I like this, but do we want to remove ALL objects? Is there a way to remote just everything but the most recently cached n objects?

@markrickert yeah that would be ideal, but from what i can tell NSCache is like angsty teenager and it's eviction policies are just suggestions. I'm still uncomfortable with this because SDWebImage should never be causing system memory events. I found this actually can be managed with #1140 At this point the implementer has clearly mismanaged his memory and it's already too late to be algorithmic about it.

For over a year I have been fighting a bug where a long session would run up the memory and crash/shutdown when a memory intensive action followed. I was always under the assumption that this was perfectly normal to run up this cache as the system would reclaim it when needed, i figured it was just some memory leak inside other libraries like GPUImage when in reality NSCache would drive up memory pressure and the device would send a memory warning, NSCache would ignore it, and the app would explode on a memory intensive alloc.

I hate this PR, but short of completely turning off NSCache as #1130 or rewriting a real LRU eviction policy, we are being terrible netizens at the moment.

A good start might be to leave a message in the README about carefully configuring SDWebImage on the context of the app and to carefully monitor memory so that maximum cost and count limit are set correctly

I agree, the right fix is to implement an LRU cache. I don't have the bandwidth to do that right know, but if someone want to send a PR, I have the feeling that a lot of people would be grateful.

In the meantime I'm gonna accept this PR as it is a good meantime tradeoff. Could you please remove the part of the PR that belongs to anothers?

closing for #1143