This is a bugfix for issue #4999.

It follows a tightening up in numpy 1.10 when indexing one array with a second boolean array which is smaller, such as

x = np.asarray([1.0, 2.0, 3.0, 4.0])
index_array = np.asarray([False, True])
y = x[index_array]

Here index_array is shorter than x. Prior to numpy 1.10 this was acceptable, with index_array padded out with False values. In numpy 1.10 it issues a VisibleDeprecationWarning.

The problem occurs in TriAnalyzer for a Triangulation that contains points that are not referenced by any triangles, which can occur if triangles are masked out. The solution is to make sure the boolean array is of the correct size, which is easily accomplished in this case using the minlength kwarg which was added to numpy.bincount in version 1.6.

I have not added an explicit test as the existing test_triangulation.test_tritools already finds the bug.

Thanks to @WeatherGod for reporting the problem.