[Rd] Accessing data files w/ --use-zip-data

Prof Brian Ripley ripley at stats.ox.ac.uk
Fri Oct 24 08:36:51 MEST 2003


On Fri, 24 Oct 2003 w.huber at dkfz-heidelberg.de wrote:

> > The issue that Duncan suggested I raise is whether or not it should be
> > considered accepted behaviour for a package author to be accessing files
> > in <pkg>\data (at this time I don't know the reasoning behind the specific
> > example here, I just know that that's what they've done) or if this should
> > be considered a Bad Thing.
> 
> Examples are the makecdfenv und matchprobes packages, part of which I
> wrote. They deal with data formats that are external to R, namely vendor
> (Affymetrix) specific file formats, and provide tools for importing these
> data into R. For the examples / vignettes, I need to provide example data
> files. Up to now I am putting them into the data directory, and access
> them with
> 
> specializedImportFunction(file.path(.path.package("matchprobes"),
>                           "data", "HG-U95Av2_probe_tab.gz"), ...)
> 
> Is this considered offensive to the R package structure? Whereelse to put
> data that comes in specialized (not .rda, .txt, .csv) formats?

It is broken, as Windows users have found.  Please use a different
installation directory, as in my earlier reply, or make
specializedImportFunction() handle zipped data directories.

I have been proposing, and may implement as soon as 1.9.0, to process all 
the data directory at install time to .rda files in a database (details 
depend on Luke's final choice of implementation)  which is load-on-demand 
and perhaps unload if not currently used.  So it not a good idea to use 
the data directory for files other than those to be accessed by data().

-- 
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 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595



More information about the R-devel mailing list