Optimising Android application before release [closed]

Answer 1

At some point you are going to get to the point where using known tricks will hit their limits. The best thing to do at this point is profile your code and see what areas are the bottle-necks based on your specific requirements.

Investigating RAM usage using MAT and Using Traceview: an article on how to use the tools to profile your application.

Answer 2

If you have network operations try re-using same httpclient instance. Avoid using regular expressions.

Answer 3

Optimize your PNG pictures with tools like OptiPNG and PNGCrush to shave off some (kilo)bytes from APK size. Image optimization tips for websites apply here too: use appropriate picture formats, play with JPG compression, consider using binary transparencies instead of 8-bit transparencies etc.

If you ship big PNGs with alpha channel, you can trade some APK size for startup speed and use separate JPGs for RGB and A channels.

If you're making HTTP connections, check that your HTTP client utilizes content compression. If it caches received HTTP responses, check that it understands and uses caching-related HTTP headers correctly.

Answer 4

Track and squash allocations. The more you allocate, the more often the garbage collector will need to run, stopping your process from doing anything else for relatively long periods of time, such as 100ms or so.

The best tool I know for this is the Allocation Tracker included in DDMS.

Not only GC can have an impact on the user experience, but superfluous allocations and GC do consume some computing resources.

Here's an example and a small trick. In my app, I have a clock which shows the current (audio) time, including tenth of seconds. This is updated often. And TextView performs allocations internally whenever you call setText() with a CharSequence. But it doesn't allocate anything with the setText(char[] text, int start, int len) variant. This isn't documented, and no one answered when I asked about it.

There are many ones like this. And this is one of the reasons why my app contains 50% native code (but there are other reasons).

Apart from this, I can recommend that you experiment with ProGuard. It performs several optimization passes, and logs such informations as unused methods within the project, which can help you removing leftovers in your code.

Answer 5

When using SQLlite, put special attention on indexes. Don't assume anything. I got tremendous speedups in Zwitscher, when I put indexes on columns commonly used for search.

Answer 6

Avoid using XPath if you can parse your XML input by using string manipulation routines to get the text between the tags. I have tested and can confirm a 10x improvement in this case, on a 50000 items data set, on a HTC Desire.