Handle CSV response format from SOS about owslib

[tomkralidis]

cc @kwilcox

@rsignell-usgs thanks for the info.

What about something like owslib.util.guess_format, which can then provide a hint to the caller on how to parse the response. In the end, it's either delimited, JSON, or XML (others)?

I would see something like this can be used by the entire codebase.

def guess_format(mime_type):
    if mime_type.find('xml') != -1:
        dformat = 'xml'  # parse with lxml.etree
    elif mime_type.find('csv') != -1:
        dformat = 'csv'  # parse with csv
    elif mime_type.find('tab-separated-values') != -1:
        dformat = 'tsv'  # parse with csv and tab delimiter
    elif mime_type.find('json') != -1:
        dformat = 'json'  # parse with simplejson
    return dformat

[rsignell-usgs]

That sounds great to me.

[kwilcox]

Is it important for OWSLib to parse the response from a GetObservation request at all? Would we be more inline with the spec to just return the raw results and let the user do what they want with it? They already need to provide a responseFormat parameter to the get_observation method so would have a good idea of what the result mimeType should be.

We could wrap a test for ows:ExceptionReport in a try block and return the raw response if the request was successful.

Something like:

response = openURL(base_url, data, method, username=self.username, password=self.password).read()
try:
    tr = etree.fromstring(response)
    if tr.tag == nspath_eval("ows:ExceptionReport", namespaces):
        raise ows.ExceptionReport(tr)
except ows.ExceptionReport:
    raise
except BaseException:
    return response

@tomkralidis Thoughts?

[tomkralidis]

@kwilcox I just realized that we return a string of the response, not an lxml.etree object of any kind.

In this case, agreed, forget the util.guess_format approach and +1 to your try/except block above.