May 232012

We recently started on a fairly large mobile app. Our previous apps have been written using Ansca’s CoronaSDK ( For the most part Corona has met our needs, however recently we have had a lot of issues with stability. These issues have caused some negative reviews in the iTunes App Store. Also, there are some other features we would like to take advantage of that Corona does not currently support and may never support. There is no clear timeline for upcoming features.

Stability Issues
Some of the stability issues in the past few months include:

  • Full Screen Inneractive Ads lock up the app. This was recently fixed, but I reported this back in January. It is now May. Given that Ansca and Inneractive have both pushed using the Inneractive ad system since it was first introduced, you would have thought that this would have been a high priority bug to fix and not one that took 5 months to address.
  • Nook Color crashed at start up. Fixed in Build 819, but all that was required was a simple regression test to know it was broken. Read more here.

Additional Features Needed
These items really aren’t anything to complain about since there are plenty of apps that have no need for the same things we do, but for us these are issues worth noting:

  • LUA modules cannot be placed in sub folders
  • Texture memory is not optimized
  • Unable to extend the Corona API

LUA modules cannot be placed in sub folders
One of the key elements in our upcoming app is that it is essentially a collection of games in a single app. This means a LOT of assets, and a lot of different code modules to handle each games logic. Currently Corona does not support putting lua files in sub folders for all target devices. I’ve had success getting that to work on iOS devices, but it fails miserably on Android devices.

Texture memory is not optimized
Another issue with Corona has to do with texture memory. Corona uses one of two functions to load a texture:

-- Assume the image is 1024x768 and we want to display it
-- on a 480x320 device

local image1 = display.newImage( "image1.png" )
image1.scaleX = 480/1024
image1.scaleY = 320/768

local image1 = display.newImage( "image1.png", 480, 320 )
-- no need to scale since that is what the two params did,
-- BUT it still takes the same memory as the first call

In both cases the memory that the system will use for the texture is 1024x768xdepth. Ideally, we would want the second version to only take up 480x320xdepth bytes. This would mean we could have ONE high resolution image and use it on all platforms. Yes, we can do that now, but on the low resolution devices it will take the same amount of in app memory as the high resolution devices. This means that we have to create separate versions of the art for each resolution we want to support in order for the game to fit into memory. I really do not understand why they implemented it this way when all that had to be done was load the image into memory, resize it in code, then use the resized image for the actual texture display memory.

So the way we handle it is to create low, medium and high resolution versions of most images and put those in folders named “low”, “medium” and “high”; and a few we just have one version and dump those in a “common” folder.

Corona has something similar with their dynamic content scaling solution. However, this doesn’t address the texture memory issue, nor the bloat that comes from having to have multiple versions of key images.

With a proper load function we would be able to just have a common image and use it for whatever resolution we want by dynamically scaling the width and height for the actual device display resolution. Nice and memory efficient and we don’t double our APK or IPA size with copies of assets at different resolutions.

Unable to extend the Corona API
This issue along with the stability problems are probably the biggest reasons to consider moving away from Corona. Because Corona is not open source, we have to way to add features using platform API’s. I realize that is probably not the typical use of Corona as it was really targeted for Game Designers and not Game Programmers.

It would be fine if the community could request a feature and have a reasonable expectation that Ansca would add it to their API, but something as simple as pixel access to textures is still not in Corona and people have requested it for over a year.

Part 2 coming soon

P.S.: Like this article?! Want to support?! Be free to donate with bitcoins! 19r39vr2zs14aHW6DMNtPjmRga9xERQTTP

 Posted by at 2:28 am

  4 Responses to “Moving away from Corona SDK – Part 1”

  1. Have you made the switch from Corona, if so what cross platform SDK did you end up using. We are still deciding on whether to use Corona, and your post has certainly made us think twice..

    • We opted for Cocos2d-x. This is a C++ implementation of Cocos2d. We just completed a 4 month project using it and I have to say that it beats the Corona SDK by a long shot.

  2. I have used sub-folders extensively to organize the LUA files in Corona, never had any problems with this.
    We deploy to both Android and iPhone, and sub-folders never caused an issue.

  3. Kenneth, Corona finally got around to supporting this a few months back. I haven’t updated the post because we switched to Cocos2dx.

Sorry, the comment form is closed at this time.