DataLink involving a DynamicMap

[DancingQuanta]

I want to view a image of 3D data and use a PointDraw to mark pixels of the image.
With @philippjfr help, I am able to make the points snap to the center of a pixel. The points is created by a DynamicMap returning hv.Points.
Finally I wanted to view the coordinates of the points in a hv.Table. To do this, I used a DataLink to link the dynamic Points to the table.

However, the DataLink appears not to work. The table did not update upon adding new points.

Full example code

spatial_dmap = hv.DynamicMap(lambda X: hv.Image(np.random.rand(10, 10)), kdims=['X']).redim.values(X=[1, 2, 3])
stream = hv.streams.PointDraw(data={'x': [], 'y': []})

def snap_points(data):
    key = spatial_dmap.callback.args or tuple(d.values[0] for d in spatial_dmap.kdims)
    img = spatial_dmap[key]
    closest = img.closest(list(zip(data['x'], data['y'])))
    return hv.Points(closest)

points = hv.DynamicMap(snap_points, streams=[stream])

table = hv.Table(points.data, ['x', 'y'])
DataLink(points, table)

(
    spatial_dmap
    * points
    + table
)

Here is another example code that does not snap points to the center of the pixels. This example code works where the table do get updated with new points.

spatial_dmap = hv.DynamicMap(lambda X: hv.Image(np.random.rand(10, 10)), kdims=['X']).redim.values(X=[1, 2, 3])

points = hv.Points([])
point_stream = streams.PointDraw(source=points)

table = hv.Table(points.data, ['x', 'y'])
DataLink(points, table)

(
    spatial_map
    * points
    + table
)