Best practices for passing configuration to a qooxdoo app - javascript

A complex qooxdoo (desktop) application may need to obtain deployment-specific configuration at startup, for example:
REST/WebSocket backend URL (which may be different from script/document location);
branding data;
enabled/disabled features;
performance tweaks
and so on. At the moment, we do the following:
<head>
<script type="text/javascript">
config = {
key1: "foo",
key2: "bar",
key3: "<%= getParameter("myapp.key3") %>"
};
</script>
<script type="text/javascript" src="script/myapp.js"></script>
</head>
As you've noticed, it's a JSP that provides dynamic config.key3, along with static config.key1 and config.key2. I somewhat dislike this approach because 1) config top-level object is not recognized in qooxdoo application code, and this results in compiler warning, 2) you can accidentally mess up with another top-level object of same name. Can anyone suggest any better, cleaner way to do the same? It's important that a combination of static and dynamic configuration is supported. For example, a developer hardcodes his specific config into some VCS-ignored JSON file; production environment serves config data from the database; staging/QA environment can use both sources for different parts of config.
The problem is not limited to qooxdoo, of course. I think this may as well be topical for any standalone JavaScript application that may need deployment-specific customization.

Use below method to solve both problems (warning and top level object):
a. in index.html before loading your application's JS:
window.qx = { $$environment: {
"myapp.key1": "foo"
} };
b. later in your Qooxdoo application
var key1 = qx.core.Environment.get("myapp.key1");

For pre-defined environments I suggest to use normal Qooxdoo configuration, not the hack #geonik presented. Here follows excerpt from normal config.json:
"config-warnings" :
{
"job-shadowing" : ["common", "source-script", "build-script"]
},
"jobs" :
{
"common" :
{
"environment" :
{
"appname.version" : "1.0.0",
"appname.timeout" : 30000,
"appname.analyticsId" : "UA..."
}
},
"source-script" :
{
"environment" :
{
"appname.serviceUrl" : "http://127.0.0.1:8080/service",
"appname.crossDomain" : true
}
},
"build-script" :
{
"environment" :
{
"appname.serviceUrl" : "/service",
"appname.crossDomain" : false
}
}
}
With the above you have one settings in development build and other settings in production build. You can add more environments, like QA. with other desired settings. Then access them with normal qx.core.Environment.get.
I don't suggest to mix runtime settings with Qooxdoo configuration which is inherently static. Just load them on application start with any qx.io.* way and handle in your classes.

Related

How to override the naming of on_failure screenshots?

Unfortunately due to using cucumber and writing out scenarios I end up with incredibly long file names which windows complains about. Is it possible to override the naming for the files?
I'm assuming you're using the Nightwatch framework since you've tagged it in your post.
It looks like screenshot filenames are defined in nightwatch/lib/api/client-commands/end.js at line 26 of the latest repo const prefix = '${this.api.currentTest.module}/${this.api.currentTest.name}';.
The screenshot path is defined in your nightwatch.json:
{
"test_settings" : {
"default" : {
"screenshots" : {
"enabled" : true,
"on_failure" : true,
"path" : "./screens"
}
}
}
}
It appears your have a few options:
Modify the code in end.js to use a custom naming scheme, and live with a custom framework
Shorten your test module names or test names
Shorten your screenshot destination path in nightwatch.json - unlikely to solve anything since you're saying it's the filenames that are the problem.
This should be possible by using the filename_format option, since:
#2023 was merged

Load app.js before rest of application

I'm trying to figure out how I can load app.js before allowing the user to get the actual application. What I'm attempting to do is load a user's configuration file before all of my class Ext.defines fire... the reason I want to do this is because the Ext.defines actually depend on values in the user's configuration. So for example, in an Ext.define, I could have the title property set to pull from this global user configuration var. And no, I don't want to have to go through and change all of these properties to use initComponent... that could take quite some time.
Instead, what I'd like to do is load the configuration, and then let the Ext.defines run, but I will need Ext JS and one of my defined classes to be loaded before the rest of the classes. Is this possible? I've been looking into Sencha Cmd settings, but I've been extremely unsuccessful with getting this to work. I was playing with the bootstrap.manifest.exclude: "loadOrder" property, which loads classic.json, and doesn't define my classes, but unfortunately, that also doesn't fully load Ext JS, so Ext.onReady can't be used... nor can I use my model to load the configuration.
I have a very high level example below (here's the Fiddle).
Ext.define('MyConfigurationModel', {
extend: 'Ext.data.Model',
singleton: true,
fields: [{
name: 'testValue',
type: 'string'
}],
proxy: {
type: 'ajax',
url: '/configuration',
reader: {
type: 'json'
}
}
});
// Pretend this would be the class we're requiring in our Main file
Ext.define('MyApp.view.child.ClassThatUsesConfiguration', {
extend: 'Ext.panel.Panel',
alias: 'widget.classThatUsesConfiguration',
/* We get an undefined value here because MyConfigurationModel hasn't
* actually loaded yet, so what I need is to wait until MyConfigurationModel
* has loaded, and then I can include this class, so the define runs and
* adds this to the prototype... and no, I don't want to put this in
* initComponent, as that would mean I would have to update a ton of classes
* just to accomplish this */
title: MyConfigurationModel.get('testValue')
});
Ext.define('MyApp.view.main.MainView', {
extend: 'Ext.Viewport',
alias: 'widget.appMain',
requires: [
'MyApp.view.child.ClassThatUsesConfiguration'
],
items: [{
xtype: 'classThatUsesConfiguration'
}]
});
Ext.define('MyApp.Application', {
extend: 'Ext.app.Application',
mainView: 'MyApp.view.main.MainView',
launch: function() {
console.log('launched');
}
});
/* In app.js... right now, this gets called after classic.json is downloaded and
* after our Ext.defines set up, but I basically want this to run first before
* all of my classes run their Ext.define */
Ext.onReady(function() {
MyConfigurationModel.load({
callback: onLoadConfigurationModel
})
});
function onLoadConfigurationModel(record, operation, successful) {
if (successful) {
Ext.application({
name: 'MyApp',
extend: 'MyApp.Application'
});
}
else {
// redirect to login page
}
}
I call this "splitting the build", because it removes the Ext.container.Viewport class's dependency tree from the Ext.app.Application class. All Ext JS applications have a viewport that is set as the main view. By moving all requires declarations of the core of the application to the viewport class, an application can load the viewport explicitly from the application class, and the production build can be configured to output two separate files, app.js and viewport.js. Then any number of operations can occur before the core of the application is loaded.
// The app.js file defines the application class and loads the viewport
// file.
Ext.define('MyApp.Application', {
extend: 'Ext.app.Application',
requires: [
// Ext JS
'Ext.Loader'
],
appProperty: 'application',
name: 'MyApp',
launch: function() {
// Perform additional operations before loading the viewport
// and its dependencies.
Ext.Ajax.request({
url: 'myapp/config',
method: 'GET',
success: this.myAppRequestSuccessCallback
});
},
myAppRequestSuccessCallback: function(options, success, response) {
// Save response of the request and load the viewport without
// declaring a dependency on it.
Ext.Loader.loadScript('classic/viewport.js');
}
});
-
// The clasic/viewport.js file requires the viewport class which in turn
// requires the rest of the application.
Ext.require('MyApp.container.Viewport', function() {
// The viewport requires all additional classes of the application.
MyApp.application.setMainView('MyApp.container.Viewport');
});
When building in production, the viewport and its dependencies will not be included in app.js, because it is not declared in the requires statement. Add the following to the application's build.xml file to compile the viewport and all of its dependencies into viewport.js. Conveniently, the development and production file structures remain the same.
<target name="-after-js">
<!-- The following is derived from the compile-js target in
.sencha/app/js-impl.xml. Compile the viewport and all of its
dependencies into viewport.js. Include in the framework
dependencies in the framework file. -->
<x-compile refid="${compiler.ref.id}">
<![CDATA[
union
-r
-class=${app.name}.container.Viewport
and
save
viewport
and
intersect
-set=viewport,allframework
and
include
-set=frameworkdeps
and
save
frameworkdeps
and
include
-tag=Ext.cmd.derive
and
concat
-remove-text-references=${build.remove.references}
-optimize-string-references=${build.optimize.string.references}
-remove-requirement-nodes=${build.remove.requirement.nodes}
${build.compression}
-out=${build.framework.file}
${build.concat.options}
and
restore
viewport
and
exclude
-set=frameworkdeps
and
exclude
-set=page
and
exclude
-tag=Ext.cmd.derive,derive
and
concat
-remove-text-references=${build.remove.references}
-optimize-string-references=${build.optimize.string.references}
-remove-requirement-nodes=${build.remove.requirement.nodes}
${build.compression}
-out=${build.out.base.path}/${build.id}/viewport.js
${build.concat.options}
]]>
</x-compile>
<!-- Concatenate the file that sets the main view. -->
<concat destfile="${build.out.base.path}/${build.id}/viewport.js" append="true">
<fileset file="classic/viewport.js" />
</concat>
</target>
<target name="-before-sass">
<!-- The viewport is not explicitly required by the application,
however, its SCSS dependencies need to be included. Unfortunately,
the property required to filter the output, sass.name.filter, is
declared as local and cannot be overridden. Use the development
configuration instead. -->
<property name="build.include.all.scss" value="true"/>
</target>
This particular implementation saves the framework dependencies in their own file, framework.js. This is configured as part of the output declaration in the app.json file.
"output": {
...
"framework": {
// Split the framework from the application.
"enable": true
}
}
https://docs.sencha.com/extjs/6.2.0/classic/Ext.app.Application.html#cfg-mainView
https://docs.sencha.com/extjs/6.2.0/classic/Ext.container.Viewport.html
https://docs.sencha.com/cmd/guides/advanced_cmd/cmd_build.html#advanced_cmd-_-cmd_build_-_introduction
As far as I know, this is not possible with Sencha Cmd, because while Sencha Cmd can load framework and application separately, it is not possible to tell the production microloader to wait with the second file until the code from the first file has done something (presumably loaded something from the server?).
So the only approach would be to get the options outside ExtJS, before loading ExtJS.
You would have to write your own javascript that loads the configuration into a global variable using a bare, synchronous XmlHttpRequest, and include that into the index.html before the ExtJS script. That way, the script is executed before ExtJS is loaded at all, and you have completely consistent behaviour across development, testing and production builds without modifying any framework file that may be overwritten during framework upgrades.
I guess this is what you are searching for.
So how I did it: In index.html, I added a custom script that fills some global variables:
<meta name="viewport" content="width=device-width, initial-scale=1, maximum-scale=1, user-scalable=no">
<script type="text/javascript">
APIURI = '../api/', // <- also used in ExtJS.
var xhr = new XMLHttpRequest();
xhr.open('GET', APIURI+'GetOptions', false);
xhr.setRequestHeader('Accept','application/json');
xhr.send(null);
try {
var configdata = eval("(" + xhr.responseText + ")");
} catch(e) {
// snip: custom code for the cases where responseText was invalid JSON because of a shitty backend
}
if(configdata.options!=undefined) Settings = configdata.options;
else Settings = {};
if(configdata.translations!=undefined) Translations = configdata.translations;
else Translations = {};
Translations.get=function(str) {
if(typeof Translations[str]=="string") return Translations[str];
return "Translation string "+str+" missing.";
};
</script>
<link rel="icon" type="image/vnd.microsoft.icon" href="../favicon.ico">
<title>Application</title>
<script id="microloader" data-app="1a7a9de2-a3b2-2a57-b5af-df428680b72b" type="text/javascript" src="bootstrap.js"></script>
Then I could use in Ext.define() e.g. title: Translations.get('TEST') or hidden: Settings.HideSomeButton or url: APIURI + 'GetUserData'.
However, this has major drawbacks you should consider before proceeding.
After a short period of time, new feature requests emerged and settings I had considered fixed should change at runtime, and I realized that always reloading the application when a setting changes is not good user experience. A while later, I also found that Chrome has deprecated synchronous XmlHttpRequests, and that this approach delays application startup time.
So, the decision was made that in the long run, the only sane approach is to be able to react to changes of any configuration value at runtime, without a full reload of the application. That way, settings could be applied after loading the application, and the requirement could be dropped to wait for settings load before proceeding with the application.
For this, I had to completely work out everything needed for full localization support, so the user can switch between languages without reload of the application, and also any other setting can change at runtime and is automatically applied to the application.
Short-term, this is quite some work, which didn't really matter to me because I was scheduled to rework the whole application layout, but long-term, this will save quite some time and headache, especially when someone decides we should start polling for changes to the settings from the server, or that we should use an ExtJS form for login instead of good old Basic authentication (which was by then already asked for multiple times, but we couldn't deliver because of said shitty ExtJS app architecture).
We actually do use a Sencha CMD approach. As #Alexander mentioned, we also use a global variable for keeping the application's configuration. This approach also implies that the server returns the actual declaration of the global config variable.
If you dig into the app.json file, and find the js config key, you will see that in the description it says
List of all JavaScript assets in the right execution order.
So, we add the configuration's endpoint before the app.js asset
"js": [
{
"path": "data/config",
"remote": true
},
{
"path": "${framework.dir}/build/ext-all-debug.js"
},
{
"path": "app.js",
"bundle": true
}
]
also specifying remote: true.
// Specify as true if this file is remote and should not be copied
into the build folder
The "data/config" endpoint returns something like:
var CONFIG = {
user: {
id: 1,
name: 'User'
},
app: {
language: 'en'
}
}
And now we can have a reference to the CONFIG variable anywhere in our classes.

Dojo build doesn't include dojo/dom, dom/when, dojo/dom-class and about 100 other modules

I have a problem with my build in Dojo. It does build, and most of all widgets seems to be included in dojo.js after the build.
But when I test the built project it still loads about 100 files on demand.
I think the common denominator for the files that doesn't get build, is that they don't use return declare(
But instead returns functions or objects.
I attach a print-screen of some of the modules that doesn't get bundled in the build.
Dump from Firebug NET-console
The question is, is there some way of bundling these files into dojo.js, and avoid the 100+ extra requests?
Dojo builds are a pain in my neck. There are several different ways to configure them.
Generally, if you're trying to build everything (including Dojo) into one Javascript source file make sure your layer has "customBase" and "boot" set to "true".
build.profile.js
var profile = (function() {
return {
layers: {
"my/layer": {
customBase: true,
boot: true
}
}
}
}();
That should catch all of the Dojo source files. Otherwise, if something somehow slips that's what the "include" option is for. It's an explicit list of modules that get built into the layer.
build.profile.js
var profile = (function() {
return {
layers: {
"my/layer": {
include: [ "dojo/dojo", "dojo/date", ... ]
}
}
}
}();

Different settings for debug/local ("grunt serve") vs. dist/build ("grunt")?

I want to define some application settings, but I want to provide different values depending on whether I'm running in 'debug' mode (e.g. grunt serve), or whether the final compiled app is running (e.g. the output of grunt). That is, something like:
angular.module('myApp').factory('AppSettings', function() {
if (DebugMode()) { // ??
return { apiPort: 12345 };
} else {
return { apiPort: 8008 };
}
});
How can I accomplish this?
The way I handle it in my apps:
move all your config data for one environment to a file: config.js, config.json,... whatever your app finds easy to read.
now modify your config file to turn it into a template using grunt config values, and generate the file with grunt-template as part of your build - for example: app.constant('myAppConfig', {bananaHammocks: <%= banana.hammocks %>});
finally, add grunt-stage to switch grunt config values depending on environment: create your different config/secret/(env).json files, update your template (app.constant('myAppConfig', {bananaHammocks: <%= stg.banana.hammocks %>});), and then grunt stage:local:build or grunt stage:prod:build
I find this the good balance between complexity and features (separation between environments, runtime code not concerned with building options,...)

Prevent browser cache issue on Javascript files with RequireJS in SeedStack

using SeedStack 14.7 we are facing a cache issue when uploading a new version on servers: every user have to clear their cache to get the last version of files.
I tried to use "urlArgs": "version=2" in the requireConfig part of the fragment JSON file. It do the job by adding argument on every files and so we can use it when changing version, but it also affect the urls in the config of each modules !
As we are using this config to pass the REST base url to each module, it breaks all REST requests by adding the argument to the base url.
My fragment JSON file :
{
"id": "mac2-portail",
"modules": {
"gestionImage": {
"path": "{mac2-portail}/modules/gestionImage",
"autoload": true,
"config": {
"apiUrl": "muserver/rest"
}
}
},
"i18n": {...},
"routes": {...},
"requireConfig": {
"urlArgs": "version=2",
"shim": {...}
}
}
Any idea to solve the cache issue without breaking REST requests ?
EDIT : it is not a duplicate of Prevent RequireJS from Caching Required Scripts. Yes SeedStack uses RequireJS and this configuration solve the cache issue, but it also affect other modules defined in the fragment so I need to find another solution to prevent browser to cache files
The module configuration values, like apiUrl in your example, are not touched by RequireJS unless you call require.toUrl() on them explicitly. I think this is what is happening in your case. To avoid this problem, you should always do the concatenation first and only then call require.toUrl() on the full resulting URL.
So, instead of doing:
var fullUrl = require.toUrl(config.apiUrl) + '/my/resource';
Do this:
var fullUrl = require.toUrl(config.apiUrl + '/my/resource');
By the way, instead of setting the version directly in the RequireJS configuration, you can simply add the version of your application to the data-w20-app-version attribute on the <html> element of the master page:
<html data-w20-app data-w20-app-version="2.0.0">
This will provide the same behavior but will work correctly in the case of Angular templates in $templateCache. If your master page is automatically generated by the backend, this is done automatically. Check this page for the details.

Categories

Resources