Re: Most efficient way for a big XForms file

Florian Lacoste:
> 
> Victor Engmark wrote:
> >
> >I cannot claim to be very experienced, and I don't know how much
> >functionality and how large instances you are going to handle (link?),
> 
> The instances are not that large, like 30 elements/sub-elements, and on 
> ly 10-15 submitted for the biggest one for the moment. But the UI part 
> for one of the big requests is around 500-600 lines of code. I've only 
> completed 5 request (only a big one) and I've like 10 more to do, 4 or 
> them a lot more complex.
> I'd also like to find a way to batch requests, but i've absolutely no 
> idea how.

You can use a "context" or "batch" instance, which by default only
contains an empty template for the data which will be submitted. You can
then operate on that with an "Add" button, and have a final "Submit"
button which will send the whole instance.

> >but here are some observations from working with the Chiba web module:
> >- Try to separate all read-only data from the rest, and group them into
> >	independent instances as much as possible. This eases the process of
> >	changing data, and marking them read-only makes loading faster 
> >	because
> >	little or no validation is needed.
> >- Try to minimize the number of objects in the read-only data; this is
> >	the only serious bottleneck in my own system now. You might consider
> >	serving it dynamically.
> >- Also, loading as little data as possible for each request type could
> >	help you avoid scaling problems.
> >- (In Chiba, you can gain performance by putting read-only data into a
> >	separate model. This will not change your XPaths (models are
> >	"transparent"), but they will not be revalidated when the other 
> >	models
> >	change.)
> I've not used readonly until now, but as i'm also using chiba, i'll keep 
> that in mind. Thanks.
> 
> I'm doing a similar thing with relevancy data. My instance looks like:
> <data>
> 	<request>
> 		...
> 	</request>
> 	<relevant>
> 		<item1/>
> 		...
> 		<itemN/>
> 	</relevant>
> </data>
> 
> 	When there are optional element/attributes, I use a select with only 
> one item (value:true) bound to /data/relevant/itemX and by selecting it 
> (or not) the user decide if he wants to give a value for this element. 
> The control bound to this element become relevant when this item is 
> selected. And I only submit the /data/request part, so the <relevant> 
> element is not in the request. The element is required, to force the 
> user to unselect if he doen't want to enter a value and avoid having 
> null attributes in the request. It works, but if someone has a better 
> way, i'd like to hear it. ^^
> 
> But my problem is when this optional attribute is inside a <repeat> 
> (lots of problems with repeat or is it just me ? ;p), cause if I bind 
> the select to an element outside the repeat, its value will be the same 
> for each repeated element (then my attributes will all become relevant 
> or not at the same time). For now I created a new relevancy element 
> inside my request and defaulted its value to "true"; if the user 
> unselect it or enter a value, it becomes irrelevant and thus does not 
> appear. But if the user unselect it, there is no way to get it back. you 
> have to delete this repeated element and create a new one.
> 
> In fact I'm looking for a way to have a control that would be relevant 
> in the UI while its value would be irrelevant in the instance. But one 
> can't use differents binds for the instance data and the control :/
> 
> I know I could delete all unwanted elements in the submited data using 
> XSLT, but it would be great if I could have a well formed request using 
> only XForms.

I'm not completely sure I understand how this works, and what you wish
to accomplish. Do you have a link to the XForms?

PS: If you want to look at an example using relevant, repeats, and
context instances, here's a shameless plug:
http://isscvs.cern.ch/cgi-bin/viewcvs-all.cgi/?cvsroot=moi /
CVSROOT :ext:isscvs.cern.ch:/local/reps/moi

-- 
Victor Engmark
"Quid quid latine dictum sit, altum viditar" - "What is said in latin,
sounds profound"