Support Mercator.GOOGLE and cartopy.crs.epsg() projections

[sanzoghenzo]

I'm trying to add a basemap in the same fashion as #48.

I'm using GeoPandas data (gdf1 to gdfX) with local crs to compose a map and add a background map from web tiles.

    # ...
    import geoplot as gplt

    proj = gplt.crs.Mercator()
    gdf1 = gdf1.to_crs(epsg=4326)
    ax = gplt.pointplot(gdf1, projection=proj)
    lims = [ax.axis()]
    for gdf in gdf2:
        ax = gplt.choropleth(gdf.to_crs(epsg=4326), hue='vul_max', 
                             cmap='RdYlGn_r',  alpha=0.5, ax=ax, projection=proj)
        lims.append(ax.axis())
    # ... some other layers added ...
    ws, es, ss, ns = zip(*lims)
    ax.axis((min(ws), max(es), min(ss), max(ns)))
    ax = add_basemap(ax)

def add_basemap(ax, webmap='ST_TONER_LITE'):
    import contextily as ctx
    url = getattr(ctx.sources, webmap, ctx.sources.ST_TONER_LITE)
    w, e, s, n = ax.axis()
    img, ext = ctx.bounds2img(w, s, e, n, url=url)  # , ll=True)
    ax.imshow(img, extent=ext, interpolation='bilinear', zorder=-1)
    ax.axis((w, e, s, n))
    return ax

The problem is with Mercator() that centers the coordinate system to the center of the plot and messes up the boundaries (ax.axis returns "garbage").
Since we're talking about webtiles, that always use EPSG 3857, there should be a way to directly specify that projection.

Maybe you already know, but I found that cartopy has the Mercator.GOOGLE projection that could be used for this goal.

Using directly that class doesn't work because of the "load" method missing.
Manually passing the parameters works:

proj = gplt.crs.Mercator(central_longitude=0, 
                         min_latitude=-85.0511287798066,
                         max_latitude=85.0511287798066,
                         globe=Globe(ellipse=None,
                                     semimajor_axis=WGS84_SEMIMAJOR_AXIS,
                                     semiminor_axis=WGS84_SEMIMAJOR_AXIS,
                                     nadgrids='@null'))

It would be nice to have a convenience class for this directly in geoplot.

There's also a cartopy.crs.epsg() function to create the projection from an arbitrary EPSG (I'm thinking about outputting the maps in the original CRS), supporting it could open to many possibilities.

[ResidentMario]

Hmm. This library predates contextily, and I'm not sure what the best pattern for getting geoplot/geopandas/cartopy/contextily to play nice with one another is. There's even a PR perma-open against this library adding contextily support.

I can naively add support for the Google projection, probably. It irks me that cartopy doesn't expose this projection in their list in the documentation.

[ResidentMario]

Hi, closing this issue now because I am tracking these feature requests in #146 and #147. Thanks for reporting! I don't have a timeline yet for these things, but they're pretty high on the list of priorities for after this next geoplot@0.3.0 release.

[ResidentMario]

A special gcrs.WebMercator projection has now been added to geoplot (#169) that provides an EPSG:3857 projection, which addresses your original request. Support for arbitrary EPSG coordinate reference systems is still on the to-do list. If making webmaps is your goal, there is now a gplt.webmap plot type that does just that, using the WebMercator projection. 🎉