The Picasso library (which we use to load images from the network) automatically caches the images in local storage. This is generally a nice feature, but it seems to cache the images without any bounds. We should investigate the possibility of hooking into Picasso's caching logic, and trim the cache when it exceeds a certain limit.
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | Dbrant | T87073 Out of space errors causing app to crash | |||
| Resolved | • Mholloway | T89834 Limit the filesystem cache size of Picasso library. |
Event Timeline
Comment Actions
Looks like this should be easy:
http://stackoverflow.com/questions/20090265/android-picasso-configure-lrucache-size
Comment Actions
Oh wait. The above link is just for RAM cache.
To limit disk usage:
https://github.com/square/picasso/issues/278
http://stackoverflow.com/questions/18944773/how-to-implement-my-own-disk-cache-with-picasso-library-android
Comment Actions
Per Bernd's link above, Picasso will simply use OkHttp's disk cache, if configured, so limiting its size looks like a matter of adjusting line 15 of OkHttpConnectionFactory.java:
private static final long HTTP_CACHE_SIZE = 16 * 1024 * 1024;
The question is, how much smaller do we want it to be? Is there an error or something that I can reliably trigger and then try to eliminate by adjusting the cache size downward?
Comment Actions
@Mholloway I guess you want to browse to a lot of pages and see if the app cache stops growing at some point.
It would be good to (at least temporarily) add some debug logging of the OkHttp Cache sizes. That should help get a better picture of it.
Comment Actions
I did some testing/logging. Under the hood, OkHttp's Cache is a DiskLruCache with some added optimization functionality, and it definitely maxes out at the value set in HTTP_CACHE_SIZE (currently 16MB).
Comment Actions
See my writeup in the task above.
tl;dr: I don't think this limit applies to the series of events causing the crashes.
Comment Actions
Based on the above comment, I'm fairly convinced that this is no longer an issue. My guess is that the reports we've gotten about running out of storage space were from devices that were already critically low on storage, and wouldn't have benefited from additional constraints on our caching.