Re: plot and image demos

[Ben Abbott]

On May 27, 2009, at 12:48 PM, John W. Eaton wrote:

> Using the current development version I just did
>
>  more off
>  rundemos plot
>  rundemos image
>
> and stepped through all the demos and everything looks good except
> for the following:
>
>  * the demo for comet causes the plot window to grow in the vertical
>    dimension (hmm, this happened twice, but then it seems to not be
>    100% repeatable).
>
>  * the first surfl demo, the contrast demo, and the first imshow demo
>    all showed the warning
>
>      warning: range error for conversion to character value
>
>    but that only happened when running all the demos with rundemos
>    plot followed by rundemos image, not when running them
>    individually.
>
>  * sombrero produced warnings from gnuplot
>
>      multiplot> et origin 0, 0
>                 ^
>                 line 1092: invalid command
>
>      Press <enter> to continue:
>      multiplot> u
>                  nset view;
>                 ^
>                 line 1092: invalid command
>
>    but this only happened once (?!).  Other runs of this demo seemed
>    to work OK.  So I don't know what might have caused problems
>    sending data to gnuplot.
>
> Does anyone else see these problems?  Any clues about what could be
> causing the warnings?  Is something sometimes not reset properly when
> a new figure is created?
>
> jwe

I'd thought the growing window had been fixed. I had encountered this  
when adding support for the figure position. Unfortunately, each time  
the position and size of the figure was specified via gnuplot's "set  
term  ..." command the window had some probability of growing by one  
character height. The solution was to only specify the window position  
and size when the figure was created (i.e. only for the original "set  
term ...")

If I recall correctly you are running gnuplot 4.2.5. Perhaps this  
version of gnuplot behaves a bit different than 4.2.4/4.2.3?

Regarding "warning: range error for conversion to character value" is  
there a way to turn this specific warning into an error? In situations  
like this, such a feature would be very helpful.

The errors you encountered with sombrero occur for me on occassion.  
They've generally occurred for me when the prior plot failed. A "close  
all" has always cleared it up for me. In any event, it still looks  
like a bug. If someone can find a way to repeat the problem (an my  
observation is correct), it should be possible to use unwind_protect  
to avoid the problem, or to do a better job initializing the plot  
stream each time it is drawn.

I have a busy morning ahead (I'm 12 hrs off your time zone), but hope to

Ben