Auto-resize Dojo Mobile Charts on Orientation Change

The best I can tell, Dojo’s dojox.mobile.Charts2D do not auto-resize on their own when the phone’s orientation changes. I posted a question on how to get around this on the Dojo Community Forum and never got an answer. So, I had to cobble together my own solution.

I have to point out that the functionality I built by hand is inherent in Flex and Silverlight, and you wouldn’t even bat an eyelash thinking about this. So, from a productivity standpoint I spent about double and maybe even triple the time I should have needed in order to sort through why things weren’t working as they should, and to build my own best practice for handling it. 

I do consider what I built as a hack, so caveat emptor. It should at least give you a good starting point to improve on what I’ve already done. There are some important things to note.

  • Here’s a sample demonstrating the functionality: https://andygup.net/samples/realestate/
  • Dojo does not provide any State properties on the View. So, I had to build that.
  • Dojo does not provide any way to bind a dijit to a mobile View. In other words, this enables the Chart to take action automatically when something happens in the View. Check…yep, I bolted that in.
  • Dojo, as far as I know, does not provide a way to detect when the phone’s orientation changes. So you have to listen for that at the window object level. I’m fairly certain that the pattern I used is not completely reliable across all platforms, but it’s what I had to work with. So, I built that too.
  • I also had to detect if there was no orientation change prior to a View transition. This was so that I didn’t unnecessarily redraw the chart and make it appear to flicker. This check was important because my chart is in a secondary View. There seems to be a bug in charts redraw() function in that the chart may self destruct if you try to redraw it from a different View.
  • There’s a bug in the Android native browser that passes the previous orientation event object to the listener. You actually have to set an event timer so that you retrieve the final, and most recent, orientation event object.
Here’s how you initialize the chart. In this case, I’m using a pieChart. This snippet also includes the html markup:
pieChart = new com.agup.PieChart("chart1","statsView").pieChart;

<div id="chart1ParentDiv" dojoType="dojox.mobile.RoundRect">
        <div id="chart1" style="width:100%; height: 350px;"></div>
</div>

Here’s the PieChart Class that I built to encapsulate the functionality I described above:

dojo.declare("com.agup.PieChart",null,{
    pieChart:null,
    orientationChanged:null,
    constructor:function(chartDiv,chartView){
        this.pieChart = this._createChart(chartDiv);
        this.orientationChanged = false;
        if(chartView)this._setTransitionListener(chartView);
        if(chartView)this._setOrientationListener();
    },
    _createChart:function(chartDiv){
        //create the chart
        //Had problems with using just HTML markup, so creating it here and piping to DIV
        var pieChart = new dojox.charting.Chart2D(chartDiv);
        //set the theme
        pieChart.setTheme(dojox.charting.themes.PlotKit.blue);
        //add plot
        pieChart.addPlot("default", {
            type: "Pie",
            radius: 100,
            fontColor: "black",
            labelOffset: "-20"
        });

        pieChart.isVisible = false; //NOTE: this is a new public property that we inject

        return pieChart;
    },
    _setTransitionListener:function(/* DIV of dojox.mobile.View where chart resides - typeof String  */view){
        var test = dijit.byId(view);
        var pieChart = this.pieChart;
        dojo.connect(test, "onAfterTransitionIn",null,
                dojo.hitch(this,function(){
                    pieChart.isVisible = true;
                    if(pieChart != null && this.orientationChanged == true)var time = setTimeout(function(){pieChart.resize()},700);
                })
        );

        dojo.connect(test, "onAfterTransitionOut",null,
                function(){
                    pieChart.isVisible = false;
                }
        );
    },
    _setOrientationListener:function(){
        var supportsOrientationChange = "onorientationchange" in window,
                orientationEvent = supportsOrientationChange ? "orientationchange" : "resize";

        window.addEventListener(orientationEvent,
            dojo.hitch(this,function(){
                var pieChart = this.pieChart;
                var orientationChanged = this.orientationChanged;
                if(pieChart != null && pieChart.isVisible == false){
                    orientationChanged = true;
                }
                if(pieChart != null && pieChart.isVisible == true){
                    orientationChanged = false;
                    var time = setTimeout(function(){pieChart.resize()},700);
                }
        }), false);
    }
});

The problem with JavaScript Obfuscators and Minification – Tracking down errors

JavaScript obfuscators and minifiers do their job well. In fact, some obfuscators have anti-debugging features. However, if you are a legitimate developer building applications against one of these libraries, chances are you’ve gotten an indecipherable error such as “z=null line 14300” and it brings your development efforts to a halt. Error messages like this provide no useful information on what the problem really is, or give any hints on how you might be able solve it. You’ve probably even looked at the jumbled source code in a last ditch attempt to make some sense out of the error. And, whether it’s your own library or a mainstream ones as jQuery or Dojo, it doesn’t matter. The amount of productivity lost because of these errors in probably very large, not to mention the frustration it causes.

I hope the the developers of these obfuscators are reading this…because I have a proposed solution to the problem.

Now, I want to start out by mentioning that I fully understand why obfuscators exist for reasons such as source code protection and decreasing download size. What I propose takes this fully into account, yet makes your library developer friendly in a secure way:

During the obfuscation process create an index file that maps each variable, function and class to a real line number and store this file in a web folder.  Then create a small html file that lets you search the index and return the real line number. Provide an option for return the variable, function or class name, too.

The concept is that if there is an error, like the  “z=null line 14300” I mentioned above, developers can then at least have some hope of narrowing down the general area of the code where it might be occurring.

The bonus is, if you own an obfuscated commercial library, now your tech support people can also look up the general area where a customer might be having a problem. For security reasons you don’t have to share the index file, But, even then, there isn’t enough information in it to de-compile the library. Now, if I post my error to the forum:  What is “z=null line 14300”? Tech support will be able to tell me that I’m missing a custom property on a widget’s HTML DIV element. It’s a win-win situation.

What do you think?

Presenting at LinuxFestNW 2012 on HTML5 Geolocation

This weekend I’ll be presenting at LinuxFestNW 2012. We (Esri) will also have a booth there, so if you are in the area of Bellingham, WA swing on by.

My topic will be HTML5 Geolocation and I’ll be talking about building applications using the HTML5 Geolocation API from A to Z. I’m going to start out looking at basic coding patterns, then we’ll discuss accuracy across different browsers, look at many common gotchas, and I’ll be giving tips on how to effectively implement it with local and remote database. If you attend, come prepared to get a full dose of HTML5 goodness.

Here’s the link to the prezi: https://prezi.com/kawyhcqxjdip/html5-geolocation-api-location-location-location/