An alternate rendering: where DBI::dbFetch() gives the tidy
long-form data frame, xr_raster() gives the slab in the same
in-memory form as gdalraster read_ds() a plain vector of pixel
values in left-to-right, top-to-bottom order, band-sequential, with
a gis attribute carrying type, bbox, dim, srs, and
datatype. No dataset is created and no bytes go through GDAL; the
array converts directly via numpy (no pandas hop), and gdalraster
is not needed at all.
Usage
xr_raster(
x,
statement = NULL,
as_list = FALSE,
max_cells = getOption("xrdbi.max_cells", 1e+07),
force = FALSE,
params = NULL
)Arguments
- x
an XarrayResult, or an XarrayConnection (with
statement)- statement
a Python expression, when
xis a connection- as_list
if TRUE, a list of band vectors instead of one band-sequential vector
- max_cells
guard threshold, as elsewhere
- force
bypass the guard
- params
named list of values the statement refers to, when
xis a connection (see XarrayResult)
Value
a vector (or list of band vectors) with attribute gis,
interchangeable with the output of gdalraster read_ds
Details
The selected object must be a DataArray that squeezes to 2 or 3
dimensions: (y, x) for a single band, (band, y, x) for a stack
(e.g. a time series of slabs). y and x are taken by position, the
last two dimensions, whatever their names: an array stored (lon, lat)
renders transposed, so put it in order in the statement
(.transpose('lat', 'lon')). The y and x coordinates must be 1-D,
numeric and regular (to their storage precision: float32 grids are
fine); the half-cell bbox expansion assumes cell-centre coordinates.
Ascending or descending coordinate order is handled on either axis
(rows and columns are reordered into top-left pixel order).
Curvilinear grids have no affine rendering, and vector data cubes
(an xvec geometry dimension) no raster at all; both error with the
reason.
srs is the CRS found in this order: an index carrying one (xproj's
CRSIndex, as gdalxarray sets; rasterix's RasterIndex); WKT in a
crs_wkt or spatial_ref attribute of a spatial_ref/crs
coordinate, of the variable named by the CF grid_mapping attribute,
or of a dataset variable named spatial_ref or crs; and last the
CF parameters of that grid_mapping variable, through pyproj when it
is installed. "" when none is found.
See also
xr_chunks() for the storage chunks under a slab;
XarrayResult for statements and params; xrdbi-package for an overview.
Examples
if (FALSE) { # \dontrun{
## needs a Python environment with xarray (see ?xrdbi)
f <- system.file("extdata/autotest/trmm.nc", package = "xrdbi")
con <- dbConnect(xarray(), f)
r <- xr_raster(con, "ds.pcp.isel(time=0)")
attr(r, "gis")$dim
attr(r, "gis")$bbox
## or from a result
res <- dbSendQuery(con, "ds.pcp.isel(time=0)")
r <- xr_raster(res)
dbClearResult(res)
} # }