Selector problems with reversed dims

[Mattriks]

I tried to extract time series from an ocean dataset e.g. ersstv5.nc that has Lat = Float32[88.0, 86.0, …, -86.0, -88.0]. The point 100°E, 40°N is clearly land, but extracting this point shows ocean temperatures:

sst = NCDstack("ersstv5.mnmean.nc")[:sst]
sst[Lat(At(40)), Lon(At(100))]

Float32[13.2772, 14.0097, 13.6713,   …  11.1825, 11.3001, 12.3442]

And the reflected point (which is in ocean) has all missing values:

sst[Lat(Near(-40)), Lon(Near(100))]

Float32[-9.96921f36, -9.96921f36, -9.96921f36,  …  -9.96921f36, -9.96921f36, -9.96921f36]

[rafaqz]

Yep I'm just in the process of trying to track down the last of these bugs.

Can you upload the file somewhere? That's probably the easiest way to fix this.

[Mattriks]

ftp:https://ftp.cdc.noaa.gov/Datasets/noaa.ersst.v5/sst.mon.ltm.1981-2010.nc (1mb)

[rafaqz]

It plots upside down too. My tests are are all with files with forward order latitudes, so thanks for catching this. There's a lot of detail to getting this all right for every format...

[rafaqz]

Ok fixed :)

I'll push it through sometime today.

[rafaqz]

Ok that's on master now. Let me know any other issues you have, I'm keen to make a release this week.

[rafaqz]

And by the way - for some reason NCDatasets,jl is not picking up the missing value in that file.

You can set it manually by passing the kwarg for the array constructor to the stack:

A = NCDstack(file; childkwargs=(missingval=-9.96921f36,))[:sst]

And convert it to missing with replace_missing(A) if you need to.