Add bokeh.plotting.gmap to create GMapPlot correctly and easily

[ahamiltonian]

When using GMapPlot with DataRange1d, the points plotted do not appear in the correct location when the data is updated. See example code below. The solution is to use Range1d. It would be helpful to give a validation error so the developer can be advised to use Range1d instead of DataRange1d.

Thanks,
Andrew

from bokeh.io import output_file, show
from bokeh.models import GMapPlot, GMapOptions, ColumnDataSource, Circle, DataRange1d, WheelZoomTool, Select
from bokeh.plotting import curdoc
from bokeh.layouts import layout, row, widgetbox

# set up map / plot  (my google maps API key: AIzaSyAh1cmN_U6smbyt5uCpMIhttQOYrE4JtLY)
map_options = GMapOptions(lat=49.74, lng=-123.117, map_type="roadmap", zoom=14)
plot = GMapPlot(
    x_range=DataRange1d(), y_range=DataRange1d(), map_options=map_options,
    api_key='PUT YOUR GOOGLE MAPS API KEY HERE'
)

c=['red','green','blue']

# create an initial source object (one color and size on each point)
source = ColumnDataSource( data=dict(   lat = (49.746504,49.735954,49.744864),
                                        lon = (-123.117835,-123.114313,-123.114587),
                                        color = c ) )

# set the circle parameters from the source elements
circle = Circle(x="lon", y="lat", fill_color="color", size=10,  fill_alpha=0.5, line_color=None)

# add the source to the plot as circles
plot.add_glyph(source, circle)

# throw on some tools
plot.add_tools( WheelZoomTool() )


def update(attrname, old, new):

    if color.value == 'RBG': c=['red','green','blue']
    if color.value == 'Greyscale': c=['black','grey','white']
    if color.value == 'CMY': c=['cyan','magenta','yellow']

    # create the source object with the data to be plotted
    source.data=dict(   lat = (49.746504,49.735954,49.744864),
                        lon = (-123.117835,-123.114313,-123.114587),
                        color=c)

    # update the plot
    plot.update()


# add the selectors
color = Select(title='Color', value='None', options=['RBG','Greyscale','CMY'])
color.on_change('value', update)

# add the controls
controls = widgetbox([color], width=200)
layout = row(controls, plot)

curdoc().add_root(layout)
curdoc().title = "My Google Map"

[ahamiltonian]

Sorry for making a mess of the code, I'm new to posting to GitHub. The 'Insert Code' <> button gave me the mess above (I should have previewed...)

[bryevdv]

No worries, I fixed it up :) FYI code quoting on GH is triple-backticks (on their own lines before/after)

[bryevdv]

@bokeh/dev thoughts? This could be a validation warning, or we could fail harder earlier with an exception... it really doesn't make sense to ever have a DataRange with a GMap, since we are not in the driver's seat. GMaps insist on controlling the ranges directly, and we respond to those changes.

[mattpap]

If it doesn't make sense, don't allow it. There is asserts() method on Property for this task.

[bryevdv]

@mattpap Well, the base Plot just has this:

    x_range = Instance(Range, help="""
    The (default) data range of the horizontal dimension of the plot.
    """)

    y_range = Instance(Range, help="""
    The (default) data range of the vertical dimension of the plot.
    """)

So we'd have to factor everything except ranges out into a PlotBase or something, so that GMapPlot can define those differently. I'm not opposed to that, it's just a chunk of work.

[mattpap]

Not necessarily, GMapPlot could override those properties. This way we don't even need asserts(). Just I doubt this is sound under OOP (edit: given that models are mutable).

[bryevdv]

Possible solution: extend Override functionilty

[bryevdv]

Going further, we can also make a GMapRange1d that has only read-only start and end