[Kst] Re: Class changes

[C. Barth Netterfield]

On Thursday 29 July 2004 07:02 pm, you wrote:
> My concern here is that we are sacrificing functionality
> for a "clean" class design. However, the functional specification
> should always preceed any code design and implementation.
> Some desired functionality results in complex code (but this
> is not the same as "dirty" code). I think the fits was a case
> in point. At the cost of an additional class we tied together
> the curves and labels. I think this is something very desirable
> to do and its only a matter of time before we get a complaint
> about not doing it.

I agree that we have reduced functionality in the fits, but I hope it is 
temporary.  

For example, the problem where, if we change a fit,  it doesn't change the 
lable is a problem which is generic to any label associated with a data 
object.  It is the same problem with auto-generated X and Y labels in plots, 
for example.  I would like to think about a generic solution to this.  

What I am imagining is for KstDataObject is to get a virtual QString 
automaticLabel(), which can be implemented by different child classes.  For 
fits, we could "if (plugin->_isFit) ..." in KstPlugin::automaticLabel().   

Then we can refer to a data-object's automaticLabel in any label; changing the 
data object will change the label.  (Eg, change a vector in a curve, and the 
X and YLabels in the plot, if they refer to the autmaticLabel, automatically 
change)  This could be awsome for equations as well - well formated equation 
names could be exported for label use.....

  A similar problem is wrst exported scalars - what happens when you change 
the source of an automatic scalar, or change the name of a vector which is 
generating automatic scalars.  This is a general problem which arrises from 
our refering to a label by its tagname, and it needs a general solution.

Thoughts?