[R] Summary: Future directions for R on Windows

Prof Brian D Ripley ripley at stats.ox.ac.uk
Thu Apr 22 23:03:41 CEST 1999

On Thursday 15 March I wrote:

> Now rw0640 is released [...] we are now thinking of possible
> enhancements, and would like your comments on the following ideas.

We have had 25 replies, and congratulations to all but one person for
following the instructions on where to send them. They were very
constructive and raised a lot of points not taken up here but which we
will discuss amongst ourselves and R-core.

Here is a summary of opinion, and our current views.

   o   Use a standard MDI interface like MS Office's, say. Would you 
       prefer an MDI interface to the current one, and if so is the
       `standard' `look and feel' important?

  Pretty evenly divided between `be as standard as possible' and
  `happy to be different'.  For those who didn't know, MDI is
  `multiple document interface' and means lots of sub-windows in one
  large frame.	

  Guido has been busy, and a MDI version is running as a
  prototype. The next rwxxxx will allow the current (SDI) or MDI
  styles via a configuration file / command-line flag.  We are looking
  at building a different front-end with the `standard' look, as an

   o   Replace the help system by standard Windows help files. Would
       you like this instead of text help, html help or both?

  Some of you want this, some of you hate it. Someone who hates them
  wants a help file system that allows full text searches, apparently
  unaware that Windows help files are the one choice that does allow them.

  The hooks are now in, and I will work slowly on Rd -> hlp
  conversion scripts. Then those who want it can use it.

   o   rw0640 implements offline printing via latex.  Do you anticipate
       using this?  (It needs lots of extra files.)  Would printing from
       a Windows help file be an adequate replacement?

  A few people use it / or will use it. One thinks people who use
  Windows do not use LaTeX, but two high-quality freeware projects
  (fptex, MiKTeX) show otherwise. Some of you would prefer PS or PDF.

  Keep distributing the latex files.  The future of help formats for R
  as a whole is under discussion.

   o   Build an R dll that can be linked into other applications to drive
       effectively an R command-line by passing input strings. Possibly
       versions without and without on-screen graphics.

  Those who saw a use were very enthusiastic.

   o   Implement inter-process communications, either
       -  Windows-style (DDE and ActiveX automation) or
       -  CORBA-style  (real standards)
       This might allow R to be called from a spreadsheet or Visual Basic 
       front-end, for example.

  Those who saw a use were very enthusiastic, and a few wanted CORBA,
  but most want to be able to communicate with the likes of Excel.

  This is linked to the previous point and to the overall design of R.
  We will investigate what might be possible, but do not expect
  progress real soon.

   o   Would you prefer an installer as well as / instead of the 
       current set of smallish zip files?

  Many people said they don't want one but knew people who would like
  one.  Many people don't trust them! Many people want < 1 floppy chunks.

  To build an installer like that used by GSView, that unpacks the zip
  files for you to a place you choose and allows you to select the
  parts you want. I think this might be helpful for installing
  packages too.  The pieces will remain in the 1Mb region.

   o   How important is it that R can be built from source with free
       tools?  (We are not contemplating requiring any new tools to
       install packages, but a Windows-standard interface probably 
       requires a commercial compiler with MFC.)

  Not much interest in building R from source from the respondents but
  some strong views from within R-core.  A few people said they
  thought this should be possible but would not actually do it.

  We will experiment.  There are some concerns that the port should
  not depend on one or two people, but I feel requiring a serious
  developer of R itself to buy a particular compiler is not really an
  issue. (I have done this many times for other systems.  My time is
  much more expensive than compilers.)

  We would also welcome other suggestions.  Some of this appears easy,
  some rather hard, so there are no guarantees that anything will be done.

  There were lots of points. The most common requests were for

  o  Some sort of programmable menu system to add commands to menus.

  o  A spreadsheet-like data editor in R.

  o  Customizable graphics (I take this to mean editing graphics after
     they are drawn).

  o  Connections to external databases, ability to read/write Excel and
     dBase files, ....  (To some extent having ActiveX would avoid any
     need for this.)

  o  Ability to use VC++ to compile for R. [This one is already there:
     see the V&R2 `R Complements' for a worked example, and I have
     added that info to readme.packages.]

  o  Some of you want more optimized code. This might be possible via
     compiler options although we think the compiler we use is quite
     good and our version compares well with the Linux version on the
     same hardware. There are performance tuning issues for R, but not
     just this version and not yet.

and lots of not-specific-to-Windows things like gam(), flexible memory
management, three-dimensional and Trellis-like graphics.  Perhaps
unsuprisingly the first four points are common in other Windows
statistical systems, including the `R-like' one.

Some of those things are hard for us to see how to do, and we are not
enamoured of low-level Windows programming. We would appreciate help

  -- Examples of database connections / spreadsheet imports from open
     source packages we could use for inspiration, and any examples of
     open source/other code that already does that.

  -- Windows programming. Someone who is quite familiar with Windows
     programming could do many of these things rather quickly without
     knowing anything about R internals.

As before, private comments only please, and remember that you will only
get what the volunteers (guido + I + ...) want to work on.


Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch

More information about the R-help mailing list