[Kss-devel] Marshalling of form data

Godefroid Chapelle gotcha at bubblenet.be
Fri Feb 23 13:41:47 CET 2007


Balazs Ree wrote:
> 2007. 02. 22, 16.10 Hedley Roos wrote:
>> I am working on a project using PloneAzax and I make use of marshalling 
>> (:list, :record, :records) in my page templates. Azax cannot handle 
>> this, which means that in my Python scripts I do not have access to the 
>> structures. There are other ugly solutions like assembling special tag 
>> attributes and decoding them, but I wanted a clean solution.
>>
>> I heavily modified forms.js and kukit.js and made a single char change 
>> to resourcedata.js. See attached patches and examples to understand the 
>> discussion below.
>>
>> In my KSS file I had to make use of a naming convention. For exmaple 
>> consider the line
>>
>> kss_MarshallTest-value_record|record: currentFormVar('arecord:record')
> 
> I think we don't need the | convention in the variable name for that.
> Because that is already known from the type of the data to be sent over
> (= the output of currentFormVar). The marshaller can put the record tags
> automatically.
> 
> This is actually already used with :list and maybe with :record too. So
> what we actually need to do is
> 
> - to make the marshaller marshall the :records types too
> 
> - to make the formvar logic smarter so that it can select the composite
> variables that belong to one varname only, from the form.
> 
> I am not saying the final word on this, but changing the syntax in the
> suggested way is something we would like to avoid. So if we can do
> without it...
> 

At the end of the sprint, Balazs and I discussed syntax for a few 
features that we did not know how to implement.

We suddenly recalled that CSS has stuff like the following :

border: solid 1px black;

and that those space-separated values were a solution to the features we 
wanted to implement.

I think it is the case this time as well.

I propose that we explore the thing shown below.

kss_MarshallTest-value_record: currentFormVar('arecord:record') record


This would be a non invasive zopism :

frameworks that do not pass type like Zope does would not state type and 
they would even not know...

Your opinion ?

PS: I do not have a lot of time this week to discuss more. Please hassle 
me in the coming days


-- 
Godefroid Chapelle (aka __gotcha)- BubbleNet  http://bubblenet.be



More information about the Kss-devel mailing list