with these changes, pydap sets endianness correctly, and ONLY ONCE (from 1st chunk header). With these changes, pydap can correctly read DAP4 data from tds and hyrax without any extra logic, and importantly, in a correct way.

Before the following only worked for Hyrax (it did not for the tds):

url_tds = "https://thredds-test.unidata.ucar.edu/thredds/dap4/dev/d4icomp/SimpleGroup.nc4.dap?dap4.ce=/SimpleGroup/X&dap4.checksum=true"
url_hds = 'http://test.opendap.org:8080/opendap/dap4/SimpleGroup.nc4.h5.dap?dap4.ce=/SimpleGroup/X'

X is an array on int16 spanning from 0 to 39.

With pydap I get:

dataset_tds['/SimpleGroup/X']
>>> <BaseType with data array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
       17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
       34, 35, 36, 37, 38, 39], dtype=int16)>

and

dataset_hds['/SimpleGroup/X']
>>> <BaseType with data array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15, 16,
       17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33,
       34, 35, 36, 37, 38, 39], dtype=int16)>

Still draft mode. I need to refactor several things. And test against other data types in that same file.

ready for review! It sets the endianness of the data by reading the first chunk (before the DMR). It makes the assumption that the endianness does not change (tds approach). I check the datasets on both tds and hyrax servers and they look identical to the original dataset (which I created). Both 1D and 2D arrays (int16 and float32).

It removes any logic identifying whether the data is served by a TDS or a Hyrax DS.

@ndp-opendap