subset-user-input.ipynb: fix output for Thredds v5

[tlvu]

Only merge if bird-house/birdhouse-deploy#413 can be merged.

This is mostly to have a nice notebook diff.

15:47:13  ___ pavics-sdi-master/docs/source/notebooks/subset-user-input.ipynb::Cell 12 ___
15:47:13  Notebook cell execution failed
15:47:13  Cell 12: Cell outputs differ
15:47:13  
15:47:13  Input:
15:47:13  # gather data from the PAVICS data catalogue
15:47:13  catalog = "https://boreas.ouranos.ca/twitcher/ows/proxy/thredds/catalog/datasets/gridded_obs/catalog.xml"  # TEST_USE_PROD_DATA
15:47:13  
15:47:13  cat = TDSCatalog(catalog)
15:47:13  data = cat.datasets[0].access_urls["OPENDAP"]
15:47:13  data
15:47:13  
15:47:13  Traceback:
15:47:13   mismatch 'text/plain'
15:47:13  
15:47:13   assert reference_output == test_output failed:
15:47:13  
15:47:13    "'https://pav...rcan_v2.ncml/'" == "'https://pav...s/nrcan.ncml'"
15:47:13    Skipping 70 identical leading characters in diff, use -v to show
15:47:13    - _obs/nrcan.ncml'
15:47:13    + _obs/nrcan_v2.ncml'
15:47:13    ?           +++
15:47:13

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[tlvu]

@tlogan2000 is nrcan_v2.ncml and nrcan.ncml equivalent?

[tlogan2000]

@tlogan2000 is nrcan_v2.ncml and nrcan.ncml equivalent?

It's not the same data no

[tlvu]

@tlogan2000 is nrcan_v2.ncml and nrcan.ncml equivalent?

It's not the same data no

So it is not expected that with the new Thredds v5, nrcan_v2.ncml is used instead of nrcan.ncml?

[tlogan2000]

@tvlu THREDDS sorting of datasets seems to have changed between v4 and v5. Two options to fix.

1. Leave as is and accept that THREDDS v5 finds nrcan_v1 in the [0] index of the catalogue search. The example is still valid in terms of operations (subsetting and calculations etc etc)

or

2. changing the line

data = cat.datasets[0].access_urls["OPENDAP"]

to

data = cat.datasets[-1].access_urls["OPENDAP"]

should ensure that v2 is found instead of v1

[tlvu]

Is there an access by name instead of by index? Otherwise, the day we have nrcan_v3.ncml this will change again !

Also, whether we choose cat.datasets[0] or cat.datasets[1] we can not merge immediately and have to wait until Thredds v5 go-live to merge, else it won't be backward compatible with the existing Thredds v4.

[tlogan2000]

Is there an access by name instead of by index

Nothing built in but I change the code to search for nrcan_v2 should be future proof now

[tlvu]

Nothing built in but I change the code to search for nrcan_v2 should be future proof now

Super, I'd do a "refresh" with Thredds v5 and test against our current Thredds v4. If it works, we can merge this immediately and be forward compatible with Thredds v5 !

[tlvu]

Opps, closed by mistake.

[tlvu]

Replace by #342