featuresAt with radius 0/not defined only returns Polygons

[lbud]

…bug?
Can only find LineString/Points if radius >= 1.

[lbud]

https://github.com/mapbox/mapbox-gl-js/blob/mb-pages/js/data/feature_tree.js#L103-L116
https://github.com/mapbox/mapbox-gl-js/blob/mb-pages/js/data/feature_tree.js#L136-L145

[lbud]

The above two functions (lineContainsPoint, pointContainsPoint) leave absolutely zero tolerance for what seem to be imperfect computations (see #874 — it's impossible to mouse over a Point or LineString and see a result). Should we add a tiny bit of tolerance (i.e. change this line to

    if (distToSegmentSquared(p, v, w) < (r || 1)) return true;

and similarly in the point calculation)? Or is that…cheating?
@jfirebaugh

[jfirebaugh]

Well radius is supposed to provide that tolerance, right?

See #316 for relevant discussion -- for many/most uses cases of featuresAt, I think what we really want is paint-property-aware hit testing.

[lbud]

Well radius is supposed to provide that tolerance, right?

I suppose so, yes, but it's weird that it's only possible to find Polygon features without providing a radius. It's not even that it's very hard to click on an exact LineString/Point; it's literally impossible to do so because of these calculations. The assumption would be that with radius: 0 you should be able to find a feature on exact click/hover.

[lbud]

maybe even a Math.floor would be an ok answer here — I've logged the calculations at least from lineContainsPoint and when mousing over a line I've been able to log something like 0.16