TDSCatalog does not include base path in access_url about siphon

[jm-cook]

Might be related to #114

However that was fixed?

[jm-cook]

I looked some more into this since I was curious as to why the fix to #114 does not solve my issue. I ran my little test on the oceandata opendap catalog link in #114 (now updated to https://oceandata.sci.gsfc.nasa.gov/opendap/SeaWiFS/L3SMI/2000/0101/catalog.xml) and see the same issue, ie opendap/hyrax is not inserted into the url.

What is happening is that when access_urls is constructed in make_access_urls(), the server_base url is correctly constructed as 'https://oceandata.sci.gsfc.nasa.gov/opendap/hyrax', the url_path is obtained as the absolute path /SeaWiFS/L3SMI/2000/0101/SEASTAR_SEAWIFS_GAC.20000101.L3m.DAY.CHL.chlor_a.9km.nc but then on the next line

                    access_urls[subservice.service_type] = urljoin(server_base,  self.url_path)

urljoin will create the absolute url, so the opendap/hyrax part of the server_base url is lost.

This seems to be how hyrax is presenting the paths (ie with a leading slash).

According to the documentation here: https://docs.unidata.ucar.edu/tds/5.2/userguide/basic_client_catalog.html, the service base, urlPath, and dataset should be concatenated together.

The code below shows the incorrectly constructed access_urls (the urls created cannot be accessed):

from siphon.catalog import TDSCatalog
cat_url = "https://oceandata.sci.gsfc.nasa.gov/opendap/SeaWiFS/L3SMI/2000/0101/catalog.xml"
cat = TDSCatalog(cat_url)
for cds in cat.datasets:
    url = cat.datasets[cds].access_urls['dap']
    print(f'access url = {url}')