S3 methods st_as_stars.CFDataset() and st_as_stars.CFVariable()

[pvanlaake]

This PR is in response to #756. It adds two new S3 methods, as mentioned in the title.

Additionally, it fixes two other issues:

1. The refsys "CFtime" was changed to include the CF calendar like "CFtime_noleap". This created problems with testing for string equality of the refsys in three locations.
2. There was a simple error in test-ncproxy.R, line 107: "GMT" changed to "UTC".

[edzer]

Great, thanks; time handling is now done in create_dimension(), my commit also passes on units in dimension values.

[edzer]

However, following R-CF/ncdfCF#13 this doesn't work:

library(ncdfCF)
library(stars)
# Loading required package: abind
# Loading required package: sf
# Linking to GEOS 3.12.2, GDAL 3.11.3, PROJ 9.4.1; sf_use_s2() is TRUE
ds <- open_ncdf("TEMPO_O3PROF_L2_V04_20250921T210803Z_S012G02.nc")
# There were 22 warnings (use warnings() to see them)
vars <- ds$var_names
op <- ds[[ vars[1L] ]]
class(op)
# [1] "CFVariable" "CFObject"   "R6"        
(op_ll <- op$subset(longitude = NA, latitude = NA, layer = 24, .resolution = 0.04))
# <Variable> ozone_profile 

# Values: [-0.156455 ... 13.96814] DU
#     NA: 542558 (77.3%)

# Axes:
#  axis name      long_name                  length values                     
#  X    longitude pixel index along slit      649   [-71.691716 ... -45.771716]
#  Y    latitude  scan mirror position index 1081   [17.398387 ... 60.598387]  
#       layer                                   1   [24]                       
#  unit         
#  degrees_east 
#  degrees_north
#               

# Attributes:
#  name         type     length value                         
#  comment      NC_CHAR  23     retrieved ozone profile       
#  units        NC_CHAR   2     DU                            
#  valid_min    NC_FLOAT  1     -100                          
#  valid_max    NC_FLOAT  1     100                           
#  _FillValue   NC_FLOAT  1     -1.26765060022823e+30         
#  coordinates  NC_CHAR  50     time ozone_profile_pressure...
#  actual_range NC_FLOAT  2     -0.156455, 13.968141          
s = st_as_stars(op_ll)
s
# Error in dim.stars(x) : length(d) == length(dim(x[[1]])) is not TRUE
# Calls: <Anonymous> ... print.stars -> cat -> dim -> dim.stars -> stopifnot
# Execution halted

because

> dim(s[[1]])
[1]  649 1081

and

> st_dimensions(s)
          from   to offset delta    refsys x/y
longitude    1  649 -71.71  0.04 OGC:CRS84 [x]
latitude     1 1081  17.38  0.04 OGC:CRS84 [y]
layer        1    1     24    NA        NA    

don't match: dim(s[[1]]) needs to be c(649,1081,1) or layer needs to be dropped:

> st_dimensions(s) = st_dimensions(s)[1:2]
> s
stars object with 2 dimensions and 1 attribute
attribute(s):
                         Min.  1st Qu.   Median    Mean  3rd Qu.     Max.
ozone_profile [DU] -0.1564551 6.509749 7.033861 7.18038 8.031126 13.96814
                     NA's
ozone_profile [DU] 542558
dimension(s):
          from   to offset delta    refsys x/y
longitude    1  649 -71.71  0.04 OGC:CRS84 [x]
latitude     1 1081  17.38  0.04 OGC:CRS84 [y]

[pvanlaake]

Ok, I know where this goes wrong. Degenerate dimensions in the data array are dropped and the corresponding dimension should thus not be created. I'll work on a solution.