Reproject coordinates from a matrix or data frame by explicitly specifying the 'source' and 'target' projections.
# S3 method for sc
reproj(x, target = NULL, ..., source = NULL)
# S3 method for mesh3d
reproj(x, target, ..., source = NULL)
# S3 method for quadmesh
reproj(x, target, ..., source = NULL)
# S3 method for triangmesh
reproj(x, target, ..., source = NULL)
reproj(x, target, ..., source = NULL, four = FALSE)
# S3 method for matrix
reproj(x, target, ..., source = NULL, four = FALSE)
# S3 method for data.frame
reproj(x, target, ..., source = NULL, four = FALSE)
reproj_xy(x, target, ..., source = NULL)
reproj_xyz(x, target, ..., source = NULL)coordinates
target specification (PROJ.4 string or epsg code)
arguments passed to proj4::ptransform()
source specification (PROJ.4 string or epsg code)
if TRUE, and PROJ version 6 is available return four columns xyzt (not just three xyz)
numeric matrix of the transformed coordinates, either 2, 3, or 4 columns depending on the
shape of the input, or the argument 'four' in reproj(). Use reproj_xy() or
reproj_xyz() for those specific 2- and 3-column cases.
We currently use the proj4 package.
The reproj() and related functions drive proj4::ptransform() and sort out
the requirements for it so that we can simply give coordinates in data frame
or matrix form, with a source projection and a target projection.
If using PROJ, reproj can pass in a wider variety of source and target strings, not just "proj4string" and we are completely subject to the new rules and behaviours of the PROJ library. We always assume "visualization order", i.e. longitude then latitude, easting then northing (as X, Y).
The basic function reproj() takes input in generic form (matrix or data
frame) and returns a 3-column matrix, by
transforming from map projection specified by the source argument to that
specified by the target argument. Only column order is respected, column
names are ignored.
This model of working also allows adding methods for specific data formats
that already carry a suitable source projection string. Currently we
support types from the silicate and quadmesh and rgl packages, and only the
target string need be specified.
This model has obvious flexibility, for packages to import the generic and
call it with the correct source (from the data format) and the target
from user, or process controlled mechanism.
The source argument must be named, and if it is not present a light check
is made that the source data could be "longitude/latitude" and transformation
to target is applied (this can be controlled by setting options).
The function reproj() always returns a 3-column matrix unless four = TRUE, and PROJ package is available then a 4-column matrix is returned.
Functions reproj_xy() and reproj_xyz() are helpers for reproj() and always
return 2- or 3-column matrix respectively.
Note that any integer input for source or target will be formatted to a
character string like "EPSG:<integer_code>" as a simple convenience. Note that
there are other authorities besides EPSG, so the pattern "AUTH:code" is a general
one and you should really be explicit.
Until recently the proj4 package was the only one available for generic
data that will transform between arbitrary coordinate systems specified by
source and target coordinate systems and with control over 'xy' versus
'xyz' input and output. This package adds some further features by wrapping
the need to convert longitude/latitude data to or from radians.
Other R packages for transforming coordinates are geared toward data that's in a particular format. It's true that only GDAL provides the full gamut of available geographic map projections, but this leaves a huge variety of workflows and applications that don't need that level of functionality.
The PROJ package is a stub atm and is not used.
The proj4 package works perfectly well with the PROJ-lib at versions 4, 5, 6, or 7 and if this is preferred reproj can be set to ignore the PROJ R package (see reproj-package).
The behaviour is controlled by user-settable options which on start up are
reproj.assume.longlat = TRUE and
reproj.default.longlat = "OGC:CRS84".
If the option reproj.assume.longlat is set to FALSE then the source
argument must be named explicitly, i.e. reproj(xy, t_srs, source = s_srs),
this is to help catch mistakes being made. The target is the second
argument in reproj though it is the third argument in proj4::ptransform.
This function also converts to radians on input or output as required.
If the option reproj.assume.longlat is set to TRUE and the input data
appear to be sensible longitude/latitude values, then the value of
reproj.default.longlat is used as the assumed source projection.
See reproj-package for another option set reproj.mock.noproj6 for package
testing for expert use.
There are a number of limitations to the PROJ library please use at your own risk. The sf package provides a better supported facility. The libproj package will be used if it makes it to CRAN.