PLplot

On Thu, Aug 05, 2004 at 12:07:42PM +0200, Rafael Laboissiere wrote:
> 
> This error message comes from the following lines in plgridd.c:
> 
> #ifdef HAVE_QHULL
>     grid_dtli(x, y, z, npts, xg, nptsx, yg, nptsy, zg);
> #else
>     plabort("plgriddata(): you must have Qhull to use GRID_DTLI.");
> #endif
> 
> Autoconf does detect the presence of the Qhull library and set the
> HAVE_QHULL conditional appropriately.  However, I think that the call to
> plabort above is too strong and should be downgraded to plwarn (as in other
> places in plgridd.c).
> 
> What do the other developers think?

I agree plabort is a bit strong. plwarn would probably do. Perhaps we
could expand on the message a bit to make it a bit clearer to new users
and avoid panic? Should we also fall back to some other method instead
in this case?

Andrew

* Andrew Ross <andrewross@...> [2004-08-05 12:25]:

> I agree plabort is a bit strong. plwarn would probably do. Perhaps we
> could expand on the message a bit to make it a bit clearer to new users
> and avoid panic? Should we also fall back to some other method instead
> in this case?

Unless Joao is around and hearing to this thread, could you take care of it,
please?  

-- 
Rafael

On Thu, Aug 05, 2004 at 04:22:59PM +0200, Rafael Laboissiere wrote:
> * Andrew Ross <andrewross@...> [2004-08-05 12:25]:
> 
> > I agree plabort is a bit strong. plwarn would probably do. Perhaps we
> > could expand on the message a bit to make it a bit clearer to new users
> > and avoid panic? Should we also fall back to some other method instead
> > in this case?
> 
> Unless Joao is around and hearing to this thread, could you take care of it,
> please?  

I've commited a change to CVS to do this. It will fall back to using the
Nearest Neighbors "Around" Inverse Distance Weighted method if either
CSA or QHULL is not available. 

Andrew

* Andrew Ross <andrewross@...> [2004-08-05 16:27]:

> On Thu, Aug 05, 2004 at 04:22:59PM +0200, Rafael Laboissiere wrote:
>
> > Unless Joao is around and hearing to this thread, could you take care of it,
> > please?  
> 
> I've commited a change to CVS to do this. It will fall back to using the
> Nearest Neighbors "Around" Inverse Distance Weighted method if either
> CSA or QHULL is not available. 

Great.  Thanks.

-- 
Rafael