Using $sce in a function to generate ng-include location leads to infinite digest

[KamBha]

Plunkr: http://plnkr.co/edit/A7jqE4pG8h9n5fPQXPl6?p=preview

Basically, if you have an ng-include which calls a function that generates an template URL which does something like this:-

    $scope.findTemplate = function(templateName) {
      return $sce.trustAsResourceUrl(templateName);
    };

Then you will get an infinite digest loop. This worked fine in v1.4.1, but stopped working in 1.4.2 onwards. I worked around this problem by caching the result from $sce.trustAsResourceUrl and ensuring I return the same object everytime.

I did not read anything in the changelog about breaking changes, so I thought I would check if this is how we should use $sce..

[gkalpak]

This was introduced in #12240. Needs further investigation.

[Narretz]

@KamBha does the stack strace provide any usable information?

Oh, I didn't see the plnkr.

[Narretz]

So the watch on ngInclude previously wrapped the watched src expression with parseAsResourceUrl which creates a function from the expression, which checks internally if the expression value is trusted in the context, and then returns the trusted value (string).
In #12240 the trusted check was moved into the templateRequest fn.
What that means now is that if you watch a fn that returns $sce.trustAsResourceUrl(templateName), it will return a new object on every digest, instead of a the src string, which causes infdig.

I think caching the result of the trustAsResourceUrl call is the recommended migration strategy. If you think about it, if you are trusting a specific url, you only need to do that once, not every time the url is requested.

I don't see an easy way to fix that otherwise.

[Narretz]

Actually, the easiest fix is to use equality watch on the include src. This might cause performance problems, though.

[gkalpak]

In this specific usecase, you could also remove the $sce.trustAsResourceUrl() call and use $sceDelegateProvider.resourceUrlWhitelist to white-list specific URLs.

[KamBha]

I am happy to continue to cache the result or use the delegate, but can you clarify if this applies to all methods on $sce. Also, is this breaking change going to be documented somewhere? Not sure what the policy is on retroactively modifying the changelog, but that sounds like the right place to do it.

[KamBha]

@gkalpak ,
Just want to confirm, if we do the white listing, we don't need to use $sce?

Thanks.

Kamal.