Do not call shiftdata for scattter plots

[guziy]

I've added a test for the scatter function and modified the decorator to eliminate the call to shiftdata for scatter in 1-d case, since I am not really sure why it is needed....

Addresses the issue #197.

Cheers

[guziy]

The tests are added to show that it works with 1 and 2 points. (#191)

[pelson]

@WeatherGod - did you have a case where this is not a good idea?

[WeatherGod]

I don't think so. It was a few years ago that I reviewed this original
feature. I can't remember why we applied it to scatter plots. The only
thing I can think of is one is doing an image and scatter. One would want
to be able to have the domains for them match up, right?

On Thu, Sep 3, 2015 at 5:07 AM, Phil Elson [email protected] wrote:

@WeatherGod https://github.com/WeatherGod - did you have a case where
this is not a good idea?

—
Reply to this email directly or view it on GitHub
#202 (comment).

[guziy]

@WeatherGod I'll try to generate a test case for this, to see if it really is a problem.

Cheers

[guziy]

Apparently, shiftdata is needed for shifting when longitudes are further than +/- 180 from the lon_0:

https://github.com/guziy/PyNotebooks/blob/master/test_scatter_and_shiftdata.ipynb

I was supposing that it is done in the m(lon, lat) call....

One way to patch it would be a special shiftdata for scatter, which does not care about relations, but only changes coordinates to be within lon_0 - 180 and lon_0 + 180 range... Or change the call method, however, I am not sure how it would impact other plotting methods.

Cheers

[guziy]

The plot with shiftdata does not seem right when you look at the latitudes of the rightmost part...

[guziy]

I was thinking about this a bit more, I think there should be a better way to solve this, not hacking only scatter and then maybe others... I'll try to understand shiftdata and fix it if possible...

Cheers