Possible memory leak in ngRepeat

[daanporon]

I think i found a memory leak issue in angular 1.2.x at first we where working on 1.2.23 where i had the issue then i saw we where behind and updated to 1.2.26. But the issue was not resolved.

I created a plunkr where it's possible to reproduce the problem.
http://plnkr.co/edit/pA43qJd92eCxMLbtWUf5?p=preview

I have 2 states, state A and state B and on state B i have an ng-repeat. When i transition from state A to state B and then back to state A. The scope of the controller of state B is still in memory. When i put my ng-repeat inside another directive which creates it's own scope (isolated or regular) it's also leaking the scopes of every item in the ng-repeat. When i remove the ng-repeat from my state- or directive template the controller is GC'd correctly.

Also tested this with an ng-if and ng-switch and couldn't reproduce the problem with them.

Also tested it by upgrading to the latest angular 1.3 and there the issue seems resolved, but i still have to support IE8 so :(

I thought it had something to do with this PR (#9281) so i tried backporting the changes to an 1.2.26 version, but this doesn't solve it.

Anyone any idea what could be the issue or can someone point me in the right direction? Already spent a few hours trying to solve this, and i'm a bit stuck :(

[Narretz]

Hi, have your tried with the most recent version, 1.2.26?

[daanporon]

@Narretz yup i have the same issue there ... also the plunkr i created is based on 1.2.26 ...

oops, modified my issue i was talking about 1.2.23 and 1.2.26 :/ stupid typos ...

[petebacondarwin]

I ran this plunker in Chrome and watched the timeline...

When I hit Garbage Collect, the memory dropped back down.

[petebacondarwin]

Moreover if I just keep alternating long enough then the GC kicks in automatically and clears everything out...

[petebacondarwin]

My specific fork of your plunker is http://plnkr.co/edit/0HLmALy9AgadDPkAjAcp

Can you provide more evidence of the memory leak?

[daanporon]

@petebacondarwin thanks for looking in to this. I attached a screen why i think this is a memory leak.

• Snapshot 1 is a heap dump on page 'foo'
• Snapshot 2 is a heap dump on page 'bar'
• Snapshot 3 is a heap dump on page 'foo' again

Now you can see that the scopes which where made when getting to page 'bar' are still available while you are on page 'foo' again. In Angular 1.3.x i don't have this problem. Also in angular 1.2.16 the problem wasn't there ... In angular 1.2.17 the cached class was added for the childScope and since then i noticed this problem.

In this case this scopes and detached dom nodes won't take up much memory, but we have a large application with large ngRepeats and our memory keeps increasing ...

[daanporon]

I created a second plunkr:
http://plnkr.co/edit/7R534RcXt4NSjDSmOGQ9?p=info

In this plunkr i tried to identify a second problem. When i tried to put the fix for the cached scope class in our app, i saw this didn't resolve it. It was resolved in plunkr but not in our application. After some research i came to the following conclusion. When there is an ngRepeat in the page with an undefined or empty collection the scopes of my other ngRepeat aren't GC'd.

There are flags in the index.html

• cacheScopeClass: when you set this to true it will use the Scope class caching as is done in angular 1.2.17+ and the GC will keep the scopes ... when you set this to false the code from 1.2.16 will be used. By doing this the scopes won't be leaked anymore.
• undefinedNgRepeat: when you set this to true the array of the ngRepeat will be set to undefined and the Scopes of the ngRepeat in the state won't be GC'd correctly. When you set this to false an array with items will be used and everything is GC'd correctly (at least when cacheScopeClass is also set to false).

This is the first time i'm doing memory profiling in the browser ... and it's driving me nuts :p can't find anything useful ...

In most cases this probably won't be a big issue, but we have big tables with lots of data on them ... and very complex DOM. And if this isn't GC'd correctly it becomes a big issue :(

[daanporon]

It seems i'm also having the same issue as the first one in angular 1.3 (the one with the cached class) ... i thought i tried it before and it didn't happen there, but now i can keep reproducing it in 1.3.3 also ...

The one with the empty/undefined ngRepeat isn't happening in 1.3

[daanporon]

I identified the second issue to be resolved in v1.3.0-beta11(commit hash b87e5fc) ... but there are a lot of changes made to transclude and stuff ... so i don't think this will be easy to backport this to 1.2.26 :(

[mrmarktyy]

Having very similar experience, just wondering what's the status of this issue ? is it confirmed still an issue in 1.3.x ? thanks

[Narretz]

Closing, as this is an old issue without a clear repro applying to old versions.