[R-sig-Geo] metatdata

[Barry Rowlingson]

On Mon, Nov 9, 2009 at 7:01 AM, Agustin Lobo <alobolistas at gmail.com> wrote:
> Hi!
> Is anyone thinking on designing metadata for R spatial objects?
> I've seen there are metadata facilities in Bioconductor for genomics,
> perhaps we could get some inspiration from them?

 I think you need to give us some examples of what you'd like to do. A
lot of work on metadata is data-type-agnostic, so is a general R
question anyway and hence something for R-dev - for example
implementing Dublin Core metadata for all R objects. This already
seems to be implemented for the tm packages:

http://finzi.psych.upenn.edu/R/library/tm/html/meta.html

 As Roger has said, it's tricky to keep attributes and metadata stuck
to R objects once you start manipulating them, assuming that's what
you want to do. It may be that you just want a set of classes for
handling spatial metadata (and this I think is a prerequisite to being
able to store them attached to R objects anyway). I've just had a look
at the OGC web site and there's a few standards for metadata there -
mostly in conjunction with catalogue service design. Check that out.

 I find the best thing anyone can do when saying 'can we do something
with X' is to give a 'user story' - including (fictional) transcripts,
example:

# fred wants to identify himself as the creator of this map:
 map = readOGR("foo.shp")
 metadata(map,"author")="Fred Foo"

 - that's an example of tightly-coupled metadata, the kind that R will
probably drop if you do map2=map1[1:10,]

# fred wants to keep track of who is the creator of this map:
 map = readOGR("foo.shp")
 mapMetadata=DublinCore(author="Fred Foo",...)

 - that's an example of loosely-coupled metadata, stored in a separate
object and subject to formal constraints on structure. Any updates are
the responsibility of the user.

Barry