[rkward-devel] plot history: 3rd attempt: request for feedback

[Stefan Rödiger]

Am Freitag 10 September 2010, 21:27:14 schrieb Prasenjit Kapat:
> Hi,
> 
> I've just committed another attempt at the plot history feature(s).
> Please test it. Some notes follow:
> 
> 1. Implemented action:
> 1a. first / prev / next / last
> 1b. Remove plot
> 1c. Forcefully append plot, irrespective of its status and type
> 1d. Clear history
> 1e. Drop down menu
> Insert and Overwrite actions haven't been implemented and I do not
> intend them for this release.
> 
> 2. par (mfrow/mfcol) and split.screen have been accounted for, almost!
> 
> 3. Although there is limit on the size of individual plots, one
> crossing that limit the user is now presented with an option to save
> or ignore this over-sized plot... This way, such plot are not "lost."
> 
> 4. As usual, play around with the icons / menus and make sure the
> behavior is _NOT_ harmful and "less confusing." Stuff to try out:
> 
> . open multiple new (empty) devices
> . add plots using regular plot calls
> . update plots using title (), grid () etc...
> . duplicate device (with saved and unsaved plots)
> . remove plots
> . browse using the arrow buttons
> . browse using the drop down menu and make sure that the order /
> indices are not screwed up
> . close device (with saved / unsaved / _no_ plot)
> . perform actions on inactive device and make sure that the state of
> the device does not change
> . perform actions on _one_ device and make sure the display on other
> devices are not changed
> . exceed the hist limit
> . reduce the hist limit from Settings > RKWard > Output to something
> below the current hist length
> . plot large sized plots (for ease of use, you may reduce the plot
> size in settings)
> . call par (mfrow = ...) or split.screen (...). It is _not perfect_
> but see if is "satisfactory"..
> . if you use lattice then mix lattice and standard calls...
> . for lattice: try out update (trellis.last.object (), ...) calls
> . try preview devices.. they should not be tracked at all
> ....
> 
> 5. And MOST importantly, find plotting functions which does not get
> recorded (technically, ones that do not call plot.new or persp or
> print.trellis). In such cases (even otherwise) try using the "Append
> plot" action.
> 
> 6. By default all debug output is turned off. If you are so inclined
> to see (esp on error) call:
> 
>     rk.record.plot$.set.rk.rp.debug (TRUE) # FALSE to turn off
> 
> This will create a whole lot of messages.
> 
> 7. The label strings for the drop down menu are by default restricted
> to 50 characters. If you need to change it, call:
> 
>     rk.record.plot$.set.call.lab.len (60) # or 20 ...
> 
> See how the it fares on different screen sizes.
> 
> 8. Summary of all the saved plots (not of the devices) can be found by:
> 
>     rk.record.plot$getSavedPlotsSummary () # this is data.frame
> 
> 9. Summary of the managed devices can be found by:
> 
>    rk.record.plot$getDevSummary () # but this needs debug to be TRUE
> 
> Of course, there are a lot debug commands, although no output is
> produced, which will be removed once the performance is acceptable.
> 
> Regards,

Great I will uppdate, test and report.

Regaqrds
Stefan