Jacopo,
I did not intent my message to propose XAP as a replacement for OfBiz
screens/forms, but to present its possibilities.
Comments follow...
Jacopo Cappellato escribió:
Daniel,
thank you for your interest and comments. What I had in mind was not
to replace the form widget but instead to integrate XAP with it.
I'm sorry, I don't have a clear picture but there could be more ways
to do this:
1) including the XAP tags in form widget definitions
I haven't thought of this. I am not really aware of the effort to
integrate and maintain with the XML forms, but I don't like the idea of
mixing tags, but that's probably just my opinion.
2) let the form widgets generate XAP tags and then pass the output to
the XAP parser to get the final output (Ajax pages)
I have thought of this, but I doubt about the advantages over making the
same with dojo or prototype.
--
Daniel
My 2 cents,
Jacopo
On May 7, 2008, at 7:42 PM, Daniel Martínez wrote:
Jacopo,
From what I have looked about XAP (since your mail) it seems to me as
it could be used a replacement for OfBiz screen/widgets. Its most
interesting features are the declarative UI in XML (like OfBiz
screens/forms, except for the AJAX ;) and the independence of AJAX
library.
XAP is what I would like OfBiz widgets to be :)
--
Daniel
Jacopo Cappellato escribió:
Adrian,
it is really great to see you are putting effort on this.
As a side note, as I've already mentioned this in one of my mails
some time ago, I'm still wondering if the usage of XAP (one of the
incubating projects at Apache) could help us in this effort.
Unfortunately I had no time to seriously look into it but if you are
interested you can get a quick overview of the tool here:
http://incubator.apache.org/xap/overview.html
Jacopo
On May 7, 2008, at 4:48 PM, Adrian Crum wrote:
David,
I agree that there are some cool Ajax like features that won't map
exactly to existing widgets. In that case we'll create a new widget
and try to find an HTML equivalent. So, I'm not tossing out the
idea of new widgets that have improved features, I'm suggesting
let's start with adding new features to existing widgets.
I have seen the Ajax work done in the Example component.
Regarding the alternate HTML rendering classes, I don't think those
will be needed. My thinking right now is to just evolve the
existing HTML rendering classes.
-Adrian
David E Jones wrote:
Adrian,
This sounds great for the elements that have some sort of natural
JS/AJAX extension that doesn't change what the form means or
offers to the user but instead just improves the user experience
and/or efficiency.
With cases like I agree it would be great, and a REALLY cool
feature, to upgrade automatically and not require form changes or
anything.
However, there are many cases where we can't automatically add
JS/AJAX extensions, but instead they represent a possible widget
that is different enough from the concept behind any of the
existing form field types that it would be weird to piggy back the
functionality and try to automatically shoe-horn it into the
existing functionality.
Some of the stuff I played with recently and added, like an
auto-refresh on a screen container, is not something we would want
to automatically turn on. The other one I added recently, to
submit a form in the background and not refresh the page, is also
something that I don't think we would want to automatically turn on.
So, yes, I agree we should add some of the automatic extensions
that we can and it would be a really cool set of new features.
Stepping back to the original thingy, how does using alternate
HtmlFormRenderer classes help with this? Maybe you're not still
considering that, but I'd say if we did default fancy things
(which again, I love the idea of), we should just do those all the
time unless, like you wrote, the browser identifier is clearly one
that won't support it (some of that might need to be client side
too though... I'm not sure about all of the nuances there).
-David
On May 6, 2008, at 3:42 PM, Adrian Crum wrote:
I'm resurrecting this thread because I've spent some time looking
into the whole third party rendering library support idea and I
think I have a simple solution.
I thought about David's suggestion of having new widgets that are
effects based. I don't think that will be a good strategy because
not all browsers will have javascript enabled - which would
render those widgets useless.
A better approach I would like to propose is to use the Prototype
javascript library in combination with EXISTING widgets to
improve their response and functionality. The widget rendering
code would detect if the browser supports javascript, and output
the correct HTML to accommodate the browser.
Instead of a "live-form" widget, the existing form widget would
detect browser support, and render an improved form if the
browser supports it. The current paginated tables would use Ajax
calls to scroll through pages instead of refreshing the whole
screen.
Basically, I'd like to see the cool effects and improved response
implemented without any additional work on the widget XML files.
What do you think?
-Adrian
David E Jones wrote:
I guess this is a continuation of the discussion in the thread
"uilabels and screenlet widget", and is related somewhat to part
of the stuff in issue OFBIZ-1648.
The general goal of the widgets is simple: no platform specific
artifacts. Unfortunately this isn't entirely possible, which is
why we have a very big and ugly "platform-specific" tag to
delineate things that are not generic and provide for the
possible of having alternative platform things specified
together so that when rendering for a different target the
appropriate option can be selected.
As far as that applies to this topic, I'd say the best approach
is to never have any element or attribute called "dojo" or
"ajax" or "rico" or anything. In the dojo attribute for the
container elements, I'm not sure what you'd propose to put in
it, ie the "some Dojo data", but in general I'd prefer to never
have anything that is so dependent on a particular underlying
technology, the widget artifacts gain efficiency by their focus
on different effects, with the underlying software taking care
of the "causes", or rather how the effects are brought about.
In other words while we wouldn't want elements that have
anything to do with "dojo" or "openrico" we would want elements
to describe the effects from those libraries we'd like to have
available through the widget, and the most appropriate is
probably the Form Widget with different form and field types
(though some would certainly go elsewhere and are not form
related).
Examples of that would be a new form type like "live-grid" or a
new form field type like "live-combobox" (or "dynamic-combobox"
or "server-side-combobox" or something). If we add elements like
that then it doesn't matter which AJAX library we use underneath
and generate HTML/etc for, and we can change libraries without
requiring any change to the higher level artifacts, like the
form definitions.
-David
On Feb 16, 2008, at 1:34 PM, Adrian Crum wrote:
In order to accommodate 3rd party rendering libraries (Ajax,
Dojo, etc) in the screen widgets, we need to discuss how that
support will appear in the screen widget XML files.
I'll start things off with a suggestion I made in another
thread. Everyone is welcome to join in and offer their ideas.
When we reach an agreement, we can submit the results to Jira
and begin building it out.
I was thinking we could simply extend the existing widgets with
additional attributes. The new attributes would pass 3rd party
specific data to the rendering classes. The new attributes are
ignored by rendering classes that don't need them. All
rendering classes render all widgets in some form - some
rendering classes might have additional bells and whistles
based upon the additional attributes, while others downgrade
gracefully and still provide a usable screen rendering.
So, the widget XML would look something like this:
<container id="some-id" style="some-style" dojo="some Dojo
data" ajax="some Ajax data" foo="some foo data">
...
</container>
The additional attributes could be applied to any screen widget
element, not just the container element.
The advantage I see to this approach is it is fully backwards
compatible. We can add attributes to any screen widget element
without breaking existing rendering code.
That's it. Like I said, please add your ideas.
-Adrian
---------------------------------
Looking for last minute shopping deals? Find them fast with
Yahoo! Search.
