NOTES

NOTES

7/7/23

There are serious problems with the netcdf-c library version 4.9.1 and
4.9.2 that have broken EDL authentication. 4.9.0 does work, but trying
to use that with a conda env was just too hard. 

I switched to the python venv system and installed packages using pip
which, via stroke of good fortune, worked. 

7/10/23

I added pip-tools so that requirements.in --> requirements.txt works.
This should streamline running the notebooks using binder.

Pulled this out of the larger collection of notebooks to focus on this
jst about NASA's data and OPeNDAP servers.

7/13/23

Building a conda env using conda-forge and _not_ loading netCDf4 or xarray
worked with Binder. THere was no need to use a pre-built container nor was
there a need to separate the environment and content parts of the tutorials
into two repos (might be later...). I loaded netCDF4 and xarray in via pip
in the notebooks. 

Loading only xarray results in a broken `values` method/field. The scale_factor
attribute of the SST variable, e.g., is a string and not a number. using 
the netCDF library pulled in by xarray, the application of the scale factor 
fails with an inscrutable error message about loops. The fix is to load netCDF
4.9.3 or newer from pip.

Here is a StackOverflow post about the error message: https://stackoverflow.com/
questions/42013903/why-do-i-get-ufunc-multiply-did-not-contain-a-loop-with-signature-matching-ty