[Kst] Re: filters and stuff...

[Barth Netterfield]

A) I think filters should be plugins.
B) Filtersets should be (inherit from?) plugins.
C) There needs to be an 'apply filter' widget/dialog which can apply any one 
filter (including 1 filter set).  This will be used in datawizard, and in 
various curve related context menus.
D) there needs to be a widget/dialog to define filter sets which are chained 
filters.  Similar to the current filter set dialog.

If we ignore the concept of filter sets, then filters are just a subset of all 
plugins which can be used with a reduced/tuned user interface, but which 
appear in the data manager, etc exactly as any other plugin.  The concept is 
parallel to fits.

I imagine that filter widget (C) will look a lot like the current plugin 
dialog, in that the input parameters (except the vector you are filtering 
which will be fixed) are all defined here, etc.

QUESTION: should the parameters for filter sets be set in the filter set 
creation dialog, or when they are used.

ANSWER: Perhaps if one is chaining filters, one already has in mind the values 
for all of the parameters.  So these parameters are fixed for a given filter 
set.  So a filter set looks like a parameterless filter (1 in, 1 out) in all 
plugin UI's.

QUESTION: given the above, if one edits the parameters of an existing/used 
filter set, do the vectors which are using it change to match the new 
parameters?  (Sort of like reaching down into the plugin an recoding it mid 
stream....)

ANSWER: yes.

On June 1, 2004 05:29 pm, you wrote:
> Hi Barth,
>
> Rick was going to start looking at filters.
> Is the plan still to change their interface
> back to that used by plugins? What about the
> UI?
>
> Andrew