[R] A long digression on packages

[Sander Oom]

Maybe some of this confusion about search opportunities and pros/cons 
could be avoided if the search page on CRAN 
(http://cran.r-project.org/search.html) would be extended to cover all 
main search tools!

Quickly scanning the discussion, I found these:
1- simply Google: some tips and tricks have been mentioned and would be 
usefully for most users;
2- R site search (external to CRAN) 
http://finzi.psych.upenn.edu/search.html;
3- from R prompt: help.search();
4- browser supported search through local help files: 
R/doc/html/search/SearchEngine.html.

ad1. Google is normally my first source using broad keywords for a 
method or problem.
ad2. just discovered this today!
ad3. help.search() provides a simple overview, printing the command with 
the providing package:
 > help.search("rose")
Help files with alias or concept or title matching ‘rose’ using
regular expression matching:
hirose(boot)            Failure Time of PET Film
rose.diag(CircStats)    Rose Diagram
rose.diag(circular)     Rose Diagram
windrose(circular)      Windrose Generator
rosavent(climatol)      Wind-rose plot
Kinship82(clue)         Rosenberg-Kim Kinship Terms Partition Data
HolidayDatabase(fCalendar)
                         Holiday Calendars and Utilities
conc(ineq)              Concentration Measures
Type 'help(FOO, package = PKG)' to inspect entry 'FOO(PKG) TITLE'.

ad4. when installing all package locally, the results produced by the 
browser supported search can be overwhelming. Even searching within the 
results often does not help. If commands were printed along with the 
providing package would be a good improvement. Then the apparent random 
order of commands listed might also reveal some order.

Hope this is a useful addition to the debate,

Sander.

Chris Evans wrote:
> On 5 Jun 2005 at 18:44, Jari Oksanen wrote:
> 
>>There are diverse opinions about netiquette. One of the most basic, in
>>my opinion, is this: if someone posts starts a discussion in a certain
>>forum, you shall not divert it to another forum where it may be hidden
>>by most readers, perhaps even by the originator of the thread.
> 
> With the greatest of respect for Duncan and the R-devel list, I think 
> Jari has a point here.  This is one of the most important issues I've 
> seen raised on this list (R-help) in recent months and I think it may 
> be a structural problem for the development of R, in common with that 
> of much FLOSS s'ware, that there's a separation of users and authors 
> that needs thought.  There are no perfect answers but too big a 
> separation and projects go "techno" and it's hard for those of us who 
> can't code C and who are mere "users" to help those outstanding 
> people on whom we depend hear what we need: sometimes they are so 
> clever, so specialised in their knowledge, or simply in the realm of 
> genius not the ordinary, that they can't see our problem.  I have 
> slowly come to respect that a pretty brusque style from our 
> authorities is the only way to prevent this list being a madhouse but 
> I think that Jim's point may fall into that class that is worth some 
> duplicate bandwidth here.
> 
> I know I've found the problem Jim highlights very confusing and 
> unhelpful at times.  Views, which I didn't know, seem helpful but not 
> a real solution to the key problem: they may tangentially help by 
> ensuring that if your needs fit into a view, it becomes more likely 
> that you'll install the packages you need and a local search may tell 
> you what you need.  I've taken the inefficient route which suits me 
> of installing just about every package to make it less likely I'll 
> miss something of use to me. That means my search for "kappa" and 
> "Cohen" (with ignore.case=FALSE) turns up at least three 
> implementations of aspects of Cohen's kappa.
> 
> It may already exist, but a web interface that did a help.search over 
> all the packages in the current release version would be great.  (If 
> it does exist, sorry, but I'm no dunce and use R nearly every day and 
> try to read much of r-help every day and don't know it, which may say 
> something!)
> 
> I think there may be a need for some R improvement and automated 
> updating of what I think is Frank Harrell's function finder:
> 	http://biostat.mc.vanderbilt.edu/s/finder/finder.html
> Though I'm not absolutely sure how fitting your works into something 
> like that could be imposed on developers!
> 
> Another thing that might help would be for a system by which ordinary 
> users would volunteer to pair up with developers for packages and try 
> to suggest adaptations of the help and such like that might make the 
> packages more user friendly.  I wouldn't want to do that for the 
> whole of a huge and vital package like MASS or Hmisc (or base or 
> stats!) but I'm up for pairing with a developer on a smaller package 
> if anyone thinks that would be helpful.
> 
> Thoughts for what they're worth.  Thanks a million to all developers 
> ... asbestos suit on!
> 
> Chris


-- 
--------------------------------------------
Dr Sander P. Oom
Animal, Plant and Environmental Sciences,
University of the Witwatersrand
Private Bag 3, Wits 2050, South Africa
Tel (work)      +27 (0)11 717 64 04
Tel (home)      +27 (0)18 297 44 51
Fax             +27 (0)18 299 24 64
Email   sander at oomvanlieshout.net
Web     www.oomvanlieshout.net/sander