One of the problems of mobile development is that it’s a human resource driven activity that’s expensive. Multiply that by iOS and Android and the costs double (or more). Wouldn’t it be good if you could develop on Android or iOS and a tool could automatically generate for the other platform? Well, MyAppConvertor has that ambition. Really. This isn’t some limited, poor html-based or intermediate language-based ambition. It’s the real thing - Objective c to Java and vice versa.
MyAppConverter presented at the recent DroidCon UK 2014. You can signup with SkillsMatter to see the presentation.
MyAppConverter is currently in beta and is limited to online conversions via their web site. It only works for games, only converts from iOS 7.1 to Android 4.4.2 and only on Samsung Galaxy S5. However, the aim is to eventually support IDE-based conversions, for all types of app, to and from all major mobile platforms and versions of those platforms.
The presentation outlines their approach based on models and meta models to not only directly map code but to also do it in such a way that creates the right algorithmic and UI idioms/patterns on the respective devices. While the converter can create source code for completion by a developer, they are aiming for 100% conversion so that anyone can use the tool. Anything less is much less useful as it then needs a developer and the ‘extra stuff’ the developer adds needs to be somehow added every time there’s a new version of the app which would get difficult to manage.
The intention is impressive and hugely ambitious. However, it’s so ambitious I question whether it’s viable. The Droidcon presentation hints that they have already spent 5000 man days on iOS to Android transformations, just for games APIs. Games are the simplest, low hanging fruit, case as they don’t use that many APIs and are mainly drawing code.
This isn’t just about mapping APIs which is complex in itself. It’s also about knowing all the nuances of the APIs that are documented in places like stackoverflow. It’s also about mapping idioms and patterns of doing things, somehow mapping 3rd party libraries, knowing fragmented differences between (target) devices and knowing differences in APIs across OS versions. Even then, it won’t cover apps that use ‘real native’ c/c++ NDK code - previously estimated by Intel to be of the order of 25%. Once all this is solved, consideration has to be given to keeping all this up to date into the future. APIs can become deprecated over time and new API’s are created. For example, in Android Lollipop there are over 5000 new API’s. New devices will also become available with new nuances.
I don’t think what they are trying to achieve is possible in a reasonable time and/or with a reasonable team size. The World really needs tools like MyAppConverter. I sincerely hope they prove me wrong.
If I am wrong and the tool comes to fruition then there are some interesting possibilities MyAppConverter might not have considered. It might be possible to create a new tool to create apps by visually (or otherwise) building up the pre-existing meta models. These meta models, together with some kind of configuration data (the static, unchanging data in an app), would then be able to auto generate to multiple platforms with no actual developer coding thus putting developers such as myself out of business. Lay people, capable of wiring things together, could create complex apps.