$.extend(true, ....) won't work on sources that contain circular references.

[jdalton]

Try this page and open your web inspector console: ​http://jsbin.com/ehisub

      var object = {
        'foo': { 'a': 1 },
        'bar': { 'a': 2 }
      };

      var source = {
        'foo': { 'b': { 'foo': { 'c': { } } } },
        'bar': { }
      };

      source.foo.b.foo.c = source;
      source.bar.b = source.foo.b;

      var actual = $.extend(true, object, source); // stack overflow

I handle this in Lo-Dash here: ​https://github.com/bestiejs/lodash/blob/b5b8523257b46aafa180e9d6e256e6db2f3a5d2c/lodash.js#L2034-2045

using the stack array to hold objects I've iterated over and then return the corresponding assigned value.

[jdalton]

The current $.extend has a basic check to attempt to avoid stack overflows but it's not enough.

[dmethvin]

I'm not sure this is something we need to guard against. Keeping a stack of visited objects seems like overkill. It feels like our internal methods are being forced to handle cases we don't encounter in our own usage. When might we encounter something like this?

[dmethvin]

Our "internal" utility methods seem to be pulled into dealing with all sorts of edge cases lately. Thoughts?

[mikesherov]

-1, That's what Lo-Dash is for.

[Rick Waldron]

-1, I really appreciate the report, but I'd rather point users to Lo-Dash (as Mike noted)

[gnarf]

+0, If the impact to code size is small enough, this seems like one of those edge cases people might run into.

[jdalton]

Just to clarify deep $.extend is not an internal method. It's used by devs wanting a deep clone equiv.

The bug report was to fix an existing chunk of code in jQuery which seems to be in place to handle this issue (circular references), but does not. So this isn't about adding new support, it's about fixing the existing support.

Keeping track of visited objects is nothing new or extreme. Underscore does this in its _.isEqual method as well.

Lo-Dash, with its extra checks and circular reference support is still faster than jQuery (I'm guessing because Lo_Dash punts on sparse arrays). ​http://jsperf.com/deep-extend-perf#chart=bar

[dmethvin]

@jdalton, the reason I quoted "internal" was because yes we've documented this method and people are using it as a generalized any-kind-of-object extension mechanism where we even try to determine array-likeness to decide how to extend it. The method was really not originally designed to carry that load, it started as a simple method to merge simple plain objects. So it's just me expressing frustration that these methods get pulled into progressively more complex scenarios.

Would you be interested in filing a pull request for this?

[jdalton]

Would you be interested in filing a pull request for this?

Yap, can do, I'll try not to let this fall in the cracks.

[jaubourg]

#13526 is a duplicate of this ticket.

[dmethvin]

I'm inclined to let this be the domain of lodash, since nobody has demonstrated a case where our internal use needs to protect against it. I created an api docs ticket.

​https://github.com/jquery/api.jquery.com/issues/332

[dmethvin]

#14821 is a duplicate of this ticket.