Fix syntax error on Python 3.5 and wrong data type

[uvchik]

1. On Python 3.5 the comma at the end of the constructor raises a SyntaxError.

2. The data type of the '**extra_imshow_args` is not dict but undefined or various or just empty.

3. Shouldn't it be named extra_imshow_kwargs as it collects "keyword arguments" and not unnamed "arguments". But this is not very important, so I did not changed by now. I can change it if you agree.

[coveralls]

Coverage remained the same at 92.938% when pulling bc9799f on uvchik:master into dc6746f on darribas:master.

[darribas]

This looks good to me but will defer on some of the technical details to @jorisvandenbossche as he's more knowledgeable about these standards...

[jorisvandenbossche]

The fix itself is fine, but, in general I think we need to decide if we want to support Python 3.5, and then either test it, or either officially drop support for it.
We currently still support it according to setup.py:

https://github.com/darribas/contextily/blob/a2ab3d7f4fb41bfb299cb2604ce24d69a45b5d42/setup.py#L32

But personally I think it would be fine to drop (eg also pandas will drop py 3.5 in the next release)

[jorisvandenbossche]

Shouldn't it be named extra_imshow_kwargs as it collects "keyword arguments" and not unnamed "arguments". But this is not very important, so I did not changed by now. I can change it if you agree.

Yes, that's true. Since it is not an argument you actually specify as a user, I don't think it is that important to fix, but feel free to do a PR to rename!

[uvchik]

But personally I think it would be fine to drop (eg also pandas will drop py 3.5 in the next release)

As debian 10 is released, I do agree. I used to use debian 9 with python 3.5 until last week.

[darribas]

Thanks both!!!

My view is that if it's a small thing like the one raised in this issue, we could keep it compatible, as a one-off. But by not officially supporting/testing 3.5 should send the message that it might work, but if it doesn't we won't sweat to fix it. Particularly with pandas dropping it, it should be come less and less of a common option.