[rkward-devel] some separate queries related to the plot history feature

[Prasenjit Kapat]

Hi,

On Sun, Sep 12, 2010 at 1:36 PM, Thomas Friedrichsmeier
<thomas.friedrichsmeier at ruhr-uni-bochum.de> wrote:
> Hi,
>
> On Sunday 12 September 2010, Prasenjit Kapat wrote:
>> The "Append this plot" is supposed to be a safe guard action: append
>> _any_ displayed plot no matter what its state / type is, using
>> recordPlot (). For example, any call that does not go through,
>> plot.new () / persp () / print.trellis () should still be appended to
>> the history by this action. In such cases it may not possible, at
>> least not with the current implementation, to extract any information
>> at all.
>
> And so we cannot assume that the last known plot-call was what generated this
> plot. Hence it gets the "fallback" label, instead of "plot (2, 2)", correct?
> Hm, probably not fixable, then (and not important at all).

Right.
This situation also arises when re-enabling the plot history. As
currently implemented, any displayed plot will be appended to the
history. Here again, we can not identify the plot call at all because
recording was disabled until then!

> Another thing: I just added a test to rkward_application_tests.R. Basically

Thanks.

> this opens two devices and calls some of the history actions. Then it checks
> whether the current plots are the expected ones. Somehow, when I try to mix in
> trellis plots (e.g. for plots 2 and 4), the test starts to fail. I have not
> really investigated, exactly what goes wrong, though. Perhaps you can have a
> look, too?

I'll look into that. One has to make sure that the lattice calls are
made _only_ when the lattice package is available _and_ loaded.

Regards,
-- 
Prasenjit