Defered.setFinal

Discussion:

Defered.setFinal

Amit Mendapara

2009-02-10 13:30:36 UTC

Hi Per,

I have just started again improving the MochiKit Extensions. While
creating tests for the Ajax module, I found one problem (not bug, but
specific to the feature I'm trying to implement). I registered a
callback which should be fired at the end of all registered callbacks.

I achieved by modifying Async.js with a new method `setFinal` (not
addFinal as there should be only one finalizer) which gets fired when
`chain.length == 0`. It's simple. I would we happy if you add one such
function in MochiKit.Async.Defered...

Regards
--
Amit

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "MochiKit" group.
To post to this group, send email to ***@googlegroups.com
To unsubscribe from this group, send email to mochikit+***@googlegroups.com
For more options, visit this group at http://groups.google.com/group/mochikit?hl=en
-~----------~----~----~----~------~----~------~--~---

Per Cederberg

2009-02-10 16:04:56 UTC

Permalink

I think this is a good idea. I needed something similar too, so I
ended up writing an ugly hack that worked most of the time...

d.addBoth(function (res) {d.addBoth(finalFunc); return res; });

It adds new callback once the first deferred result drops in,
hopefully after all the other callbacks have been added...

But a more formally correct solution would probably be a good idea.

Cheers,

/Per

On Tue, Feb 10, 2009 at 2:30 PM, Amit Mendapara

Post by Amit Mendapara
Hi Per,
I have just started again improving the MochiKit Extensions. While
creating tests for the Ajax module, I found one problem (not bug, but
specific to the feature I'm trying to implement). I registered a
callback which should be fired at the end of all registered callbacks.
I achieved by modifying Async.js with a new method `setFinal` (not
addFinal as there should be only one finalizer) which gets fired when
`chain.length == 0`. It's simple. I would we happy if you add one such
function in MochiKit.Async.Defered...
Regards
--
Amit

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "MochiKit" group.
To post to this group, send email to ***@googlegroups.com
To unsubscribe from this group, send email to mochikit+***@googlegroups.com
For more options, visit this group at http://groups.google.com/group/mochikit?hl=en
-~----------~----~----~----~------~----~------~--~---

Bob Ippolito

2009-02-10 16:48:44 UTC

Permalink

Finalizing a Deferred should ensure that no further callback/errbacks
are registered and it should attach a default error handler (success
should be no-op). The most common problem I've seen with Deferreds is
that an error occurs but nobody attached an error handler that far
down the stack. In Python they work around this by having a finalizer
so that you see the error when the object gets GC'ed, but that's not
really possible in JS since you don't have finalizers or weak
references.

Post by Per Cederberg
I think this is a good idea. I needed something similar too, so I
ended up writing an ugly hack that worked most of the time...
d.addBoth(function (res) {d.addBoth(finalFunc); return res; });
It adds new callback once the first deferred result drops in,
hopefully after all the other callbacks have been added...
But a more formally correct solution would probably be a good idea.
Cheers,
/Per
On Tue, Feb 10, 2009 at 2:30 PM, Amit Mendapara

Amit Mendapara

2009-02-10 17:58:20 UTC

Permalink

Per,

Your hack won't work with my case as it would mark the deferred as
`chained` so consecutive addCallback/addErrback won't work. Considering
Bob's suggestion, I modified my patch to ensure no further
callback/errbacks can be registered.

Regards
--
Amit

Post by Bob Ippolito
Finalizing a Deferred should ensure that no further callback/errbacks
are registered and it should attach a default error handler (success
should be no-op). The most common problem I've seen with Deferreds is
that an error occurs but nobody attached an error handler that far
down the stack. In Python they work around this by having a finalizer
so that you see the error when the object gets GC'ed, but that's not
really possible in JS since you don't have finalizers or weak
references.

Per Cederberg

2009-02-10 19:52:30 UTC

Permalink

You mean like this?

setFinalizer: function(cb) {
this._finalizer = cb;
}

...

if (this.chain.length == 0 && this._finalizer) {
this.finalized = true;
if (res instanceof Error) {
this._finalizer(res);
}
}

Would that be acceptable to Amit as well? In that case we would add a
default error handler (called a "finalizer"), which would also be what
I'm planning to use this for.

Other opinions?

/Per

Amit Mendapara

2009-02-11 05:58:06 UTC

Permalink

It's not just error handler, it should always be called at the end
once result is available and all callbacks/errback (including default
errback) are fired.

setFinalizer: function(cb) {
this._finalizer = cb;
}

if (this.chain.length == 0 && this._finalizer) {
this.finalized = true;
this._finalizer(res);
}

--Amit

Post by Per Cederberg
You mean like this?
setFinalizer: function(cb) {
this._finalizer = cb;
}
...
if (this.chain.length == 0 && this._finalizer) {
this.finalized = true;
if (res instanceof Error) {
this._finalizer(res);
}
}
Would that be acceptable to Amit as well? In that case we would add a
default error handler (called a "finalizer"), which would also be what
I'm planning to use this for.
Other opinions?
/Per

Amit Mendapara

2009-02-11 12:52:21 UTC

Permalink

Even better...

if (this.chain.length == 0 && this._finalizer) {
this._finalized = this._finalizer(res);
}

This will allow us to finalize the deferred based on some
conditions...

- Amit

Per Cederberg

2009-07-20 09:26:12 UTC

Permalink

Slow response here, but I finally got around to committing this to
Subversion (r1533):

http://trac.mochikit.com/changeset/1533

Modified the patch a bit further to account for some additional cases.
Also added tests and documentation.

@Bob: Can you please review this change? I have the feeling that I've
still missed one or two corner cases... Also, the mochikit.com and
trac.mochikit.com site seems slow from here... High server load
recently?

@Amit: Please check that my additional controls doesn't break you use case.

Cheers,

/Per

Post by Amit Mendapara
Even better...
if (this.chain.length == 0 && this._finalizer) {
this._finalized = this._finalizer(res);
}
This will allow us to finalize the deferred based on some
conditions...
- Amit

Amit Mendapara

2009-07-20 10:03:55 UTC

Permalink

Hi Per,

I have just tested the changes with the Ajax extension module and it
works as expected. Thanks for considering adding this feature. This
would help me creating jQuery compatible Ajax module for Mochikit
extensions project.

Currently, I'm quite busy with some another work but supposed to come
back on MochiKit Extensions soon...

Regards
--
Amit Mendapara

Post by Per Cederberg
Slow response here, but I finally got around to committing this to
http://trac.mochikit.com/changeset/1533
Modified the patch a bit further to account for some additional cases.
Also added tests and documentation.
@Bob: Can you please review this change? I have the feeling that I've
still missed one or two corner cases... Also, the mochikit.com and
trac.mochikit.com site seems slow from here... High server load
recently?
@Amit: Please check that my additional controls doesn't break you use case.
Cheers,
/Per

Giulio Cesare Solaroli

2009-07-25 22:51:59 UTC

Permalink

Hello all,

In order to handle this case, we have overwrote the 'callback' method
in our subclass of the Deferred class, to add a default Errback that
at least logs the error.

A similar solution could be probably used also in MochiKit, allowing
the user to optionally specify a specific error handler / finalizer.
This will allow any further check, as once the Deferred is triggered
it is already impossible to add further callback/errback.

I will check anyway the fix added to the repository to see how to
improve our current extension with the committed fix.

Giulio Cesare

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "MochiKit" group.
To post to this group, send email to ***@googlegroups.com
To unsubscribe from this group, send email to mochikit+***@googlegroups.com
For more options, visit this group at http://groups.google.com/group/mochikit?hl=en
-~----------~----~----~----~------~----~------~--~---