Subscribe to GameFromScratch on YouTube Support GameFromScratch on Patreon

19. June 2012

 

A couple months back took a long look at Appcelerator and initially I loved what I saw.  For those of you unaware, Appcelerator allows you to develop “native” applications for Android and iOS in JavaScript that execute in a special version of the V8 JavaScript Engine. I started developing a prototype app of a simple Android game for my daughter and got surprisingly good results quickly.  It was a very encouraging start, but then the wheels fell off.  I notice a number of people end up on this site from search engines looking for an Appcelerator review ( for some reason? ) so I thought I would share my experiences.

 

For whatever reason, I got distracted and put it down for a while, I came back to it recently and intended to finish off that prototype.  Here is the kicker, the exact same code from a couple months ago, unchanged, no longer works on my Transformer tablet.

 

I will say, this is frustrating to say the least.  Now, since the code didn’t change it was obviously an update to Android that broke things, I can’t exactly blame that on Appcelerator Titanium ( oh the joys of Android device compatibility… ) but I can blame them for what came next…

 

You can’t debug on device.  At all.  With the exception of printing comments to logcat, you are basically blind.  This obviously leads to a problem, as here is what happens when I run the app.  ( Which by the way, still runs perfectly fine on my Galaxy Note running Android 2.3.x, its just on my Asus Transformer that it breaks ):

 

appcell

 

Since I can’t debug on device, this is the results I get from connecting with the Delvik Debug Monitor:

 

06-19 09:56:51.090: I/TiApplication(1823): (main) [0,0] checkpoint, app created.
06-19 09:56:51.130: I/TiApplication(1823): (main) [46,46] Titanium 2.0.2 (2012/05/30 10:21 2ff31a3)
06-19 09:56:51.170: D/dalvikvm(1823): GC_CONCURRENT freed 268K, 5% free 6703K/7047K, paused 2ms+2ms
06-19 09:56:51.380: D/dalvikvm(1823): GC_CONCURRENT freed 314K, 6% free 6898K/7303K, paused 2ms+2ms
06-19 09:56:51.450: E/TiApplication(1823): (KrollRuntimeThread) [318,364] APP PROXY: ti.modules.titanium.app.AppModule@4112e910

 

 

And, that is it.  Since I can’t debug on the device, I am pretty much DOA here.  The forums answer nothing ( although I did find other people having similar problems ).  I of course could debug in the emulator with two exceptions.  a) the error only occurs on the actual device b) the debugging experience on Android is still awful.

 

This is another area with Appcelerator Titanium I always found infuriating… the emulator.  Getting it to actually run with your running emulator was a nightmare, you had to match up the Android build version and run configuration exactly with the version of the running emulator, then it was still a 50/50 chance that it would actually use your emulator, or start a new one.  If you’ve worked with Android before, you know how fun waiting for the emulator is!  There is light and the end of the tunnel with the recently release Intel emulator running a heck of a lot quicker, but unfortunately the JavaScript V8 virtual machine Appcelerator relies on doesn’t work with it.

 

It’s a shame to be tripped up on such a thing, but actually being able to debug on your device is an absolute requirement, as is demonstrated by this situation.  There are other annoyances with using Appcelerator certainly…  the binaries it generates are “large” to be extremely kind, performance can be a bit spotty at times and you virtually have to write two applications if you want to support iOS and Android, but none of these was enough to make me give up. 

 

Not being able to debug on device, coupled with some remarkably meh support.. that is.  Combined with the fact my application was broken by an Android update… if I can’t get my code working with my three devices reliably, what happens when I release to the masses with their millions of different devices?

 

Which is disappointing, as I really liked working with Appcelerator.  I even liked Appcelerator Titanium as an IDE, which is shocking as I hate Eclipse, which it is based on. 

 

There is potentially light at the end of the tunnel, as there is an open ticket for debugging on device.  Unfortunately, its been assigned medium priority and has already been open for almost a year.

 

Appcelerator Titanium is an interesting and encouraging product, but for the above reasons, I just can’t commit to using it.  Add native debugging support, and I will certainly reconsider.

General , , ,

17. June 2012

 

In this tutorial we are going to look at adding keyboard and mouse support to your cocos2D application.  cocos2D was traditionally used on the iPhone, so keyboard and mouse support is very touch centric, so if you are used to desktop development and things seem a bit strange, this is why.

Alright, let’s jump right in.  We are going to create a simple game with our jet sprite from tutorial 3.  The big difference is it will respond to your mouse and keyboard activity.  Clone your previous project, this time I am calling it MyThirdApp.  Yeah, tons of thought went into these names! 

 

Now create a new file in the Classes directory named JetSprite.js. Inside, write the following code:

 

var JetSprite = cc.Sprite.extend({
    _currentRotation:0,
    ctor:function(){
        this.initWithFile("images/jet.png");
    },
    update:function(dt){
        this.setRotation(this._currentRotation);
    },
    handleKey:function(e)
    {
        if(e === cc.KEY.left)
        {
            this._currentRotation--;

        }
        else if(e === cc.KEY.right)
            this._currentRotation++;

        if(this._currentRotation < 0) this._currentRotation = 360;
        if(this._currentRotation > 360) this._currentRotation = 0;
    },
    handleTouch:function(touchLocation)
    {
        if(touchLocation.x < 300)
            this._currentRotation = 0;
        else
            this._currentRotation = 180;
    },
    handleTouchMove:function(touchLocation){
        // Gross use of hardcoded width,height params.
        var angle = Math.atan2(touchLocation.x-300,touchLocation.y-300);

        angle = angle * (180/Math.PI);
        this._currentRotation = angle;

    }
});

 

Here we are deriving our new class JetSprite from cc.Sprite using the extend() function.  During the constructor, we first call back to cc.Sprite’s constructor, then load our sprite from the file jet.png located in the images folder.  ( If you like me started your project by cloning MySecondApp, it should already be there, otherwise make sure you have an image there ).

 

The update() function is going to be called once every frame, this will make sense later.  Essentially this is where you put your object’s logic.  In this case we are simply rotating our sprite to the value in _currentRotationhandleKey(), handleTouch() and handleTouchMove() are three user defined functions that are going to be used to pass I/O events into our JetSprite object.

 

handleKey() is passed the key value ( it’s a number ) of the currently pressed key.  cocos2D defines a number of key values to represent each key in the cc.KEY value.  If you want more details check out CCKeyboardDispatcher.js in the cocos2D directory for more details.  In the event the left arrow key is held down, we decrement our rotation value, if the right arrow key is down we increment the rotation value.  Finally, if we roll over 360 or under 0 degrees, we update accordingly.  Effectively this rotates our sprite left and right on arrow key press.

 

handleTouch() is very simple.  It is passed in the touchLocation, which is an x,y value representing where on screen was touched ( or in our case, clicked ).  If its on the left half the screen, we point up ( 0 degrees ), if the touch occurred on the right half of the screen, we rotate to straight down ( 180 degrees ).  One limitation of cocos2D seems to be the lack of ability to handle right clicks, but this may just be an oversight on my behalf.  Right now, a “touch” is any mouse button click.

 

handleTouchMove() is passed in the x,y coordinates of the mouse’s currently location.  As you can tell by the comment, I rather crudely used a hardcoded value for representing half the screen width ( 300 ), obviously in non-demo code you wouldn’t do this!  What we want to do here is figure out the angle of the mouse relative to the center of our sprite ( which is located at 300,300 ).  This is done by taking the atan2 of the distance of the mouse’s coordinate from sprite’s center.  We then convert that value ( atan2 returns a value in radians ) to degrees.  Finally, we take our calculated angle and apply it as our rotation.  Essentially this will cause our sprite to rotate relative to the mouse location.

 

Now lets take a look at the heart of our application.  Create a new script file ( or rename MySecondApp.js ) called MyThirdApp.js and enter the following code:

 

var MyThirdApp = cc.LayerColor.extend(
{   _jetSprite:null,
    init:function(){
        this._super();
        this.initWithColor(new cc.Color4B(0,0,0,255));
        var size = cc.Director.getInstance().getWinSize();

        this._jetSprite = new JetSprite();
        this.setTouchEnabled(true);
        this.setKeyboardEnabled(true);

        this.setPosition(new cc.Point(0,0));

        this.addChild(this._jetSprite);
        this._jetSprite.setPosition(new cc.Point(size.width/2,size.height/2));
        this._jetSprite.scheduleUpdate();
        this.schedule(this.update);

        return true;
    },
    onEnter:function(){
        this._super();
    },
    update:function(dt){
    },
    onTouchesEnded:function (pTouch,pEvent){
        this._jetSprite.handleTouch(pTouch[0].getLocation());
    },
    onTouchesMoved:function(pTouch,pEvent){
        this._jetSprite.handleTouchMove(pTouch[0].getLocation());
    },
    onKeyUp:function(e){

    },
    onKeyDown:function(e){
        this._jetSprite.handleKey(e);
    }
});



MyThirdAppScene = cc.Scene.extend({
    onEnter:function(){
        this._super();
        var layer = new MyThirdApp();
        layer.init();
        this.addChild(layer);
    }
})

 

The bulk of this code is very similar to previous code with one major difference.  In previous examples, we created a layer, then added a color layer to it.  In this case, we simply extend from LayerColor and add our children directly to ourselves.  There is very little functional differences between the two approaches, except this one is a bit more concise.  Now let’s take a closer look at our code.

 

In our init() method, we init with a black background colour.  We then create our JetSprite object and tell cocos2D that our layer handles touch and keypad events.  In cocos2D, input is handled at the layer level, and setting these two values will result in touch and keyboard delegates being added to our object so we can handle IO events.  We then position our layer, add our _jetSprite as a child positioned in the center of our layer.  The scheduleUpdate() call is very important.  This will result in our JetSprite’s update() function getting called every frame.  If you don’t call this method, or sprite wont do anything!

 

Next up is onEnter() which is what is called when a layer becomes the active layer.  In our case, this call is very important, as it ( in cc.Layer.onEnter ) is where the touch and keyboard delegates are actually attached to our layer!  If you remove the call to this._super(), you will not receive any keyboard and mouse events!

 

Now that we are wired up to receive touch and keyboard events, we need a couple methods that will be called when they occur.  In our case we are going to use ccTouchesEnded, ccTouchesMoved, keyUp and keyDown.  When an event occurs, these methods will be called, which in turn simply call the appropriate method on our JetSprite.

 

There are two things to be aware of here.  There are ccTouch____ and ccTouches____ versions of events; to my experience, the ccTouch____ version will never be called!  If you want to handle touch/mouse events, use the ccTouches version.  The next is the way we are handling key presses,  right now since we used keyDown, holding down the key will result in it being called over and over.  If however you want to respond only once per key hit ( regardless to if it is held down or not ), use keyUp instead.

 

We also need to make sure our new classes are included in our project.  Just like the last tutorial, make the following changes to the bottom of cocos2d.js

appFiles:['MyThirdApp.js','JetSprite.js']

 

We also need to update the Scene name in main.js

var myApp = new cocos2dApp(MyThirdAppScene);

Here is the end result of all of our code.  Mouse over the canvas and the mouse will follow you.  Left and right arrow will rotate the sprite, while clicking (either button) on the left half of the screen will point the jet up, while clicking the right half will point the sprite down.

 

 


 

If you are running Chrome be aware the clicking doesn’t appear to work.  It is actually working, the sprite is being rotated.  However it is also triggering a mouse move event, immediately snapping the sprite back to where you mouse pointer is! Smile

 

As always, you can download the whole project right here.  It includes a zip of the cocos2D version I used, in case you have trouble working with the current release.

 

 

Read Tutorial 5

Programming , ,

15. June 2012

 

I have recently become more and more interesting in HTML5 game development.  However,HTML5_Logo_256 upon initial investigation I realized there are an absolute ton of libraries and tools out there.  Some of them are amazing, others… not so much.  During my research I put together notes on the key libraries, tools, books, frameworks and IDEs used for HTML5 development.  This is the results of that research.

 

So, if you are looking to get into HTML5 game development, this page could be a very good place to start.

 

This list is by no means complete, it just represents the best of what I have found so far.  If you have recommendations for inclusion ( especially for the links section! ), please let me have them!  There are links as well to non-game related frameworks like Backbone.js, YUI, Mustache, etc. I am intending to do some game tool development in JavaScript for an upcoming RPG project, so those are there as much for my use as anything.

 

I hope you find it handy.

 

Links of interest for HTML5 game developers.

Programming ,

8. June 2012

 

Just a quick note to say I’ve organized my recent Cocos2D posts into a traditional tutorial series, like this site’s C++ and PlayStation Mobile tutorials.  You can now access them from the right hand navigation list, or by clicking here

 

Right now we are at 3 tutorials, plus one related post.  More tutorials are on the way.

News , ,

8. June 2012

 

This tutorial is going to look at the basics of sprites and an area that initially confuses many people ( myself included! ), the positioning system used by Cocos2D.  I am going to build off the code we wrote in Tutorial 2, with a few minor changes.  First thing you need to do is make a copy of the MyFirstApp folder and paste it as the aptly named MySecondApp, so you end up with a folder looking like this:

 

image

 

Now we need to make a few changes. First, you can delete the file MyFirstApp.js within the MySecondApp directory as we are going to create a new one now.  Create another js file, MySecondApp.js.  We will worry about the contents in a few minutes.

 

First we need to make a couple small alterations to our existing code.  Open up Cocos2d.js and scroll to the very bottom.  We want to change:

 

appFiles:['MyFirstApp.js']

to

appFiles:['MySecondApp.js']

Now open up main.js and change:

 

var myApp = new cocos2dApp(MyFirstAppScene);

 

to

 

var myApp = new cocos2dApp(MySecondAppScene);

 

That is all the changes we need to make to our boilerplate code.  In future tutorials, I will just assume you have already performed this step.

 

Finally before we get to coding, we need a sprite to work.  In your newly created folder, create a directory named images and save an image file there.  I am using this sprite, which has the file name Jet.png.  You of course can use whatever sprite you want.

jet

 

Now lets fill in the code for MySecondApp.js

 

var MySecondApp = cc.Layer.extend(
{
    init:function()
    {
        var layer1 = cc.LayerColor.create(
            new cc.Color4B(128, 128, 128, 255), 600, 600),
            jetSprite = cc.Sprite.create("./images/Jet.png");

        layer1.setPosition(new cc.Point(0.0,0.0));
        layer1.addChild(jetSprite);
        jetSprite.setPosition(new cc.Point(0.0,0.0));

        this.addChild(layer1);
        return true;
    }
});


MySecondAppScene = cc.Scene.extend({
    onEnter:function(){
        this._super();
        var layer = new MySecondApp();
        layer.init();
        this.addChild(layer);
    }
})

 

Much of this code should look virtually identical to MyFirstApp.js, as not much has changed.  Obviously our variable name has changed, but onEnter() is otherwise identical.

 

In our init() function, we are simply creating a background layer that is a solid gray color with the dimensions of 600x600 ( the same as our Canvas tag ).  We then create a cc.Sprite object using our Jet.png.  This is all that is involved in loading and creating a sprite object.  Now it's just a matter of adding the sprite to our layer, and adding our layer to MySecondApp’s layer. 

 

Lets load up our new app in the web browser and see what’s what!

image

 

As you can see, we have a 600x600 solid layer of gray and if you look closely in the bottom corner of our screen, you can see 1/4 of our jet sprite.  Depending on your background, this result might seem remarkably odd, as there are two factors at play here.

 

First, as I said in a previous tutorial, Cocos2D puts the origin at the bottom left hand corner, not the more traditional top left.  So if you have ever worked with textures or bitmap graphics before ( outside of OpenGL ), this is going to seem a little backwards to you.

 

Why would they do it this way?  I am guessing that  they followed the OpenGL way of doing things.  Now, why would OpenGL buck the trend and set the origin to the bottom left?  Well that involves a trip back to high school math and Cartesian coordinates.  If you think in those terms, suddenly starting from the bottom left makes a bit more sense.  Remember this graph:

 

2D_Cartesian_Coordinates

 

In this particular case, if you are dealing with positive coordinates, you are in the upper right hand quadrant.  So if something is at say, position (2,2), it is above and to the right of the origin.

 

In the end it doesn’t really matter why, just be aware that the positions in Cocos2D start at the bottom left and you will be fine.

 

The next factor here is, where is the sprite’s anchor point.  Its all nice and good to say our jet is at (0,0)… but what part of our Jet is at (0,0)?  This is where the anchor point comes into effect.  The anchor point is where sprite translations ( rotating, translating, etc ) are performed relative to.  In Cocos2D, by default the pivot point is in the middle of the sprite, like illustrated by the red dot:

jetpivot

 

It is traditional to place the anchor point at the top left corner, but there are some real advantages to placing it at the center.  The least of which is rotation.  Consider the effects of rotating a sprite 1 degree per frame.

 

Here is the result with the pivot about the center:

(The code for the above animation is available here )

 

 

And here is the same image with the sprite anchor point moved to the top left:

 

(The code for the above animation is available here )

 

 

If you want to alter the anchor point, you can do it with this code:

jetSprite.setAnchorPoint(cc.PointMake(0,1));

 

The parameter to setAnchorPoint() is the new origin, in OpenGL coordinates.  0,0 == Bottom left, 0,1 == Top left, 1,0 == Top right, 1,1 == Bottom right.  Truth of the matter is, you should just adjust to the Cocos way of doing things unless you have a very good reason.

 

Alright… back to example, doing things the Cocos2D way.

 

Let say for example we wanted our sprite to be at the top right hand corner, we would accomplish this with the following code:

var size = cc.Director.sharedDirector().getWinSize(); jetSprite.setPosition(new cc.Point( size.width - jetSprite.getContentSizeInPixels().width/2, size.height - jetSprite.getContentSizeInPixels().height/2) );

 

Now here is the updated init() method with the Jet sprite centered to the screen.

 init:function()
    {
        var layer1 = cc.LayerColor.create(
            new cc.Color4B(128, 128, 128, 255), 600, 600),
            jetSprite = cc.Sprite.create("./images/Jet.png");

        layer1.setPosition(new cc.Point(0.0,0.0));
        layer1.addChild(jetSprite);

        var size = cc.Director.getInstance().getWinSize();
        jetSprite.setPosition(new cc.Point(size.width/2,size.height/2));

        this.addChild(layer1);
        return true;
    }

 

And finally, our end results:

image

 

Again, you can download the entire project here.

 

 

Read Tutorial 4

Programming , ,

Month List

Popular Comments