Common issues with Android Performance

Posted: 01 Nov 2021. Last modified on 04-Jun-22.

This article will take about 4 minutes to read.

Common issues with Android performance

Performance is a topic which is usually saved until the last minute, or treated as tech debt. This writeup goes over some of the common performance issues that happen in Android apps.

The most important one

Doing work on the main thread

Forgetting to move long-running work to background threads means that your app will hang!
Users will receive an ANR (App Not Responding) notification after 5 seconds of inactivity
Whenever you make a network call, access a database or do some other long-running task, make sure it is done asynchronously
Some options for this include rxjava or kotlinx.coroutines

Allocations in the onDraw() or onDispatchDraw() methods

All allocations should be made out of the draw loop!
Each frame will be wasting work, which increase the chance of a dropped frame
Increased allocations increase the risk of a garbage collection, which will create a longer skip

Overdraw

Overdraw happens when a portion of the screen is painted over multiple times within the same frame.
A common example of this is setting a background for an element that will always take up the full size of its viewport
This is wasted work! Why paint twice when only the results of one of them will show?
This can be visualized with the GPU overdraw tool in settings

Multiple Recyclerview Viewpools

When a recyclerview has other recyclerviews within it (think horizontal carousels), all of the views in the child carousel have to be reinflated when the carousel scrolls on screen.
This can be solved by sharing all views in the parent and child RecyclerViews, into a shared RecyclerView.RecycledViewPool

Nested RelativeLayouts and Constraintlayouts

The amount of time spent calculating the results of requestLayout() goes up exponentially when you have multiple Relativelayouts or ConstraintLayouts
There should be a top level layout that orchestrates the rest of the views instead. Use LinearLayouts or FrameLayouts to group nested content.

Unnecessarily resizing Recyclerviews on scroll

setting setHasFixedSize(true) on a RecyclerView means we can skip checking if the size of individual items changes when new items are added
Request layout called repeatedly within recyclerviews with only one type of view

Setting the background multiple times

It is possible to call setBackground repeatedly on the same views
This contributes to overdraw, which gives the GPU more work to do
If it spends more than 16.6ms drawing (60fps = 16.6ms * 60), the frame will drop and be emitted a frame late

Other UI issues

Overlays being drawn inefficiently
Not using fresco placeholders in favor of setBackground

There are more possible recyclerview optimizations, such as the ones outlined here: https://medium.com/google-developers/recyclerview-prefetch-c2f269075710

Cache miss

The data that we need is not being locally stored
If the app is making the same call time and time again, see if you can cache the response with a timeout, instead

Too much data

Some endpoints give back much more information than you will be able to use
Copying all of that information from the network takes time, and will make your app appear to be working slower
Try breaking a large endpoint into smaller, more focused ones, or moving to graphql, which will give you more control over what you request

Too little data

Some endpoints might require you to make calls to other endpoints as well, to get all of the data required to make a final submission.
Cart endpoints sometimes fall into this category
Make sure that you are making as many requests in parallel as you can, and show the user that you have received at least some information, and are waiting on the next set.

Lack of feedback

Users have a high abandon rate after 3 seconds of inactivity, so they need to know that you are making progress for them
Simple spinners can show that work is going on in the background

Notification of new data

A common pattern is to serve whatever data you have in your cache, while sending a request for new data from the server
The content will show immediately, but could be out of date
When the new data arrives, you can show the user a notification that lets them know their content is out of date.