[Kst] any advice on datasource access to current plot range?

[George Staikos]

On Sunday 06 November 2005 21:05, Ted Kisner wrote:
> For the healpix datasource, I've been thinking it would be nice to have the
> ability to easily "update matrix using current plot range".  In an ideal
> world, the matrix would be reprojected on every zoom.  This is a bit slow,
> but a good compromise is to have a button on the datasource config widget
> that says "set to current plot range".  This would require the datasource
> to query the current min/max values of its associated Kst2DPlot.

  Datasources don't have 2dplots though.  2dplots have objects, and some 
objects may depend somehow on some datasources.  Do you want something that 
walks the entire objects hierarchy resolving these dependencies?

> I'm fuzzy on the details of how to do this properly...  Do I need to get
> the global list of KstObjects and search through it until I find a
> Kst2DPlot object that contains the proper KstImage and data file?  Is there
> a better way?  What if multiple plots with different ranges are displaying
> the same matrix?

  Exactly....  There are lots of cases.  Including cases where people use 
objects but don't plot them.

> The reason that this kind of action is clunky (and perhaps impossible to
> implement) goes back to our limited types of basecurves.  We are "forcing"
> a healpix map (which is naturally expressed as a *vector* of pixel values)
> into an intermediate data product (KstMatrix).
>
> Eventually, in the future, I still think it might make more sense to have a
> separate basecurve that does a readField from the datasource to obtain a
> full-sphere vector of pixel values, and then projects this information
> directly onto the buffer upon each paint.  This type of direct projection
> from pixel vector --> buffer actually *is* fast enough to do on each zoom,
> and feels/looks *much* better for the user.

  This is definitely possible.  Curves are just siblings of all other data 
objects that happen to use data objects.  It's going to require ongoing 
genericizing of kst2dplot and other objects though.
   
> In the mean time, any suggestions for temporarily hacking in this kind of
> support are welcome...

  I'm not sure at the moment.

-- 
George Staikos
KDE Developer				http://www.kde.org/
Staikos Computing Services Inc.		http://www.staikos.net/