RKWard 0.5.5 - Enhanced robustness and many small improvements

Tuesday, 19 April 2011

On the surface, RKWard 0.5.5, released today (download) may appear to bring only few changes. Internally, however, the GUI and the R engine have been split into two separate processes. This brings better robustness, and better compatibility of RKWard esp. with R packages based on rJava and RGtk2. In addition, many bugs and quirks have been fixed in various areas. And, in case you experience a crash despite of these improvements, RKWard 0.5.5 will attempt to create an emergency save file of your data in the last moment.

Please give this new version of RKWard a try, and don't forget: We are always eager to hear your feedback!

The changes in detail:

New features and improvements

Re-organize the default toolbar actions of the main window
Detached windows are now restored in detached state when restoring the workplace layout
Workplace layout is now saved in a separate file, instead of in a hidden object inside the .RData file
Added R functions rk.save.workplace() and rk.restore.workplace() to save / restore a set of document windows
RKWard now tries to detect, when a workspace has been moved to a new directory, and adjust the paths restored script editor windows, accordingly
All pages in the package installation dialog now support sorting and keyboard search
Allow entering factor labels instead of only numbers, when editing factors in the data editor
Display logical values as "TRUE" and "FALSE" in data editor, and accept "T"/"TRUE"/"F"/"FALSE" in addition to "0"/"1" while editing
Add support for R functions loadhistory(), savehistory(), history(), and timestamp ()
Moved automated testing framework moved to a dedicated R package "rkwardtests", and added documentation
Support pasting to the middle of the current command line in the R Console
Better handling of script-editor commands while the R Console is busy
GUI frontend and R backend now run in separate processes
Re-write of large portions of the R embedding code
Attempt to save workspace on crashes
Removed the remainder of the PHP scripting backend

Bugfixes

Fixed: Crash when several environments on the search path have the same name
Fixed: Potential crashes when changing length of a data.frame that is currently opened for editing from R code
Fixed: Converting from factor to string in the data editor set values to numeric, internally
Fixed: On Windows, a wrong locale for LC_NUMERIC would be applied, resulting in malfunction of pdf() and postscript()
Fixed: Did not use system configured background color in data editor
Fixed: New columns would be always be added to the right, instead of to the left, in the data editor
Fixed: Would crash when trying to edit 0 row data.frame
Fixed: Output generated by external processes (system()) was not shown on the console
Fixed: Converting a variable to factor in the editor would drop existing levels, silently, and lead to NAs
Fixed: Cursor would keep jumping to the end, when typing a filename in the filename selection field in plugins
Fixed: Submit button would not become enabled in "Basic Statistics" plugin
Fixed: Portions of multi-line commands entered on the R Console would be run twice under some circumstances
Fixed: Lockups or crashes when using packages RGtk2 or rJava on the X11 platform
Fixed: R task callback handlers were not called
Do not load .RData file from current directory by default (configurable)
Correct interleaving of multi-line commands and the corresponding output in the R Console
Fixed: Did not respect system settings for tooltip background color in some styles
Fixed: Installation would continue, seemingly successfully, even if R support libraries could not be installed
Fixed: Function argument hints would sometimes persist after closing a script window
Fixed: Would fail to analyse structure of ReferenceClass-objects
Fixed: "Vector" mode in "Paste special" action did not work correctly
Also try to relay SIGABRT and SIGILL to the proper signal handlers
Fixed: Would crash on tcl/tk events in some cases