This dataset is not intended for users, but rather for internal use
within the gsw package. The dataset stores the 1.4M lookup
table defined in the 8.3M file src/gsw_saar_data.c in the C
library. (The .c file exceeds CRAN limitations on size.)
The data are designed to replace C elements defined as below
in src/gsw_saar_data.c:
static int gsw_nx=91, gsw_ny=45, gsw_nz=45;
static double longs_ref[91];
static double lats_ref[45];
static double p_ref[45];
static double ndepth_ref[4095];
static double saar_ref[184275];
static double delta_sa_ref[184275];
R storage is in a list named saar, with elements named
as in the C code, i.e. gsw_nx etc.
C storage for these variables is allocated as needed,
and the data are inserted, when gsw is launched.
Thus, the existing C library code "knows" about the data
as local storage, which keeps alterations to the C library to
a minimum.
The saar dataset was created by the following R code. The
netcdf file used in this code comes from the GSW-Fortran
repository (at commit baa0c09ffc7ed1f74972a1a2902d8754caa5b4cb)
and its md5 value is dacb3f981e8e710ac2e83477701b3905.
library(ncdf4)
nc <- nc_open("~/git/GSW-Fortran/test/gsw_data_v3_0.nc")
## Use as.vector() since these will all get handed into C, which does not understand matrices.
p_ref <- as.vector(ncvar_get(nc, "p_ref"))
lats_ref <- as.vector(ncvar_get(nc, "lats_ref"))
longs_ref <- as.vector(ncvar_get(nc, "longs_ref"))
ndepth_ref <- as.vector(ncvar_get(nc, "ndepth_ref"))
ndepth_ref[!is.finite(ndepth_ref)] <- -9e99
saar_ref <- as.vector(ncvar_get(nc, "SAAR_ref"))
saar_ref[!is.finite(saar_ref)] <- -9e99
delta_sa_ref <- as.vector(ncvar_get(nc, "deltaSA_ref"))
delta_sa_ref[!is.finite(delta_sa_ref)] <- -9e99
saar <- list(gsw_nx=gsw_nx, gsw_ny=gsw_ny, gsw_nz=gsw_nz,
longs_ref=longs_ref, lats_ref=lats_ref, p_ref=p_ref, ndepth_ref=ndepth_ref,
saar_ref=saar_ref, delta_sa_ref=delta_sa_ref)
save(saar, file="saar.rda")
tools::resaveRdaFiles("saar.rda")
nc_close(nc)