Re: [evolution-patches] Evolution calendar get_object_query optimizations

[Rodrigo Moya <rodrigo novell com>]

On Tue, 2004-10-26 at 13:36 +0530, Harish Krishnaswamy wrote:
> On Tue, 2004-10-26 at 00:47 +0200, Rodrigo Moya wrote:
> 
> > this is the work I am doing on the recurrences-work branch, so please,
> > give up on this, since I'll have it finished this week, ready to be
> > merged to HEAD.
> sorry.. did not mean step on your work. This patch merely replaces some
> wrapper calls with direct function calls and adds some early checks - so
> that redundant queries to back-end are not made. No attempt has been to
> modify the functionality, at least by purpose :)
> 
you didn't step on my work, you just changed some stuff I had already
changed in the branch.

> Regressed the patch only for current functionalities ('Delete All
> instances', 'Delete this instance' options for recurrences). They seem
> to work fine. Not sure if its compatible with something new you are
> planning to add.
> 
yes, it is, as I said, it's part of the changes I've done in the
recurrences-work branch.

> 
> > Also, unfortunately, this is wrong:
> > we need to use e_cal_generate_instances_for_object to get back any
> > detached recurrences, which e_cal_recur... does not return.
> > 
> fail to understand why this is wrong. 
> e_cal_recur.. in the course of its execution generates the instances and
> updates the view accordingly, ensuring the existing functionality is
> preserved. Why do we need to get the recurrences back at the
> e_cal_view_objects_added_cb function ? 
> 
if there are detached instances for the object you're generating
instances for, e_cal_recur_* won't detect them, since it's got no access
to the ECal. Thus, it will never return any detached instance there is.

e_cal_generate... on the other hand, queries the backend for all objects
with the same UID as the object you're generating instances for, and if
there are detached instances, replaces the appropriate ones from the
list of instances returned by e_cal_recur_*.

I hope I've explained it correctly :)
-- 
Rodrigo Moya <rodrigo novell com>