v3 polylines truncate "leaf nodes" in routes.

[Robert Lipe]

Hello, Google Map Gurus.

I'd sort of fallen off the boat of the API and wound up implementing a
Map that that relied on Polylines before the blog post announced they
were added. I'd been beating my head on the wall trying to figure
out why I was losing key vertices in a polyline, but I'm now
suspecting the problem isn't entirely mine.


I have a polyline that consists of almost 800 points. (From v1 api, I
know this is too many, but the map in question needs to be the
"before" in a "before and after" comparison once we simplify the
points. To Google's credit, 800 points in a polyline doesn't melt
the browser like it did back in the Old Days.) These are points
gathered from a GPS on a school bus in a rural area. School busses
in rural areas drive to where the kids are and they often turn
around.

I'm seeing a large number of these "turnaround" roads shortened or
dropped completely when I pull them into the map. I've put a screen
capture of an affected area up temporarily at http://www.gpsbabel.org/tmp/Picture%2055.png

The code in question is blindingly dumb. It's just a plain array of
vertices:

var pathCoords = [[30.705786, -96.374112],
[30.706147, -96.374148],
[30.706077, -96.374136],
[30.706073, -96.374152],
[30.706065, -96.374184],
[30.705878, -96.374093],
[ continues for completely too long ]

and I just added the markers (cut and pasted from the examples) to
"prove" that the GPS tracks captured what I thought they captured.
The key function is the pretty obvious:

function loadup(a, path, map) {
for (i = 0; i < a.length; i++) {
lat = a[i][0];
lng = a[i][1];
pnt = new google.maps.LatLng(lat, lng);
path.insertAt(path.length, pnt)
// for debugging, add a marker.
var marker = new google.maps.Marker({
position: pnt,
map: map,
title:"Hello World!"
});
bbox.extend(pnt);
}
}

I suspect that the polyline code is simplifying the polyline and not
accounting for the "doubleback" case when computing maximum
crosstrack error. It may be data specific, but I have several
occurrences of it in only two GPS tracks, so it's not like it's
extremely hard to trigger on this type of data.

If a Google developer hasn't branded me a bozo at this point and isn't
smacking his or her (potentially multicolored) forehead at this point
and needs to see the full Javascript with data, please contact me
privately.

[Ben Appleton]

Please do publish a failing page and link here. There's little we can
do without seeing the bug.