Resuscitating this thread, as I'd like to post a patch.
While I understand that it's error-prone to merge in two places, I'm not
sure we have a choice here. Prefs passed on the fragment are by definition
not available to the server, so can't participate in a server-side merge.
Passing all prefs on the query string obviously kills caching. I see two
options:
1. We could require that anybody who generates rendering URLs manually put
all userprefs on the query string. This allows all prefs merging to happen
in-server, but is inconsistent with prefs on the fragment. That in turn has
client implications, since a container "properly" rendering a gadget would
have to include all userprefs at all times in its URL (pre-merged). True,
that happens with the Java impl today, but it's another somewhat painful
burden. Also, implementation of this mechanism will still need temporarily
to have JS and server-merges, since the JS is shared.
2. Deal with two merges, which is inevitable for a time anyway.
I'll implement the first piece of this (server merge and output, in Java)
and visit the remainder thereafter.
John
2008/9/12 John Hjelmstad <[EMAIL PROTECTED]>
> That's fair. So long as prefs.js doesn't parse the URL but instead accepts
> only the prefs values the server provides, we're golden.
>
>
> 2008/9/12 Kevin Brown <[EMAIL PROTECTED]>
>
>> 2008/9/12 John Hjelmstad <[EMAIL PROTECTED]>
>>
>> > I could be misunderstanding, but it sounds like that's what Ben did --
>> pass
>> > (in the server) the defaults to setMessages_ with some special format,
>> and
>> > have prefs.js read them when no up_<key> values are provided on the URL.
>> > Meanwhile, UP substitution is already handled by the server, with a
>> merge
>> > algorithm that presumably mirrors what Ben's does in JS.
>>
>>
>> Yeah, and I think that's error prone. We should be doing the merging in
>> just
>> one place, or we're going to have bugs.
>>
>>
>> >
>> >
>> > -John
>> >
>> > 2008/9/12 Kevin Brown <[EMAIL PROTECTED]>
>> >
>> > > That's not a complete solution because it skips user pref
>> substitution.
>> > The
>> > > only thing that completely allows skipping filling in the user pref
>> > > defaults
>> > > in the url is to make the user prefs merge happen server side.
>> > >
>> > > 2008/9/11 ben bonfil <[EMAIL PROTECTED]>
>> > >
>> > > > My solution was indeed to pass the default values in the
>> setMessages_
>> > > > array(like iGoogle), and modify prefs.js accordingly.
>> > > >
>> > > > Ben Bonfil
>> > > >
>> > > > On 9/12/08, John Hjelmstad <[EMAIL PROTECTED]> wrote:
>> > > > > Correct me if I'm wrong, but the only server-side code that
>> touches
>> > > > > gadgets.Prefs initializes it with Messages (setMessages_), not
>> Prefs
>> > at
>> > > > all,
>> > > > > all of which are initialized by parsing
>> > > gadgets.util.getUrlParameters().
>> > > > >
>> > > > > I don't have any strong opinion on implementation particulars -
>> > either
>> > > > way
>> > > > > server-side code spits out defaults and some corresponding JS
>> reads
>> > > > them...
>> > > > > assuming we want to support this in the first place.
>> > > > >
>> > > > > -John
>> > > > >
>> > > > > On 9/11/08, Kevin Brown <[EMAIL PROTECTED]> wrote:
>> > > > >>
>> > > > >> That's wholly unnecessary. If we want gadgets.Prefs to include
>> > > defaults,
>> > > > >> we
>> > > > >> just need to make the server side code that outputs gadgets.Prefs
>> > > > properly
>> > > > >> merge in the defaults from the spec. There's no reason to add new
>> > > client
>> > > > >> libraries.
>> > > > >>
>> > > > >>
>> > > > >> On Thu, Sep 11, 2008 at 3:01 PM, John Hjelmstad <
>> [EMAIL PROTECTED]>
>> > > > wrote:
>> > > > >>
>> > > > >> > Given that default values for a given gadget fixed in the spec,
>> I
>> > > > think
>> > > > >> the
>> > > > >> > cleaner solution is to inject some kind of defaultPrefs object
>> > into
>> > > > the
>> > > > >> > gadget when rendering it, somewhat like gadgets.config.init()
>> > does.
>> > > > >> >
>> > > > >> > As implemented today (which admittedly isn't the cleanest impl,
>> > but
>> > > > >> that's
>> > > > >> > an orthogonal problem), that would mean modifying
>> > > GadgetRenderingTask
>> > > > to
>> > > > >> do
>> > > > >> > something like:
>> > > > >> >
>> > > > >> >
>> > > > >>
>> > > >
>> > >
>> >
>> js.append("gadgets.Prefs.setDefaults(").append(jsonObjectRepresentingDefaultsOf(gadget.getSpec()).append(");\n");
>> > > > >> >
>> > > > >> > ...then augmenting features/core/prefs.js correspondingly to
>> > > > initialize
>> > > > >> > from
>> > > > >> > defaults when the up_<key> equivalent isn't available.
>> > > > >> >
>> > > > >> > --John
>> > > > >> >
>> > > > >> > On 9/10/08, Cassie <[EMAIL PROTECTED]> wrote:
>> > > > >> > >
>> > > > >> > > (in case you are interested - don't feel pressured though :)
>> > > > >> > >
>> > > > >> > > samplecontainer.js actually already makes calls to the
>> metadata
>> > > > >> servlet.
>> > > > >> > > see
>> > > > >> > > line 167 - the "requestGadgetMetaData" function. This came
>> from
>> > a
>> > > > >> > different
>> > > > >> > > patch which gave the samplecontainer the ability to show
>> gadget
>> > > > >> > > titles.
>> > > > >> > >
>> > > > >> > > - Caszsie
>> > > > >> > >
>> > > > >> > >
>> > > > >> > >
>> > > > >> > > On Wed, Sep 10, 2008 at 6:48 AM, Tamlyn Rhodes <
>> > [EMAIL PROTECTED]
>> > > >
>> > > > >> > wrote:
>> > > > >> > >
>> > > > >> > > > On Tue, Sep 9, 2008 at 6:55 PM, Cassie <[EMAIL PROTECTED]>
>> > wrote:
>> > > > >> > > > > Hey Tamlyn - do you think you could make this into a
>> patch
>> > and
>> > > > >> attach
>> > > > >> > > it
>> > > > >> > > > to
>> > > > >> > > > > a jira issue for Shindig?
>> > > > >> > > >
>> > > > >> > > > The thing is I've done it from outside of gadgets.js so
>> it's
>> > not
>> > > > >> > > > really patchable. Doing it in gadgets.js would mean making
>> an
>> > > ajax
>> > > > >> > > > call to fetch the metadata in the
>> gadgets.container.addGadget
>> > > > method
>> > > > >> > > > and I'm not really sure how that would work since
>> > > gadgets.ioisn't
>> > > > >> > > > available in the container (unless i've misunderstood).
>> > > > >> > > >
>> > > > >> > > > I'm aware that you're keen to maintain clear separation
>> > between
>> > > > >> > > > Shindig and the JavaScript code that manages the
>> layout/chrome
>> > > but
>> > > > >> for
>> > > > >> > > > most people actually using Shindig I suspect this
>> separation
>> > is
>> > > > >> rather
>> > > > >> > > > more theoretical than practical. I've been developing a
>> drag &
>> > > > drop
>> > > > >> > > > layout/framework using jQuery, PHP and MySQL and I hope to
>> > open
>> > > > >> source
>> > > > >> > > > this once I get it working. This should be some time in
>> > > September.
>> > > > >> > > >
>> > > > >> > > > Tamlyn.
>> > > > >> > > >
>> > > > >> > >
>> > > > >> >
>> > > > >>
>> > > > >
>> > > >
>> > >
>> >
>>
>
>
