Skip to content

Fields used by available.packages

4 messages · Lluís Revilla, Kurt Hornik

#
Hi all,

I have recently been researching how available.packages and
install.packages filter packages from repositories with additional fields
in their PACKAGES file.

Currently there are some default filters, but users (and R admins) can set
up their own filters by passing a list to the fields argument or adding
them to the "available_packages_filters" option.
But if the fields used by the filters are not read by default, then users
must manually add the required fields to each call to available.packages.

This makes it difficult to use new fields and to control what is installed
in highly regulated systems which want to use more fields to select what is
installed.

Current workarounds considered are:
 1) The filtering function requiring new fields intercepts the call to
available.packages and adds the desired fields via eval in
parent.environment and then adds the filters again.
 2) Import new data (remote or local) when filtering packages, merge them
and filter accordingly.
 3) Suggestions?

The first solution is complicated, while the second doesn't use the R
machinery of tools::write_PACKAGES to set up the repository with all the
fields (although how to add more fields to the repository file is a
different issue).

Would it be possible to add a new option to add fields to be read by
available.packages, similar to filters?
The same approach for fields as for filters would avoid the two workarounds
mentioned. To match it, the new option could be named
"available_packages_fields".

I look forward to hearing from you,

Llu?s
3 days later
#
Lluis,

So in available.packages() I could replace

    if (is.null(fields))
	fields <- requiredFields
    else {
	stopifnot(is.character(fields))
	fields <- unique(c(requiredFields, fields))
    }

by someting like

    if(is.null(fields))
        fields <- getOption("available_packages_fields")
    if(is.null(fields))
        fields <- requiredFields
    else {
	stopifnot(is.character(fields))
	fields <- unique(c(requiredFields, fields))
    }

?

Best
-k

        
#
Yes, I think that would be enough.
Thank you, Kurt!

Llu?s
On Wed, 12 Jun 2024 at 16:35, Kurt Hornik <Kurt.Hornik at wu.ac.at> wrote:

            

  
  
#
Thanks.  And of course add documentation ...

Will try to get this in in the next few days.

Best
-k