Over the past year, one of our major efforts for our iPhora products was to switch from using Dojo Dijit widgets to only Bootstrap widgets. One of the problems that we have seen with Dijit widgets is that they take a long time to load and are very heavy compared with JQuery based Bootstrap widgets. The advantage of JQuery-based Bootstrap widgets is that they are loaded all at once as a single file. This is great to minimize the loading speed. However, it does not lend itself to easy maintenance by multiple developers.
With Dojo you load the individual widgets are needed using dojo.require. Unfortunately, this results in multiple HTTP request in order to load the individual widgets and support files. Even if we are not using Dijit widgets but you want to use dijit._Widget to creating your own widgets, Dojo loads a whole brunch of additional files like dijit._LayoutWidget.js and doubles the loading process. That is because you need to load dijit.js in order to use dijits. So without any widgets on the page it requires you to load the following:
I highlighted in red the files that I did not need to load.
Note: I use buildRendering directing instead of using dijit._Templated. I load dojo.string when I need it and do not use dojo.cache.js since I am not using dijit._Templated.
So out of frustration, I was determined to figure out how to get rid of these unneeded files. When you require dijit._Widget it automatically forces you to load dijit._base which loads many files that is specific to dijit widgets. So I first strip this down from:
/* Copyrighted 2009-2013 Phora Group. All rights reserved.*/
if(!dojo._hasResource["dijit._base"]){
dojo._hasResource["dijit._base"]=true;
dojo.provide("dijit._base");
dojo.require("dijit._base.focus");
dojo.require("dijit._base.manager");
dojo.require("dijit._base.place");
dojo.require("dojo.uacss");
dojo.require("dijit._base.typematic");
dojo.require("dijit._base.window");
}
if(!dojo._hasResource["dijit.dijit"]){
dojo._hasResource["dijit.dijit"]=true;
dojo.provide("dijit.dijit");
dojo.require("dijit._Widget");
dojo.require('dojo.cookie');
dojo.require('iphora._ipBase');
}
It first loads all the _base files that I need and then load dijit which contains the only thing I need, dijit._Widget. As part of the process I have it also load dojo.cookie which I usually need and iphora._ipBase which has all the base iphora support methods for creating Bootstrap widgets.
This method eliminates the need to call dijit._base separately which normally creates another HTTP request that is not needed. When the call is made to load dijit._Widget, dijit._base is already loaded with the new file list definition so it uses this list instead of the one that is provided.
So as a result of this exercise the loading process requires less than half of the files.
With Dojo you load the individual widgets are needed using dojo.require. Unfortunately, this results in multiple HTTP request in order to load the individual widgets and support files. Even if we are not using Dijit widgets but you want to use dijit._Widget to creating your own widgets, Dojo loads a whole brunch of additional files like dijit._LayoutWidget.js and doubles the loading process. That is because you need to load dijit.js in order to use dijits. So without any widgets on the page it requires you to load the following:
I highlighted in red the files that I did not need to load.
Note: I use buildRendering directing instead of using dijit._Templated. I load dojo.string when I need it and do not use dojo.cache.js since I am not using dijit._Templated.
So out of frustration, I was determined to figure out how to get rid of these unneeded files. When you require dijit._Widget it automatically forces you to load dijit._base which loads many files that is specific to dijit widgets. So I first strip this down from:
if(!dojo._hasResource["dijit._base"]){
dojo._hasResource["dijit._base"]=true;
dojo.provide("dijit._base");
dojo.require("dijit._base.focus");
dojo.require("dijit._base.manager");
dojo.require("dijit._base.place");
dojo.require("dijit._base.popup");
dojo.require("dijit._base.scroll");
dojo.require("dijit._base.sniff");
dojo.require("dijit._base.typematic");
dojo.require("dijit._base.wai");
dojo.require("dijit._base.window");
}
to
if(!dojo._hasResource["dijit._base"]){
dojo._hasResource["dijit._base"]=true;
dojo.provide("dijit._base");
dojo.require("dijit._base.focus");
dojo.require("dijit._base.manager");
dojo.require("dijit._base.place");
dojo.require("dijit._base.typematic");
dojo.require("dijit._base.window");
}
}
which takes care of all the dijit._base files. The rest of the highlighted files are load by dijit.js. So instead of loading dijit.js for the page I created by own loader js file that contains:
/* Copyrighted 2009-2013 Phora Group. All rights reserved.*/
if(!dojo._hasResource["dijit._base"]){
dojo._hasResource["dijit._base"]=true;
dojo.provide("dijit._base");
dojo.require("dijit._base.focus");
dojo.require("dijit._base.manager");
dojo.require("dijit._base.place");
dojo.require("dojo.uacss");
dojo.require("dijit._base.typematic");
dojo.require("dijit._base.window");
}
if(!dojo._hasResource["dijit.dijit"]){
dojo._hasResource["dijit.dijit"]=true;
dojo.provide("dijit.dijit");
dojo.require("dijit._Widget");
dojo.require('dojo.cookie');
dojo.require('iphora._ipBase');
}
It first loads all the _base files that I need and then load dijit which contains the only thing I need, dijit._Widget. As part of the process I have it also load dojo.cookie which I usually need and iphora._ipBase which has all the base iphora support methods for creating Bootstrap widgets.
This method eliminates the need to call dijit._base separately which normally creates another HTTP request that is not needed. When the call is made to load dijit._Widget, dijit._base is already loaded with the new file list definition so it uses this list instead of the one that is provided.
So as a result of this exercise the loading process requires less than half of the files.
What you need to load depends are what you are doing with Dojo. So if you are interested in creating a faster loading process you should experiment. Have fun.
Comments