Controlled R Sessions for Each Widget

[GoogleCodeExporter]

Code enhancement.

One code enhancement that may be explored will be to allow each widget to 
control a separate R environment or subset of variables.  These could be 
individually saved and reloaded as needed.  For example once a dataset is 
subset the orriginal dataset can be saved to the tempdir and the subset can be 
worked on allowing a relative expansion of available memory.

This would involve some changes to the signals and RSession.  Session calls may 
require some extra variable names to ensure that those variables are active in 
the session.  This could be checked quickly in R.

The use of the QTimer class could be implemented to "close" or "dump" R 
variables to disk to free memory automatically.

A branch will be made to explore this issue.

Original issue reported on code.google.com by kylecovi...@gmail.com on 30 Jan 2011 at 8:22

[GoogleCodeExporter]

Issue 469 has been merged into this issue.

Original comment by kylecovi...@gmail.com on 31 Jan 2011 at 1:02

[GoogleCodeExporter]

I have tested the RObjects functionality and it seems to be working well.  I 
think this can be rolled out fairly soon.

One key aspect that this relies on is that the signal classes have access to 
the widgetID of the generating widget. 

To do this I would propose that we have all signal classes require a widget 
variable as the first argument.  So all widgets and signal classes should be 
changed to include asdf(self, data = data) instead of asdf(data = data).  This 
will allow us to implement the R variable as an update instead of a new version.

Do we have time to do this????

Original comment by kylecovi...@gmail.com on 16 Feb 2011 at 5:41

[GoogleCodeExporter]

Committed changes for each signal class needing a widget instance passed as 
first argument.  This has been fixed in the widgetSession class and in the 
widgets in base, stats, and plotting.

Should test and report any errors.

Original comment by kylecovi...@gmail.com on 21 Feb 2011 at 3:00