More reliable way to identify opendap url

[davidcaron]

More reliable way to identify opendap url

[dbyrns]

You should check this function written by Blaise : https://github.com/Ouranosinc/flyingpigeon/blob/master/flyingpigeon/utils.py#L59

[davidcaron]

The problem with doing a try/except is that when the url is not an Opendap url, it takes a second or 2 to raise an IOError. Also, I tried it a while ago and when it's not an opendap url it crashed the interpreter without raising an error... It may have had to do with my installation of netcdf, I'm not sure.

But the netCDF4 library doesn't seem to support constraint expressions, as pointed out by #20.

Knowing the structure of a constraint expression, (it seems to be located only in query parameters) we can still use the technique of appending any standard opendap suffix to the url (.dds, .das, .ver, ...) by removing the query string. But that url will still be passed to xarray or netcdf4, and it won't work unless we implement #20. So these 2 issues are related.

[huard]

Got bit by this recently. Can it be dealt with client-side ? The client could then pass the mimetype of the input, and then we would not have to guess server-side.

[davidcaron]

What was the url you sent to the server?

[davidcaron]

I did a quick research, and I think that if we make a HEAD request to the url, the Content-Description header should always start with dods if it's an opendap url.

Check the section 7.1.1 at: https://www.opendap.org/pdf/ESE-RFC-004v1.2.pdf

Edit: I also found this:

Theres an ambiguity as to whether http://server/something is an OPeNDAP or an HTTP remote file using range requests. There will be more ambiguities in the future, as other HTTP based protocols are added. Currently we do a HEAD request on http://server/something.dds, and if it succeeds, and returns a header Content-Description="dods-dds" or "dods_dds", then we open as OPeNDAP, and if it fails we try opening as an HTTP file.

https://www.unidata.ucar.edu/software/netcdf-java/current/reference/DatasetUrls.html

So I think this is the way to go.

[huard]

It was a THREDDS DAP url but starting with http. I had to write href=http... for pywps to recognize it as URL and not as data, but that may be a quirk of pywps.