor and Y.extend. But I'll stay closer to your example in the YUI forums.
There are a couple of tricks dealing with WidgetStdMod (one of the components of Y.Panel), but mostly it's just about using Y.extend and following the YUI inheritance patterns. I'll try to answer with an example:
function MyPanel() {
MyPanel.superclass.constructor.apply(this, arguments);
}
// hack: call it the same so you get the same css class names
// this is good for demos and tests. probably not for real life
MyPanel.NAME = 'panel';
MyPanel.ATTRS = {
listItems: {
// YUI now clones this array, so all's right with the world
value: []
},
bodyContent: {
// we want this so that WidgetStdMod creates the body node
// and we can insert our list inside it
value: ''
}
};
Y.extend(MyPanel, Y.Panel, {
// always a nice idea to keep templates in the prototype
LIST_TEMPLATE: '<ul class="yui3-panel-list"></ul>',
initializer: function (config) {
// you'll probably want to use progressive enhancement here
this._listContainer = Y.Node.create(this.LIST_TEMPLATE);
// initializer is also the place where you'll want to instantiate other
// objects that will live inside the panel
},
renderUI: function () {
// you're inheriting from Panel, so you'll want to keep its rendering logic
// renderUI/bindUI/syncUI don't call the superclass automatically like
// initializer and destructor
MyPanel.superclass.renderUI.call(this);
// Normally we would append stuff to the body in the renderUI method
// Unfortunately, as of 3.5.0 YUI still removes all content from the body
// during renderUI, so we either hack it or do everything in syncUI
// Hacking WidgetStdModNode is doable but I don't have the code around
// and I haven't memorized it
//var body = this.getStdModNode('body');
},
syncUI: function () {
// same here
MyPanel.superclass.syncUI.call(this);
// insert stuff in the body node
var listContainer = this._listContainer.appendTo(this.getStdModNode('body'));
Y.Array.each(this.get('listItems'), function (item) {
listContainer.append('<li>' + item + '</li>');
});
}
});

Multiple Greasemonkey Metablocks

I'm trying to write a Greasemonkey script for a hierarchy of websites such that I have a bunch of code modifications for http://www.foo.com/*, then more specific ones for http://www.foo.com/bar/*, and still others for http://www.foo.com/foobar/*.
Is there anyway for me to write all these in the same script, or do I have to make multiple?
Is there anyway for me to write all
these in the same script, or do I have
to make multiple?
Yes, just use those three #includes, then in your user script do something like (depends on specifics of script):
var currentURL = (document.location+'');
if (currentURL .match(/http:\/\/www\.foo\.com\/foobar\/.*/)) {
// do stuff for page set A
} else if (currentURL .match(/http:\/\/www\.foo\.com\/foo\/.*/)) {
// do stuff for page set B
} else if (currentURL .match(/http:\/\/www\.foo\.com\/.*/)) {
// do stuff for page set C
}
One nifty trick I was shown for dealing with different functions at different sub-locations is to use the global directory of function names as a sort of virtual switchboard...
// do anything that is supposed to apply to the entire website above here.
var place = location.pathname.replace(/\/|\.(php|html)$/gi, "").toLowerCase();
// the regex converts from "foo/" or "foo.php" or "foo.html" to just "foo".
var handler;
if ((handler = global["at_" + place])) {
handler();
}
// end of top-level code. Following is all function definitions:
function at_foo() {
// do foo-based stuff here
}
function at_foobar() {
// do foobar stuff here.
}

Resources