[Rd] CRAN Feature Request: install.data (task.view="Cluster")

[Duncan Murdoch]

On 28/10/2018 5:08 PM, Abs Spurdle wrote:
> Hi
> 
> I wish that it was possible to install datasets from R packages
> without installing the rest of the packages.
> Mainly for testing purposes or writing examples.
> 
> Assuming that there's a top level data directory (which there isn't):
>> install.data (package=c ("s20x", "R330") )
> 
> Which would install all datasets from the s20x and R330 packages,
> without installing the complete packages, with some form of
> documentation and presumably with license info.
> 
> Or:
>> install.data (task.view="Cluster")
> 
> Which would install all the datasets from the Cluster task view, again
> without installing the complete packages, with some form of
> documentation and presumably with license info.
> 
> Documentation could be the Rd files or a reduced form of the pdf/html
> files, covering the datasets only.
> 
> Also, it would be good to have search tools too.
>> search.data (task.view="Cluster", keyword="something", all.Rd.fields=TRUE, condition="n.numeric.vars >= 3")
> 
> Which would return a list of datasets from the Cluster task view with
> keyword "something" (from anywhere in the Rd file) and 3 or more
> numeric variables, including info on the (dataset) name, package
> (name), license, header, dimensions and possibly description.
> 
> Obviously, these functions would not be part of CRAN itself, however,
> CRAN would need to store packages in a different way for this to work.
> 
> Unless I'm missing something?

This is something you should write yourself.  Do it by downloading the 
package, unpacking it, then extracting the data.  I suspect everything 
you need is already built into R, in the code that installs a package. 
You just need to rewrite it to do much less.

You shouldn't ask CRAN to do it, they've got enough on their plates.

Duncan Murdoch