On Feb 14, 2013, at 7:38 AM, Matthias Wessendorf <matzew@apache.org> wrote:

On Thu, Feb 14, 2013 at 2:27 PM, Kris Borchers <kris@redhat.com> wrote:

On Feb 14, 2013, at 2:37 AM, Christos Vasilakis <cvasilak@gmail.com> wrote:

Hi,

as a follow up to my previous email, I want just to pinpoint that the issue that exists in Android where the Activity is killed during (e.g orientation changes of the device) doesn't exist in the iOS land. That is,  there is no need to reattach our view (and callback) to the running pending request.

In cases where a request is pending, and the application goes in the background (e.g. by clicking the home button), the View is not killed, when the data arrives the view is updated normally (even if it's not in the foreground).

In the case where the system decides to kill us immediately after clicking the home button, we can't do much, clicking the application again will restart the app altogether

I believe JS is similar though I'm not sure about actually continuing to update the view (I don't think it does) but if the browser is killed, there is not much we can do.

having a cancel on the pipe 'interface' does hurt, right ?
Think Cordova, once the WebView goes away, there is a hook, and you
could cancel that request(s). Shouldn't be too hard to offer that
wrapper (instead of leaving the user mixing our pipe and the
jQuery.ajax API)

Well, I'm just not quite sure how I would implement it. The Pipe has no idea of the state of a read or save or remove action. Once that method is called, everything is passed off to the Promise which is immediately returned by a call to one of those methods. Any cancel would have to happen on that object and not as a method on the Pipe which then it becomes to I call cancel or abort. See this gist for an example of what I'm talking about.

https://gist.github.com/kborchers/040710a345941c0abcd4


-M



Thanks,
Christos


On Feb 14, 2013, at 9:31 AM, Christos Vasilakis <cvasilak@gmail.com> wrote:

Hi Summers,

+1 for option #1 using the loader api and support library

As you suggest, android support library is commonly used nowadays, so won't have big impact targeting old versions of Android

Thanks,
Christos

On Feb 14, 2013, at 12:47 AM, Summers Pittman <supittma@redhat.com> wrote:

Currently, Pipes on Android are implemented using an AsyncTask and the
callbacks are fired on the UI thread.  In the case where the Activity
which made the request which launched the task is killed, an application
using Aerogear will crash.  This is bad. Additionally the internal
mechanisms of the request itself aren't easily reachable by the
developer.  This means we can't get status information (is the request
completed, is there data, cancel the request, etc) easily.

Last week at the F2F I chatted up Matzew and a few other guys and put
together a Pipe proposal (http://bit.ly/12MzeOx).  I've implemented it
and ported some example code and it works.  However, it requires a lot
of boilerplate code to manage the Activity life cycle correctly.  We can
abstract the boilerplate if we write our own extensions to the base
Android activity classes and our users extend those classes.  However
there are about 12 common classes in multiple libraries which can be
used and it is pain to maintain.

In Android 3.0+ Google provides a Loader API which handles this
boilerplate on its own.  Even better, they've extended their base
classes for us.  After doing more research today, I've come to the
conclusion that Pipe and friends can lean on Loader to do what we want
with minimal if any API changes.  However, this means on Android 2.3 our
users will have to include the Android support library.  This is usually
used by most projects anyway, but it will be something to point out in
the docs.

So my questions to the list is, which should be done?

In order of my preference:
1. Use the Loader API and require that users targeting older versions
of Android include the support libs (which they are probably doing already)
2. Use the boilerplate heavy API proposal and pass the coding onto the
user
3. Make our library 3.0+ only
4. Extend all the base classes and manage the boilerplate

3 & 4 aren't really good options, but I am including them for completeness.

Summers

_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev



_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev


_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev



--
Matthias Wessendorf

blog: http://matthiaswessendorf.wordpress.com/
sessions: http://www.slideshare.net/mwessendorf
twitter: http://twitter.com/mwessendorf

_______________________________________________
aerogear-dev mailing list
aerogear-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev