[gradebook2-dev] gb2 and dynamic localization

[John Bush]

ok, this is getting more interesting, apparently GWT deferfed binding
has a way to plug a java code generator into the framework.  If I'm
understanding this correctly, that means the calls to
GWT.create(I18nConstants.class) I could catch and generate the Impl
using the GeneratorContext, which appears to be a reflection
implementation of sorts.  Ok, GWT is freakin cool.

I think this might actually work.  So basically I catch these create()
calls with one of these generators, and then generate a impl that
makes calls the dictionary which avoid having to maintain anything
other than the interface.  Then I just make a server side component
that builds up the javascript to populate the Dictionary.  Very cool,
I think this might work, will try it out tomorrow.

see http://code.google.com/webtoolkit/doc/latest/DevGuideCodingBasicsDeferred.html


On Thu, Apr 14, 2011 at 11:09 PM, John Bush <john.bush at rsmart.com> wrote:
> Ok, missed this email, but just confirmed the bytecode manipulation
> and reflection are not part of the GWT client side support. So that
> rules out any fancy stuff and confirms my dislike for javascript, but
> I digress...
>
> I tried implementing the I18nConstants interface to delegate to the
> Dictionary and confirmed this works:
>
> public class I18nConstantsImpl implements I18nConstants {
>    Dictionary i18n;
>    public I18nConstantsImpl(){
>         i18n = Dictionary.getDictionary("I18nConstants");
>
>
>    }
>    public String actionActor() {
>        return i18n.get("actionActor");
>    }
>
>    public String actionDateFieldLabel() {
>        return i18n.get("actionActor");
>    }
>
> Then in the host html file, which is GradebookApplication.html you
> need to define the Dictionary like this:
>
>        <script>var I18nConstants = {
>     actionActor: "gwt rocks",
>         actionDateFieldLabel: "gwt rocks",
>         actionDescriptionFieldLabel:"gwt rocks",
>         actionDetails: "gwt rocks",
>         actionEntityFieldLabel: "gwt rocks"
>
> That's the bit I would replace with a call back to the server.  The
> issue is without bytecode or reflection dynamically implementing the
> I18nConstants interface is not going to be possible which would
> require devs to maintain the interface and impl, which I don't like.
> Need to think about this a little more, unless others have any idea. I
> really would like to avoid a giant refactor that doesn't  seem
> friendly, and would lead me to just push back on the requirements that
> are causing me to evaluate this in the first place.
>
>
>
> On Thu, Apr 14, 2011 at 7:29 PM, Thomas Amsler <tpamsler at ucdavis.edu> wrote:
>> See inline comments bellow:
>>
>> On Thu, Apr 14, 2011 at 5:27 PM, John Bush <john.bush at rsmart.com> wrote:
>>> I've been looking into a way to make gb2 integrate with the new
>>> dynamic loading of bundle data available in 2.8.  See
>>> https://jira.sakaiproject.org/browse/SAK-18678.
>>>
>>> I see GWT has another method other than the I18nConstants approach
>>> that allows for server side localization:
>>> http://google-web-toolkit.googlecode.com/svn/javadoc/latest/com/google/gwt/i18n/client/Dictionary.html
>>
>> Do you just want to use this approach on the server side? You mention
>> above " ...allows for server side localization".
>>
>>>
>>> >From looking at the code switching this out touch almost everything.
>>
>> We could try a refactor and see how bad it is.
>>
>>> So here's my crazy idea.
>>>
>>> Use javassist or some other byte code library to get around the fact
>>> that this call can't really be extended:
>>> (I18nMessages) GWT.create(I18nMessages.class);
>>>
>> Whatever tool/library you use, it would have to be able to build
>> within the GWT JRE emulation subset:
>>
>> http://code.google.com/webtoolkit/doc/latest/RefJreEmulation.html
>>
>>
>>> So basically I'd augment the class that GWT is generating and throw
>>> away what it does and instead delegate all the calls to the
>>> Dictionary, which can then look them up from the server.
>>>
>>> Crazy or possible, what do you think?  Going to play around with it a
>>> little tomorrow.
>>>
>>> --
>>> John Bush
>>> 602-490-0470
>>> _______________________________________________
>>> gradebook2-dev mailing list
>>> gradebook2-dev at collab.sakaiproject.org
>>> http://collab.sakaiproject.org/mailman/listinfo/gradebook2-dev
>>>
>>
>
>
>
> --
> John Bush
> 602-490-0470
>



-- 
John Bush
602-490-0470