use shapely >=1.8

[ocefpaf]

Follow up from #70 (comment)

[ocefpaf]

One test is failing with:

E             Full diff:
E               {
E                'attributes': {'geospatial_bbox': 'POLYGON ((198.669 61.777, 198.669 67.068, '
E             +                                    '174.79200000000003 67.068, '
E             -                                    '174.792 67.068, 174.792 61.777, 198.669 '
E             ?                                            ---- ^^^^^^^^^^^
E             +                                    '174.79200000000003 61.777, 198.669 '
E             ?                                             ^^^^^^^^^^
E                                                  '61.777))',
E             -                 'geospatial_bounds': 'POLYGON ((174.792 61.777, 174.926 '
E             ?                                                                 --------
E             +                 'geospatial_bounds': 'POLYGON ((174.79200000000003 61.777, '
E             ?                                                        +++++++++++
E             +                                      '174.92599999999993 62.206, 178.812 '
E             -                                      '62.206, 178.838667737 64.055605136, '
E             -                                      '178.916 64.084, 179.858 64.311, 192.86 '
E             -                                      '67.029, 196.86 67.068, 198.669 66.861, '
E             ?                                                              ----------------
E             +                                      '64.098, 192.86 67.029, 196.86 67.068, '
E             ?                                       +++++++++++++++
E             -                                      '187.753767857 64.334204193, 179.195 '
E             +                                      '197.094 67.044, 198.669 66.861, 187.784 '
E             +                                      '64.188, 179.10799999999995 62.266, '
E             +                                      '176.16899999999998 61.862, '
E             -                                      '62.395, 176.169 61.862, 174.792 61.777))',
E             ?                                       ------------------------
E             +                                      '174.79200000000003 61.777))',
E             ?                                              +++++++++++

Some are just floating points diffs like 174.79200000000003 vs 174.792 in the bbox. But in the bounds we have some that may be a mistake, like 178.838667737 64.055605136, 178.916 vs 178.812 64.098, 192.86.

[kwilcox]

I'm not surprised by the diff, it's now using different methods to "fix" the dateline crossing. I'll plot them both and make sure its still a valid "fix"...

[ocefpaf]

I'll plot them both and make sure its still a valid "fix"...

@kwilcox I know you are busy and hopefully this notebook will help you out to make a call here:

https://nbviewer.org/urls/gist.githubusercontent.com/ocefpaf/e6afad88f3063279ddcc21efa2847b89/raw/1b66d04b6381abb622e9e98e3b177b4838a99bbb/gistfile1.txt

If you are OK with the new polygons I can adjust the tests.

---

edit: I sent the commit with the value changes. We can revert if you disagree with the comparison in the notebook.