16.0 years ago by
> I would like to use some of the simple annotation aspects of
> but have had difficulty building a full package (another question
> another day). However, I would like to be able to use, for example,
> alongchromosome from geneplotter without having to explicitly build
> package. I can fairly easily make an environment that contains (for
> example) chromosome location (in BP) or chromosome number mapped to
> geneIDs (from local data sources). How can I use these simple
> constructs rather than a data package (i.e., perhaps, what are the
> conceptual and practical differences)?
an R package is an entity that includes documentation and
initialization code that ensures reasonably standard
behavior of software across all platforms on which it is
installed. when installed in an image of R, all users of
that image can make use of the package using the 'library'
command, and can get access to documentation using the
help facility. the documentation can include validating
examples that run with the examples() function.
these are strong arguments for using the packaging protocol.
if you don't need to share the software (or data object),
don't need documentation to remind yourself of what the thing is,
and have a fairly simple approach to using R (only a single
.RData or .Rprofile that ensures that you can get access
to the saved environment of interest) then it is certainly
possible that you can just have a globally available function
or object that contains the data or functions of interest
without too much risk of confusion. and even if you don't
have such a simple approach to using R, you can establish
your own protocol for getting the unpackaged resources
into whatever image of R you happen to be using.
the package protocol is a successful approach to managing
software. those who ignore it may be condemned to reinvent it.