Broken argopy link

[gmaze]

Hi Gael,
Looking at the last release, I noticed that the first argopy link at:
https://gaelforget.net/notebooks/Argo_argopy.html
is broken
it could point toward: https://github.com/euroargodev/argopy

[gaelforget]

For context : the html that was linked in the docs is the last one I managed to make work.

Update : I updated the GitHub link to https://github.com/euroargodev/argopy#readme and resurrected running the notebook example

But : the test case returns this error :

julia> argopy=ArgoData.pyimport(:argopy)
PyObject <module 'argopy' from '/Users/gaelforget/.julia/conda/3/aarch64/lib/python3.10/site-packages/argopy/__init__.py'>

julia> ds_fetcher=argopy.DataFetcher().float([6902746, 6902747, 6902757, 6902766])
PyObject <datafetcher.erddap>
Name: Ifremer erddap Argo data fetcher for floats
API: https://erddap.ifremer.fr/erddap
Domain: WMO6902746;WMO6902747;WMO6902757;WMO6902766
Performances: cache=False, parallel=False
User mode: standard
Dataset: phy

julia> ds_points = ds_fetcher.to_xarray()
ERROR: PyError ($(Expr(:escape, :(ccall(#= /Users/gaelforget/.julia/packages/PyCall/1gn3u/src/pyfncall.jl:43 =# @pysym(:PyObject_Call), PyPtr, (PyPtr, PyPtr, PyPtr), o, pyargsptr, kw))))) <class 'aiohttp.client_exceptions.ClientConnectorCertificateError'>
ClientConnectorCertificateError(ConnectionKey(host='erddap.ifremer.fr', port=443, is_ssl=True, ssl=True, proxy=None, proxy_auth=None, proxy_headers_hash=None), SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1007)'))
  File "/Users/gaelforget/.julia/conda/3/aarch64/lib/python3.10/site-packages/argopy/fetchers.py", line 490, in to_xarray
    xds = self.fetcher.to_xarray(**kwargs)
...

How can we fix this issue?

ps. I also get this error on MAC arm64, but the above ran via GitHub action on linux
pps. the argopy docs also seems to fail with erddap here and here; not sure if its the same error though

[gmaze]

Hi @gaelforget
I identified the issue a few weeks ago
this has to do with some recent updates in xarray
apparently, everything works ok for xarray < 2024.3
I'm working on fixing the issue and making argopy compatible with the most recent xarray versions
this should be available with the next 0.1.16 argopy release

[gmaze]

while pydata/xarray#8909 is not solved we have to use xarray < 2024.3

[gaelforget]

Hi @gaelforget I identified the issue a few weeks ago this has to do with some recent updates in xarray apparently, everything works ok for xarray < 2024.3 I'm working on fixing the issue and making argopy compatible with the most recent xarray versions this should be available with the next 0.1.16 argopy release

Thanks for explaining. Will keep an eye out for 0.1.16 and try again when it's out

[gaelforget]

while pydata/xarray#8909 is not solved we have to use xarray < 2024.3

I don't know how to do modify the install / use instruction of argopy to do this.

Is there an argument that you can pass to pip install argopy or conda install to impose such restriction at install time? or an argument to import or DataFetcher to select xarray 2024.3 at run time?

If we can do this in python then we should be able to modify one of the relevant lines of Julia below :

ArgoData.jl/ext/ArgoDataCondaExt.jl

Line 7 in c3b53be

Conda.add("argopy")

ArgoData.jl/ext/ArgoDataPyCallExt.jl

Line 7 in c3b53be

PyCall.pyimport("argopy")

ArgoData.jl/examples/Argo_argopy.jl

Line 95 in c3b53be

ds_points = ds_fetcher.to_xarray()

I generally expect that dependencies enforce their own dependency compatibility restrictions. In Julia that's handled by the compat section in Project.toml. For Python / argopy I see it's done in requirement.txt

https://github.com/euroargodev/argopy/blob/289a4aa8391cb4cf08a605bbe4dc23a945e9081b/requirements.txt#L1

I tried pip install git+http://github.com/euroargodev/argopy.git@master but then import argopy fails for me. Also I don't know that I can pip install from github via Conda.jl. So I think I need to wait for argopy 0.1.16 to be released.

[gmaze]

Indeed the restriction on xarray is done by requirements.txt
So installing argopy from the repo master branch with pip should make it
but that won't be as clean as a conda install, and for that you indeed need to wait for the next release

[gmaze]

rq:
if xarray > 2024.3 is installed before argopy, then I'm not sure pip will raise an error despite the requirements.txt
you may need to check when you install xarray
with pip you can pin a version or an upper bound with:
pip install xarray < 2024.3.0