d3.geo.path gives wrong values for bounds

[theevann]

I wonder if this issue may be linked with the bug #2025.

I'm plotting a map using geojson data and d3.geo. It appears that for some of the objects, the function bounds() returns wrong bounds.

Here is an example:

    var proj = d3.geo.mercator()
        .center([-0.14, 51.5])
        .scale(20000);

    var path = d3.geo.path().projection(proj);

    var obj = {
        geometry : {
            coordinates : [[[0.6673, 51.4574], [0.6759, 51.4583], [0.6815, 51.4578], [0.6815, 51.4578], [0.6873, 51.4595], [0.6673, 51.4574]]],
            type : "Polygon"
        },
        type : "Feature",
        id : "236404706",
        clipped : true,
        properties : {
            area: 118107,
            kind: "water",
            source: "openstreetmap.org"
        }
    };

    path.bounds(obj);

The bounds given are: [-62302.98385274002, -62581.853071795864] and [63311.853071795864, 63081.853071808815]
which are not correct.

With, for example, the coordinates : [[[0.6673, 51.4574], [0.6759, 51.4583], [0.6815, 51.4578], [0.6873, 51.4595], [0.6673, 51.4574]]], the bug doesn't appear.

[theevann]

Update :
Then, I wondered if that may be linked to the polygon being taken in the wrong way (problem probably linked with clockwise / antoclokwise way). So I reversed the coordinates and I got the right answer :

    obj.geometry.coordinates[0].reverse()
    path.bounds(obj);

which gives me the right answer...

[jasondavies]

d3.geo requires polygons to wind clockwise around their interior (right-hand/right-foot winding rule). Your polygon looked like it might be degenerate (self-intersecting) at first glance (I removed the coincident point for clarity):

If you zoom in, you realise there is no degeneracy, therefore the problem was simply the incorrect winding order. The numbers from 0 to 3 indicate the order, and you can see that the winding order is counterclockwise relative to the small interior.