On Oct 28, 2013, at 10:33 AM, Matthias Wessendorf <matzew@apache.org> wrote:




On Mon, Oct 28, 2013 at 3:29 PM, Lucas Holmquist <lholmqui@redhat.com> wrote:

On Oct 28, 2013, at 10:27 AM, Matthias Wessendorf <matzew@apache.org> wrote:




On Mon, Oct 28, 2013 at 2:10 PM, Lucas Holmquist <lholmqui@redhat.com> wrote:
Hello Everybody,

Now that we have support for IndexedDB and WebSQL in our Javascript library,  i've started a fallback strategy for DataManager.

There will need to be a bit of deprecation since the 2 new adapters are asynchronous and return a promise/callbacks and the other adapters are synchronous and return the actual data.

We are thinking of adding an option,  "aysnc", the name is up for debate,  which would only be valid for the memory and session/local adapters and would default to false and things would return as normal( this would also be removed in the next release and everything would be "async" ),  but if true, then these adapters would return like the 2 new adapters, using callbacks/promises.

could we make it async only ? 

it would be once we get to 1.4.0(?) since we need to deprecate

that's fine w/ me!


Side note: I think we had _similar_ discussion in the past, and unfortunately ... we removed the async API ([1] from iOS).

we have no choice since IndexedDB and WebSQL are async



 


 


Now on to the Fallback Strategy:

Here is a use case:

I'm a user,  i create an IndexedDB store,  but the device/browser i'm using doesn't support it.  I think there are 2 options we can go down here


1. The user gets an error back about IndexedDB being not support,  just like regular. And they need to "opt-in" to the fallback stuff.  this would be an option on the "add" of the Store.  Maybe the actual Constructor?
They would also be able to provide a list of adapters to try which would override our predefined list.


2. Fallback is assumed and there is no option,  it just happens, using our predefined list.


I think it would be nice if the fallback just happens; However, there might be the case that folks have a certain preference in actually specifying the option. Not sure.
As of now I think I'd support the #2
 


There is still the question of if the fallback is used,  and a "Asynchronous" store falls back to a "Synchronous" one,  what should be returned.  I'm thinking what ever the return value of the "Async" one was.

I don't see this as breaking API since the "sync" store was not meant to be created "directly"

So if they create an IndexedDB store and it fallbacks to a memory adapter,  this store will use the "async" model

+1 makes sense
See question above :-)
 


Thoughts, comments, cat pictures,

-Luke




_______________________________________________
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


_______________________________________________
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