annotate

[shaharkadmiel]

It is a bit cumbersome and unintuitive to use the annotate method with GeoAxes (see this discussion on stackoverflow).

To make the ax.annotate method more like the other GeoAxes methods (like plot, scatter, etc.), I suggest this thin wrapper:

class GeoAxes():
    ...

    def annotate(self, *args, **kwargs):
        crs = kwargs.pop('transform', self.projection)
        transform = crs._as_mpl_transform(self)
        kwargs['xycoords'] = transform
        xy = kwargs['xy']
        return self.annotate(*args, **kwargs)

This way the transform keyword is used to properly construct the xycoords transform.

Example:

ax = plt.axes(projection=ccrs.PlateCarree())
ax.stock_img()

ny_lon, ny_lat = -75, 43
delhi_lon, delhi_lat = 77.23, 28.61

ax.plot(
    [ny_lon, delhi_lon], [ny_lat, delhi_lat], '-bo', transform=ccrs.Geodetic(),
)

ax.annotate('New York', xy=(ny_lon, ny_lat), xytext=(-30, +15),
         textcoords='offset points', ha='center', va='center',
         arrowprops=dict(arrowstyle='->'), 
         transform=ccrs.Geodetic())

ax.annotate('Delhi', xy=(delhi_lon, delhi_lat), xytext=(+30, -15),
         textcoords='offset points', ha='center', va='center',
         arrowprops=dict(arrowstyle='->'), 
         transform=ccrs.Geodetic())

[QuLogic]

Seems pretty reasonable, though you have to be careful to only do so if the coords are really for data space; would you like to open a PR?

[shaharkadmiel]

Thanks @QuLogic , I can make a pull request.
Indeed, I will have to first check that xycoords is either 'data' or unset (the default is 'data') and only then assign the transform to it.