Computed property can't be data-bound from outside

[arthurevans]

When using a computed property, I can data bind it inside the element's template. However, it doesn't appear to be published such that an outside element can bind to it.

Test case:
http://jsbin.com/jifon/3/edit

The binding inside the element's template shows up. However, binding to it in the auto-binding template doesn't work. (Filtered value never appears.)

This seems wrong -- @rafaelw?

[sorvell]

I think we need to make this work, but it'll require some more sophistication in the binding system.

Right now a computed property is essentially 'bound' to the computation of its value.

[arthurevans]

Hm. Funny that it works correctly inside the element, though.

Given that this is a big gotcha, I'm inclined to put off documenting computed properties until it works, unless you folks feel otherwise -- @sorvell @sjmiles, I'm lookin' at you ;-).

[sorvell]

I suggest documenting how it works now. I think we can call this out as a
limitation that will be addressed soon.

On Wed, Jul 16, 2014 at 5:29 PM, Arthur Evans notifications@github.com
wrote:

Hm. Funny that it works correctly inside the element, though.

Given that this is a big gotcha, I'm inclined to put off documenting
computed properties until it works, unless you folks feel otherwise --
@sorvell https://github.com/sorvell @sjmiles
https://github.com/sjmiles, I'm lookin' at you ;-).

—
Reply to this email directly or view it on GitHub
#638 (comment).

[ebidel]

I think we should definitely document it now. Many people have asked about
computed properties.
On Jul 16, 2014 5:38 PM, "Steve Orvell" notifications@github.com wrote:

I suggest documenting how it works now. I think we can call this out as a
limitation that will be addressed soon.

On Wed, Jul 16, 2014 at 5:29 PM, Arthur Evans notifications@github.com
wrote:

Hm. Funny that it works correctly inside the element, though.

Given that this is a big gotcha, I'm inclined to put off documenting
computed properties until it works, unless you folks feel otherwise --
@sorvell https://github.com/sorvell @sjmiles
https://github.com/sjmiles, I'm lookin' at you ;-).

—
Reply to this email directly or view it on GitHub
#638 (comment).

—
Reply to this email directly or view it on GitHub
#638 (comment).

[jmesserly]

@sorvell just got me up to speed on this. I can take a look.

The issue is that computed properties are treated as a property binding (bindToAccessor) so they collide:
https://github.com/Polymer/polymer-dev/blob/master/src/instance/properties.js#L193

[rafaelw]

John, I actually prepared a patch for this prior to I/O, but the team didn't want any more changes while they were trying to stabilize:

https://codereview.appspot.com/102290043/.

If you want to take over this patch, I doubt landing it will be very hard.

[jmesserly]

oh, awesome! Thanks for sending that my way!

[jmesserly]

... and if you wanted to land it that works for me too :). But in case it requires merging I'm happy to revive it!

[arthurevans]

<happy-dance> !!! </happy-dance>

[jmesserly]

hmmm. I'm a little confused, but your original jsbin still seems to be failing even after the fix.
@arthurevans do you see that too? maybe I'm seeing an old cached copy, but polymer.js and platform.js both claim to be 0.4.1 ...

Maybe it's an issue with how the fix interacts with auto-binding. Just a guess because that's one difference between your jsbin compared to the test that was added.

[jmesserly]

Ah, nevermind, it looks like there was an (unrelated) change where computed properties are not considered "published" automatically. If I add "reversedvalue" to attributes= it works: http://jsbin.com/curiwocomopa/1/edit

[arthurevans]

Ah, yeah. I documented it that way: http://www.polymer-project.org/docs/polymer/polymer.html#computed-properties

One extra step, but it seems reasonable that you might not want to publish every computed property.