[rkward-devel] Subset plugin; Release plans

[Thomas Friedrichsmeier]

Hi Meik!

I'll try to write a more formal call for testing, and hopefully prepare some 
testing packages on the weekend. But this is a reminder that a new release of 
R is coming up (R 3.0.0 is scheduled for April, 3), and we will have to 
release a new version of RKWard (0.6.1) to support it.

Thus, everbody, please do some testing with the development version of RKWard, 
if possible using alpha releases of R 3.0.0. Also, translators may want to 
start working on the new strings 
(https://translations.launchpad.net/rkward/trunk). While strings are not 
formally frozen, yet, I don't expect any major changes before the release.

This also means we have to finish any features that we want to see included in 
RKWard 0.6.1, as soon as possible. One such feature is the filtering / 
subsetting plugin. Meik has already started working on this. The easiest way 
to test the latest version may be to run this script in RKWard:
  http://svn.code.sf.net/p/rkward/code/branches/external_plugins/rk.subset/inst/rkward/rkwarddev_plugin_script.R
(load the rkwarddev package, first).

Meik, do you have some time to work on the plugin, over the next week or so? 
Here's some feedback (finally):

- I think the layout looks mildly intimidating, already, and will grow even 
larger as more features get added. I would suggest splitting this into two 
tabs (or even two plugins?). Filtering columns could go to the second tab, 
IMO, as filtering cases is much more common in my experience.

- Also related to the layout, I think, if the "Filter rows by variable"-frame 
would have not further elements underneath it (only some blank space), there 
would be more flexibility for showing/hiding fields, without causing too much 
jumping in the layout.

- The filtering by columns could support both _keeping_ columns (as is, 
currently) and _removing_ columns. E.g. a radio box with options "Include all 
columns", "Include only these columns:", "Remove these columns:". Just an 
idea, though, certainly not a priority.

- Filtering by cases could make use of the <optionset> to allow for more 
complex filtering. Rows could be combined using AND or OR (left to right, 
without parentheses).

- The "Filter rows by variable"-checkbox could be removed. If using an 
optionset, this would be controlled by wether any rows are in the optionset. 
But even when sticking with the current simple solution, this could simply be 
controlled by whether or not the user has selected any object in the "Filter 
by" slot.

- For numbers, it might be more convenient to be able to specify a range, i.e. 
two <input> fields, one for min, one for max, and additional checkboxes 
"include min (>=)", and "include max (<=)". Not sure, whether this should 
replace the current <, <=, >, >=, or whether it should be an additional 
option.

- For character / factor, perhaps values could be specified in a (one column) 
<matrix>? This would avoid quoting, and seperator character issues.

- There should be an additional option "filter rows by custom expression" to 
allow specification of anything that is not covered in the GUI (importantly 
relations between more than one variable, and transformations, like var1 >= 
2*var2, etc.).

- I have also done a few small / internal changes.

If you're busy, I could try working on some of this. Otherwise, I'd like to 
focus on trying to improve startup robustness on Windows a bit (replacing the 
flakey rkward.bat with a smarter binary) in the time remaining before the 
release.

Regards
Thomas
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://mail.kde.org/pipermail/rkward-devel/attachments/20130312/f897dfa6/attachment.sig>