[R-sig-Geo] problem assigning POSIX class to AttributeList column

[Martin Maechler]

>>>>> "Edzer" == Edzer J Pebesma <e.pebesma at geo.uu.nl>
>>>>>     on Fri, 25 Aug 2006 09:54:11 +0200 writes:

    Edzer> Mike, in reaction to
    Edzer> https://stat.ethz.ch/pipermail/r-sig-geo/2006-July/001189.html

    Edzer> on http://www.geog.uu.nl/~pebesma/sp_0.9-0.tar.gz I
    Edzer> prepared an sp version where the data slot for
    Edzer> storing attribute data of classes
    Edzer> SpatialPointsDataFrame, SpatialGridDataFrame and
    Edzer> SpatialPixelsDataFrame is (again) a genuine
    Edzer> data.frame (this used to the the AttributeList).
    Edzer> There's little point in further developing the
    Edzer> AttributeList class as the problems we had with the
    Edzer> data.frame class -- excessive memory, disk and CPU
    Edzer> usage for creating huge character row.names vectors,
    Edzer> see e.g.  the little example in the R-News 5/2 --
    Edzer> have been solved in R 2.4.0; see
    Edzer> https://stat.ethz.ch/pipermail/r-devel/2006-April/037306.html

    Edzer> The upgrade to 2.4.0 seems to be binary incompatible
    Edzer> for S4 objects, 

well, 'yes'  if "binary incompatible" means that S4 objects saved
in older versions of R are not quite what they are when you
recreate them in R devel (2.4.0 to be).
I.e. if they belong to your package and are saved as <pkg>/data/<foo>.rda
you'll want recreate them eventually 
and then have your package depend on R 2.4.0 -- which is a bit
undesirable, unless you are willing to maintain too parallel
versions of your package, one for R <= 2.3.x, and another for
(current) R-devel and newer.

For the 'Matrix' package, we decided for the time being to use
./data/*.R scripts reading ./external/*.rda files which contain
the raw data and then recreate our S4 data objects at data load time.  

This has the one advantage that it does not depend on R versions
and even still works if we slightly change class definitions.


    Edzer> so that may be a good moment to make
    Edzer> this binary incompatible change in sp (see
    Edzer> https://stat.ethz.ch/pipermail/r-devel/2006-August/038811.html). In
    Edzer> general, we do recommend sp users to always keep the
    Edzer> scripts that generate or manipulate their data.

Indeed!  And that recommendation goes for all of using R, not
just "geo"-like use.

    Edzer> You, or others, may want to use it, along with R
    Edzer> 2.3.1 or 2.4.0 to solve your current problem.

    Edzer> Best regards, -- Edzer

    Edzer> _______________________________________________
    Edzer> R-sig-Geo mailing list R-sig-Geo at stat.math.ethz.ch
    Edzer> https://stat.ethz.ch/mailman/listinfo/r-sig-geo