Only attempt unbind when __bindEvents is defined

[akagomez]

An error is thrown if unbind is called on a compute that has no bindings.

This PR addresses the issue by returning early when __bindEvents is falsy, as is done in event.js here: https://github.com/bitovi/canjs/blob/master/event/event.js#L196-L198

Closes #1779

[justinbmeyer]

This error can and has served as a good warning that something is wrong. Are we sure we want to remove it?

Sent from my iPhone

On Jul 12, 2015, at 11:15 PM, Chris Gomez [email protected] wrote:

An error is thrown if unbind is called on a compute that has no bindings.

This PR addresses the issue by returning early when __bindEvents is falsy, as is done in event.js here: https://github.com/bitovi/canjs/blob/master/event/event.js#L196-L198

Closes #1779

You can view, comment on, or merge this pull request online at:

#1780

Commit Summary

Only attempt unbind when __bindEvents is defined
File Changes

M util/bind/bind.js (3)
Patch Links:

https://github.com/bitovi/canjs/pull/1780.patch
https://github.com/bitovi/canjs/pull/1780.diff
—
Reply to this email directly or view it on GitHub.

[akagomez]

I can't imagine we'd want to throw that error for the case I outlined.

What others exist? Can you think of some off hand?

[justinbmeyer]

I'd want an error thrown for that case.

Sent from my iPhone

On Jul 13, 2015, at 6:24 AM, Chris Gomez [email protected] wrote:

I can't imagine we'd want to throw that error for the case I outlined.

What others exist? Can you think of some off hand?

—
Reply to this email directly or view it on GitHub.

[justinbmeyer]

Calling unbind without arguments is invalid. Calling unbind that doesn't actually unbind is invalid.

I've never been a fan of how jQuery swallows errors. I don't think we should do it too.

Sent from my iPhone

On Jul 13, 2015, at 6:24 AM, Chris Gomez [email protected] wrote:

I can't imagine we'd want to throw that error for the case I outlined.

What others exist? Can you think of some off hand?

—
Reply to this email directly or view it on GitHub.

[akagomez]

Maybe there's a better alternative to my use case then. How I can I safely remove all bindings from a compute, regardless of whether it is or isn't bound to already?

[justinbmeyer]

Maybe is misunderstood what you are trying to do. Removing the error is different then making it possible to not pass an event handler and have all handlers removed.

Sent from my iPhone

On Jul 13, 2015, at 7:42 AM, Chris Gomez [email protected] wrote:

Maybe there's a better alternative to my use case then. How I can I remove all bindings from a compute?

—
Reply to this email directly or view it on GitHub.

[akagomez]

You're right. I assumed that calling .unbind() without any arguments would remove all of the bindings. After getting this error I then assumed that if there were bindings to be removed, they would be.

I could swear I've confirmed this before. But now I'm not sure. I probably got this mixed up with how can.Map can unbind all handlers bound to a property: http://jsbin.com/yomimiyuce/edit?js,console,output

[akagomez]

One slight clarification. Calling .unbind() without any arguments does in fact not remove any bindings.

You can remove all of the bindings of a particular type by calling .unbind([type]).

That said, doing so will still throw an error if there are no bindings at the time .unbind() is called on the compute. This PR addresses that specific issue.

Also, test added.

[justinbmeyer]

Looks good. @akagomez anyway you can resolve the merge conflicts, make sure travis passes, and merge this into master?

[akagomez]

Rebased. Tests passing. Good to merge.

[justinbmeyer]

@akagomez cc @daffl I think I misunderstood what this pull request does. @akagomez can you confirm that it does not help actually remove all event bindings? If this is the case, I still don't understand why this is useful. An error is useful.

[akagomez]

The confusion may be due to the fact that I opened the issue for the case .unbind(). That was in error. Calling .unbind() without an event "type" does nothing. See here: https://github.com/bitovi/canjs/blob/master/event/event.js#L199

I attributed that error to calling .unbind() without a type. In reality it was present even if a "type" was defined:

var c = can.compute(32);
c.unbind('change'); //-> "TypeError: Cannot read property 'undefined' of undefined
                 at can.unbindAndTeardown [as unbind] 

This PR prevents an error from being thrown if .unbind("type") is called and there are no bindings, which is consistent with most "remove" api's that I'm familiar with (e.g. el.removeClass('class'), map.removeAttr('attr'), etc)

Making .unbind() remove all event handlers of all types was not the intention of this PR.