fix broken support to py2.7 #256

[ccancellieri]

This may fix #256

[amercader]

Can you use six for the python version checking? eg

import six

if six.PY3:
    # ...
else:
    # ...

or even

_encoding = 'utf-8' if six.PY3 else str

But looking at the documentation it looks like the most portable solution would be to pass
record["xml"] = etree.tostring(mdtree, pretty_print=True, encoding='utf-8') regardless of the Python version. Does this work as expected on Python 2.7 (and 3)?

[ccancellieri]

It was 'unicode' and have been change into 'str' by:
e063522#diff-db75a495bf76416216c3b4fe8cc1038bc53d22a19cb0990cbd823b17c2d67364R104

Now I can't be sure about py3 (don't have a working environment) but you could be right.
To avoid causing problem I proposed encoding='utf-8' for py2 and 'as it is' (futurized) for py3.

Can someone confirm about py3?

[frafra]

It seems ok to me.
There are tests now, so such change could be made, and we see how the CI tests go :)

[amercader]

@ccancellieri The aim is to support Python 2 for versions of CKAN <= 2.9, and the tests are run against these versions (see eg https://github.com/ckan/ckanext-spatial/actions/runs/1394064993). If there is a Python 2 failure it means that test coverage is lacking there and something slipped through the cracks.