Geoscience Reference
In-Depth Information
client or another addressee. INTAMAP* shows how such a WPS may be configured, here for provid-
ing an interoperable framework for real-time automatic mapping of critical environmental variables
(Pebesma et al. 2011b). Another example of an OGC web standard is for an OGC Sensor Observation
Service (SOS) client, described by Nüst et al. (2011) and written as an
R
package
sos4R
.
†
The WPS and SOS examples both use
R
as part of their workflow, providing a convenient intro-
duction to the next section, in which we will show how geospatial software components may be
stacked with the
R
language and environment.
14.4 OSGeo AND R-SPATIAL
The
R
open-source programming language and environment (R Development Core Team 2011) is
understandably associated with data analysis and statistical programming. As a general interpreted
programming language, it is not limited to these tasks and can be applied to computational tasks
of many kinds. C, Fortran and C++ libraries may be interfaced with R functions through simple
mechanisms, which also permit access to R objects in compiled languages and callback to R from
compiled functions. In addition, R can be embedded within the PostgreSQL database as the proce-
dural language PL/R
‡
; it can also be interfaced with Python and other languages.
The class and method definitions used in
R
are covered in detail by Chambers (2008) and permit
data objects to be represented and handled in adequate ways. These definitions advance in waves,
with many fitted model objects using old-style classes and many data objects using new-style classes.
Old-style and new-style generic methods have also recently been brought closer together. A further
innovation of some note is the
Rcpp
package, providing a more modern interface between
R
code
and compiled C++ code, described by Eddelbuettel and Francois (2011) and in a Google TechTalk.
§
Fox (2009) reports on the development of the
R
project from a sociological viewpoint, based on
semi-structured interviews carried out in 2006 and 2007. He points to salient features of the social
organisation of the project that have enabled it to provide both a stable platform with strong conti-
nuity in functionality and a rich community encouraged to contribute software packages extending
the base functionality.
Prospects and challenges in
R
package development are discussed by Theußl et al. (2011); they
address some of the issues raised earlier with regard to the consequences of version and platform
drift for community-contributed add-ons. Contributed packages distributed through the compre-
hensive
R
archive network (CRAN) are now very numerous and have complex dependency struc-
tures. Cross-platform checks run on CRAN packages using multiple versions of
R
ranging from
released to development need careful planning to capture inter-package dependencies correctly and
to minimise the need for administrator intervention when things break, which they inevitably do
when changes are made.
14.4.1 r-S
Patial
: sp
In Bivand et al. (2008), we discuss the background for providing spatial data analysis functional-
ity in the R environment and how the need emerged for classes for spatial data (see also Câmara
et al. 2012, p. 238). From 2003, we attempted to make available mechanisms permitting a user
and developer community to grow. The R-sig-geo mailing list had in March 2012 over 2400 sub-
scribers, and Figure 14.2 shows the steady growth in the numbers of messages exchanged since its
inception. It is now the R list with most traffic, naturally except the main R-help and R-devel lists.
Mailing lists remain a vital part of open-source communities, connecting users with each other and
*
http://www.intamap.org.
†
http://www.nordholmen.net/sos4r/.
‡
http://www.joeconway.com/plr/doc/index.html.
§
http://www.youtube.com/watch?v=UZkaZhsOfT4.
